Psychophysiology 47(6)

Psychophysiology, 47 (2010), 995–1001. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01017.x

How personal earthquake experience impacts on the Stroop interference effect: An event-related potential study

JIANG QIU, YANHUA SU, HONG LI, DONGTAO WEI, SHEN TU, and QINGLIN ZHANG Key Laboratory of Cognition and Personality (SWU), Ministry of Education, Chongqing, China, and School of Psychology, Southwest University, Chongqing, China

Abstract Event-related brain potentials (ERPs) were measured when 24 Chinese subjects performed the classical Stroop task. All of subjects had experienced the great Sichuan earthquake (5/12), with 12 people in each of the Far (Chengdu city) and the Close (Deyang city) earthquake experience groups. The behavioral data showed that the Stroop task yielded a robust Stroop interference effect as indexed by longer RT for incongruent than congruent color words in both the Chengdu and Deyang groups. Scalp ERP data showed that incongruent stimuli elicited a more negative ERP deflection (N400–600; Stroop interference effect) than did congruent stimuli between 400–600 ms in the Chengdu group, while the Stroop interference ERP effect was not found in the Deyang group. Dipole source analysis localized the generator of the N400–600 in the right prefrontal cortex (PFC) and was possibly related to conflict monitoring and cognitive control. Descriptors: Stroop effect, Cognitive control, The great Sichuan earthquake, Event-related brain potentials (ERPs)

posttraumatic stress disorder (PTSD), with some evidence that this activation is localized to subregions of the amygdala and is associated with alterations in medial prefrontal activity. Previous studies have shown that control mechanisms play an important role in the organization of action and thought, and the medial frontal cortex/anterior cingulate cortex (ACC) and the prefrontal cortex (PFC) play a critical role in the central executive control system (e.g., Botvinick, Braver, Carter, & Barch, 2001; Carter, Braver, Barch, Botvinick, Noll, & Cohen, 1998; MacDonald, Cohen, Stenger, & Carter, 2000; Milham, Banich, & Barad, 2003; Milham & Banich, 2005). Some classical cognitive control tasks (e.g., the Stroop task) are often selected as experimental paradigms to explore the nature of automatic and controlled cognitive processes. The Stroop interference effect refers to an increase in response time observed when the word meaning and the stimulus hue do not match (Stroop, 1935). Results of positron emission tomography (PET) and fMRI indicated that the area of the brain that showed the most robust activation during performance of the Stroop color naming task was the ACC, including the ventral subgenual, rostral, and dorsal regions (e.g., MacDonald et al., 2000; Milham et al., 2003). Additional areas of the brain that showed increased activation included the PFC, motor cortex, inferior temporal gyrus, and superior and inferior parietal cortex (e.g., Ardi & Peter, 2002; Carter, Mintun, & Cohen, 1995). Measurement of event-related potentials (ERPs) provides better temporal resolution of neural activity than do PET and fMRI. For example, West and Alain (1999) found that a greater

After the great Sichuan earthquake (China) on May 12 (5/12), 2008, many people had long-term problems. These included obsession with the trauma, nightmares, flashbacks, emotional numbing, loss of interest in life, irritability, memory problems, and hypervigilance. After a simulation, Basoglu, Salcioglu, and Livanou (2007) predicted that many people post-earthquake would find their fear of earthquakes interfering with everyday activities, including sleeping, bathing and even walking into a building. Therefore, the events of the great Sichuan earthquake, and the experiences of the people who were in close proximity to that disaster (e.g., Deyang city), provided a unique window into the neural correlates of the influences of earthquake stress exposure on their cognitive and brain functions. For example, many functional magnetic resonance imaging (fMRI) studies (e.g., Armony, Corbo, Clement, & Brunet, 2005; Rauch, Whalen, Shin, McInerney, Macklin, & Lasko, 2000; Sharot, Martorella, Delgado, & Phelps, 2007; Shin, Wright, Cannistraro, Wedig, McMullin, et al., 2005; Williams, Kemp, Felmingham, Barton, Olivieri, et al., 2006) find that increased activity in the amygdala is the most consistently observed neural correlate of

This research was supported by the Specialized Research Fund for the Doctoral Program of Higher Education of China (200806351002), the National Natural Science Foundation of China (30800293), and the Key Discipline Fund of National 211 Project (NSKD08005). Address reprint requests to: Jiang Qiu, School of Psychology, Southwest University, Beibei, Chongqing 400715 China. E-mail: qiuj318@ swu.edu.cn 995

996 positive wave (P500) was evoked in the incongruent condition over the fronto-polar region of the brain and, within 500–1000 ms, a greater negative wave was evoked over the fronto-central region. The authors concluded that these components probably reflected ‘‘conflict discovery and conflict resolution processes’’ (West & Alain, 1999). Liotti, Woldorff, Perez, and Mayberg (2000) found that a negative wave appeared over the medial dorsal region between 350–500 ms post-stimulus, with the peak at 410 ms. Dipole source analysis suggested that an independent generator present in the ACC was mainly related to conflict discovery and conflict resolution. Moreover, Markela-Lerenc, Ille, Kaiser, Fiedler, Mundt, and Weisbrod (2004) found greater negativity in the incongruent than in the congruent condition over left fronto-central scalp regions in a time frame of 350 and 450 ms after the stimulus presentation. Immediately after this first effect (between 450 and 550 ms after stimulus presentation), a greater positive potential developed over midline fronto-central scalp regions in the incongruent than in the congruent condition. Recently, Badzakova-Trajkov, Barnett, Waldie, and Kirk (2009) also recorded high density visual-evoked potentials from 16 healthy subjects while performing a manual version of the Stroop task, and source estimation indicated that the ACC underlay the difference waveform for the incongruent-congruent contrast (in the time window of 370–480 ms). They suggested that the peak in the incongruent-congruent difference wave likely reflected identification (and subsequent resolution) of conflict in the two sources of information (the word and the color it is printed in; Badzakova-Trajkov et al., 2009). In Qiu, Luo, Wang, and Zhang’s study (2006), ERPs were measured while 11 healthy Chinese subjects performed the traditional Stroop color-word task. Their results showed that a greater negativity in the incongruent compared to the congruent condition was found between 350 and 550 ms post-stimulus over midline fronto-central scalp regions, and dipole source analysis of the difference wave (incongruent–congruent) indicated that a generator localized in the PFC, that was possibly related to conflict processing and response selection (Qiu et al., 2006). Moreover, the Stroop task and many modified versions of Stroop tasks have been utilized by researchers to explore the nature of automatic and controlled cognitive processes and disturbances in cognition resulting from various psychiatric and/or neurological disorders. Previous work had provided evidence that the effects of drug-addiction on the central nervous system may extend from conscious to unconscious processes in cognition (e.g., Fehr, Wiedenmann, & Herrmann, 2006; Goldstein, Tomasi, Rajaram, Cottone, Zhang, et al., 2007; Polo, Escera, Yago, Alho, Gual, & Grau, 2003). In addition, some previous studies had used distance from an earthquake epicenter as a way to sample populations with differing levels of earthquake exposure (Goenjian, Yehuda, Pynoos, Steinberg, Tashjian, et al., 1996; Wang, Zhang, Wang, Shi, Shen, Li, & Xin, 2009). Therefore, in the present study, 24 Chinese people (two community samples) who had experienced the great Sichuan earthquake (5/12) were selected as our subjects, with 12 people in each of the Far (Chengdu city) and the Close (Deyang city) earthquake experience groups. We used the Stroop task as the experimental paradigm to test whether the neurophysiological substrate of Stroop interference is similar to previous findings (e.g., Qiu et al., 2006), and to explore how personal earthquake experience (Far and Close earthquake experience) impacts on their cognitive control. Some previous studies also indicated that there were stress-related decreases in gray matter volume in the ACC and PFC in

J. Qiu et al. nonclinical samples (e.g., Ganzel, Kim, Glover, & Temple, 2008; Liston, Miller, Goldwater, Radley, Rocher, et al., 2006) and stress-related alterations in medial PFC function (Liston, McEwen, & Casey, 2009). Thus, we hypothesized that negative life events might have significant influence on cognitive and brain functions in people of the community sample, and the anterior systems of executive control focused on the PFC and ACC that monitor and control cognitive conflict would be deactivated in the Deyang group. Specifically, we predicted that the Stroop interference effect of longer reaction times in the color-word condition could be observed in the Chengdu group, and this interference ERP effect would be related to a fronto-central relative negativity in the incongruent compared to the congruent condition. However, the Stroop interference effect might not be found in the Deyang group.

Materials and Methods Subjects Approximately 2 months after the Sichuan earthquake, twentyfour healthy college students from Deyang city (Close earthquake experience group: 6 women, 6 men; aged 21–26 years; mean age, 23.8 years) and Chengdu city (Far earthquake experience group: 6 women, 6 men; aged 20–26 years; mean age, 22.4 years) in China participated in the experiment as paid volunteers. We know that Deyang is one of three major cities immediately surrounding the earthquake’s epicenter (Wenchuan, approximately 60 miles). There were about 4 million people; many encountered severe casualties and the current death toll stands at 6,000. However, in Chengdu city, 150 km away from the source, there were only cracks in the walls of some residential buildings in the downtown areas and no buildings collapsed. We obtained appropriate ethics committee approval for the research, and all subjects gave written informed consent. All subjects were righthanded and had no current or past neurological or psychiatric illness (i.e., instructors confirmed that they had no abnormal behavioral or psychological phenomena in the past by checking their entrance psychological archives), and all had normal or corrected-to-normal vision. Stimuli and Procedure The experiment consisted of congruent and incongruent stimuli. Congruent stimuli consisted of the four color words [ , , , (red, yellow, green, blue)] written in the same color in which the stimulus was presented (e.g., the word (red) written in red color). Incongruent stimuli consisted of the same four words with the display color not matching the word meaning (e.g., the word (green) written in red ink). Each incongruent stimulus appeared in each of the three colors not matching its meaning. The size of the Chinese characters was Song Ti No. 28 (1.41 (horizontal) ! 1.41 (vertical), and was displayed in the center of a 17-inch screen at random. Subjects were seated in a semidark room facing a monitor placed at 80 cm distance from the eyes. They were instructed to rest their left middle, left index, right index, and right middle finger on the appropriate color button (red 5 s key, yellow 5 d key, green 5 j key, and blue 5 k key) on the keyboard. They were told that a gray cross would always appear first in the center of the screen serving as a fixation point, and then one word written in different colors. The order was as follows: the fixation point

The Stroop interference effect appeared for 300–600 ms, the word appeared for 150 ms, then the empty screen appeared for 2000 ms. Subjects were asked to identify the color in which the stimulus was written as fast and accurately as possible and responded by pressing the button of the corresponding color. The experiment was divided into a practice phase and a test phase. The practice phase was designed to rehearse the mapping of colors onto fingers and pressing of the response buttons. When the accuracy rate got to 85%, the practice phase was ended. The formal test consisted of three blocks, and each block had 48 stimuli (24 congruent, 24 incongruent). Subjects were instructed to avoid blinking and eye movement of any sort and to keep their eyes fixated on the monitor rather than looking down at their fingers during task performance. They were allowed to rest at the end of each block. Electrophysiological Recording and Analysis Brain electrical activity was recorded from 64 scalp sites using tin electrodes mounted in an elastic cap (Brain Products GmbH, Gilching, Germany), with the reference on the left and right mastoids. The vertical electrooculogram (EOG) was recorded with electrodes placed above and below the left eye. All interelectrode impedances were maintained below 5 kO. The electroencephalogram (EEG) and EOG were amplified using a 0.05– 100 Hz bandpass and continuously sampled at 500Hz/channel for off-line analysis. Eye movement artifacts (blinks and eye movements) were rejected offline by using the Gratton and Coles algorithm (Brain Vision Analyzer, Version, 1.05, Software, Brain Product GmbH), which corrects ocular artifacts by subtracting the voltages of the eye channels, multiplied by a channel-dependent correction factor, from the respective EEG channels (Gratton, Coles, & Donchin, 1983). Trials with EOG artifacts (mean EOG voltage exceeding " 80 mV) and those contaminated with artifacts due to amplifier clipping, bursts of electromyographic activity, or peak-to-peak deflection exceeding " 80 mV were excluded from averaging. The averaged epoch for ERP was 700 ms including 600 ms poststimulus and 100 ms prestimulus. Only segments with correct responses were averaged, and at least 30 trials were available for each subject and condition. On the basis of the grand averaged ERP waveforms and topographical maps, the following 9 electrode points were chosen for statistical analysis: F3, Fz, F4, C3, Cz, C4, P3, Pz, P4. Latencies and amplitudes (baseline to peak) of the N1, P2, N2, and P3 were measured separately in the 80–130 ms, 130–200 ms, 200–300 ms, and 300–400 ms time windows, respectively. Mean amplitudes in the time window of 400–600 ms was analyzed using three-way repeated-measures analysis of variance (ANOVA). The ANOVA within-subjects factors were stimulus type (Congruent/Incongruent) and electrode site. Earthquake experience (Deyang city and Chongqing city) was a between-subjects factor. For all analyses, p-values were corrected for sphericity assumption violations using the Greenhouse-Geisser correction. Dipole Source Analysis Brain Electrical Source Analysis (BESA, Version 5.0, MEGIS Software GmbH, Gra¨felfing, Germany) was used to perform dipole source analysis. For dipole source analysis, the four-shell ellipsoidal head model was used. In order to focus on the scalp electrical activity related to the processing of the Stroop interference effect, the averaged ERPs evoked by congruent stimuli were subtracted from the ERPs evoked by incongruent stimuli. Principal component analysis (PCA) was employed in the inter-

997 val from 400 to 600 ms in order to estimate the minimum number of dipoles. Results Behavioral Data In the mean reaction time analysis, a very robust Stroop colorword interference effect was obtained as indicated by longer mean reaction times (RTs) for incongruent than congruent colorword presentations (F(1,22) 5 34.5, p 5 .000o.001) in both the Chengdu and Deyang groups. The mean RTs were 664 " 143 ms for Incongruent and 587 " 90 ms for Congruent in the Deyang group, and 756 " 173 ms for Incongruent and 632 " 109 ms for Congruent in the Chengdu group. However, the interaction between stimulus type ! earthquake experience was not significant [F(1,22) 5 1.91, p 5 .184.05], and the main effect of earthquake experience on reaction time was not significant [F(1,22) 5 1.75, p 5 .194.05]. For the accuracy data, stimulus type was also significant (F(1,22) 5 17.29, po.05). Post hoc tests revealed significantly more errors for the incongruent than the congruent condition. The accuracy rates for incongruent and congruent were 87.2 " 2.0% and 75.9 " 1.6%, respectively, in the Deyang group, and 92.1 " 1.1% and 84.1 " 1.4%, respectively, in the Chengdu group. Again, the interaction between stimulus type ! earthquake experience was not significant [F(1,22) 5 0.51, p 5 .484.05], and the main effect of earthquake experience on accuracy was also not significant [F(1,22) 5 1.13, p 5 .294.05]. Electrophysiological Scalp Data As shown in Figures 1 and 2, the N1, P2, N2, and P3 were elicited by two conditions. The results of the ANOVAs showed that the interaction between stimulus type ! earthquake experience was not significant [F(1,22) 5 0.09, p 5 .764.05; F(1,22) 5 1.69, p 5 .204.05; F(1,22) 5 0.51, p 5 .484.05; F(1,22) 5 0.38, p 5 .544.05], and there were no main effects of stimulus type for amplitudes of these components. However, we found that there was a significant earthquake experience effect [F(1,22) 5 5.79, p 5 .02o.05] on N1 amplitude. Results showed that both conditions in the Deyang group elicited a much more negative deflection (N1) than they did that in the Chengdu group. Between 400–600 ms, there was no main effect of stimulus type or earthquake experience for the mean amplitudes, F(1,22) 5 3.12, p 5 .094.05; F(1,22) 5 0.42, p 5 .524.05. However, there was a significant stimulus type ! earthquake experience effect [F(1,22) 5 5.44, p 5 .02o.05]. The results of a simple effects test showed that incongruent stimuli elicited a more negative ERP deflection (N400–600) than did congruent stimuli (Incongruent: 6.54 " 1.13 mV; Congruent: 8.92 " 1.27 mV) in the Chengdu group, but no difference between incongruent and congruent stimuli in the Deyang group (Incongruent: 6.71 " 0.98 mV; Congruent: 6.45 " 1.35 mV). Dipole Source Analysis The source analysis using BESA software was performed on the ERP difference wave of incongruent and congruent conditions in the Chengdu group. PCA was employed in the 400–600 ms time window. PCA indicated that two components were needed to explain 89.8% and 7.0% of the variance in the data. Therefore, two dipoles were fitted with no restriction to the direction and location of dipoles. The result (see Figure 3) indicated that the first dipole was located approximately in the right prefrontal

998

J. Qiu et al.

Figure 1. Grand average ERP for the congruent and incongruent conditions at Fz and Cz in the Deyang and Chengdu city groups.

cortex (location according to Talairach coordinates: x 5 50.8, y 5 57.0, z 5 6.9; BA10), and the second located near the medial frontal cortex (x 5 # 5.9, y 5 # 25.2, z 5 73.1; BA6). This model explained the data best and accounted for most of the variance with a residual variance of 9.8% and revealed maximal dipoles moment strength at about 540 ms. The display of the residual maps showed no further dipolar activity, and no further dipoles could be fitted in the investigated time window.

Discussion In the present study, robust behavioral and electrophysiological effects of color-word interference were seen in the Chengdu group performing the Chinese characters Stroop task, supporting evidence for the universality of the Stroop interference effect. However, the Stroop interference ERP effect was not found in the Deyang group, which had greater direct personal experience of the great Sichuan earthquake. Incongruent stimuli elicited a more negative ERP deflection (N400–600) than did congruent stimuli between 400–600 ms post-stimulus over fronto-central scalp regions in the Chengdu group. Dipole source analysis suggested that the N400–600 was mainly generated in the right PFC. To some extent, the Stroop interference ERP effect (N400–600) in our study was similar to previous findings (e.g., Liotti et al., 2000; Markela-Lerenc et al., 2004; Qiu et al., 2006). For example, the results reported by

Liotti and Markela-Lerenc demonstrated two modulations of the ERPs that are consistently associated with conflict processing in the Stroop task: the N450 wave and the sustained potential (Liotti et al., 2000; Markela-Lerenc et al., 2004). The N450 peaks between 400 and 500 ms after stimulus onset and reflects a phasic fronto-central negativity that differentiates incongruent trials from congruent and neutral trials. This ERP component seems to be related to response detection and may arise from activity in prefrontal areas. The sustained potential is elicited about 500 ms after stimulus onset and reflects a sustained parietal positivitylateral frontal negativity. In a similar study, Qiu et al. (2006) also found that incongruent condition elicited a much more negative deflection (N450) than did congruent stimuli from 350–550 ms. Dipole analysis showed that the N450 was generated in the PFC, which was probably related to conflict processing and response selection. In addition, many studies (e.g., Kern, Cohen, Stenger, & Carter, 2004; Peterson, Skudlarski, Gatenby, Zhang, Anderson, & Gord, 1999) had reported activation of the PFC and ACC in the Stroop interference color-word task. For example, Markela-Lerenc et al. (2004) found that the first Stroop-related effect originated from activity generated in the inferior PFC between 350 ms and 450 ms. Many research findings also indicate that PFC activity might be mainly related to conflict processing and be required for evaluation execution control, while ACC activity was mainly related to conflict detection and supervision (Badzakova-Trajkov et al., 2009; Bune, Hazelting, & Scanion, 2002; Rowe, Toni, & Josephs, 2000). Therefore, during performance of

Figure 2. Left: Grand average ERP for the congruent and incongruent conditions and the difference wave (Incongruent-Congruent) at Cz in the Chengdu group. Right: Topographical maps of the voltage amplitudes for the incongruent vs. congruent condition difference wave in the 450 ms and 550 ms.

The Stroop interference effect

Figure 3. Results of the dipole source analysis of the difference wave for the Chengdu group (Incongruent vs. Congruent) in the time range of 400–600 ms. The left-bottom shows the source activity waveforms, whereas the right figure displays the mean locations of the dipole. The first dipole is located approximately in the right prefrontal cortex (x 5 50.8, y 5 57.0, z 5 6.9), the second dipole near the medial frontal cortex (x 5 # 5.9, y 5 # 25.2, z 5 73.1).

the Stroop task of color naming of Chinese characters for the Chengdu group in our study, despite an intentional attempt on the subject’s part to attend to the color of words, the information of color and meaning would inevitably bring about conflict in the incongruent condition, due to the automation of the processing of word meaning. Then, activation of the right PFC might be required to control irrelevant information inference (semantic information) and make the right judgment according to the information of color. Most important and interesting, our results showed that the two conditions in the Deyang group both elicited a much more negative deflection (N1) than they did in the Chengdu group, and the Stroop interference ERP effect (N400–600) was not found in the Deyang group, which had greater direct personal experience of the great Sichuan earthquake. Previous studies had shown that the N1 was probably related to orienting of attention, feature selection, and stimulus evaluation in the visual tasks (e.g., Hillyard & Anllo-Vento, 1998; Mangun, 1998; Naatanen, Paavilainen, & Reinikainen, 1989). In Han, Yund, and Woods’ study (2003), they found that the N1 (180–200 ms) was of larger amplitude to the global than local targets, and thought that the enlarged N1 might reflect a general mechanism involved in discrimination processes based on different visual feature dimensions (e.g., Ritter, Simson, & Vaughan, 1983; Vogel & Luck, 2000). In our study, the increase in N1 amplitude in the Deyang group might indicate that these subjects needed to pay more attention to form an internal physical representation of the stimulus and to select stimulus feature related to task requirement. That is, we thought that the close personal earthquake experiences had made these subjects sensitive to physical aspects of any external stimuli (e.g., the incongruent and congruent stimuli) and the changing objective environment. This is consistent with other studies showing abnormalities in early stimulus processing in individuals with PTSD or schizophrenia (Louchart-de la Chapelle, Nkam, Houy, Belmont, Me´nard, et al., 2005; Morgan & Grillon, 1999). Previous work had reported decreased blood flow in the medial PFC in veterans with PTSD during reminder exposure tasks (Bremner, Staib, Kaloupek, Southwick, Soufer, & Charney,

999 1999). Shin and colleagues (2005) also found decreased activity in medial prefrontal regions in men with PTSD of long duration. Some studies had pointed out that the medial PFC, amygdala, and hippocampus may be especially vulnerable (Kanagaratnam & Asbjornsen, 2007; McEwen, 2005; Mitra, Jadhav, McEwen, & Chattarji, 2005; Vyas, Mitra, Shankaranarayana Rao, & Chattarji, 2002). Ganzel, Kim, Glover, and Temple (2008) found that adults with closer proximity to the disaster (9/11) had lower gray matter volume in amygdala, hippocampus, insula, ACC, and medial PFC, with control for age, gender, and total gray matter volume. They indicated that trauma exposure plays a critical role in changes to both brain structure and function, even in nonclinical adult populations. Recently, in Liston et al.’s study (2009), twenty healthy adults, exposed to 1 month of psychosocial stress, were scanned while performing a PFC-dependent attention-shifting task, and results showed that psychosocial stress selectively impaired attentional control and disrupted functional connectivity within a fronto-parietal network (e.g., PFC function) that mediates attention shifts. In short, these previous studies indicated that there were stress-related decreases in gray matter volume in the ACC and PFC in nonclinical samples (e.g., Ganzel et al., 2008; Liston et al., 2006) and stress-related alterations in medial PFC function (Liston et al., 2009). We therefore suggested that the anterior systems of executive control focused on the PFC and the ACC that monitor and control cognitive conflict might be deactivated in the Deyang group. However, behavioral data showed that there was still a robust Stroop color-word interference effect indicated by longer mean RTs for incongruent than congruent stimuli in the Deyang group. This might indicate that despite the absence of N400–600 interference effect, subjects in the Deyang group were still able to distinguish the color of stimuli, and showed normal performance results for the Stroop task in the presence of abnormalities of brain function (e.g., the larger N1 deflection, the deactivation of the PFC). Recently, Lui, Huang, Chen, Tang, Zhang, et al. (2009) found that regional activity in fronto-limbic and striatal areas increased significantly and connectivity among limbic and striatal networks were attenuated in their participants who were healthy survivors within 25 days after the Wenchuan earthquake. That is, their results indicated that traumatic experiences affect not only regional function but also dynamic interactions within brain networks. Thus, we speculated that there might be other brain activities or other abnormal brain networks (e.g., the fronto-limbic and striatal areas) instead of activation of the PFC when they performed the Stroop task. Summarily, our results indicated that the negative life events might have significant influence on brain functions in people of the community sample. However, further studies should be done using both ERPs and fMRI to reveal possible compensatory mechanisms that allowed an intact behavioral Stroop response while the PFC might be deactivated in cognitive control for these people. As Lui et al. (2009) said, ‘‘Longitudinal studies of trauma survivors may provide further insight into how alterations in brain function evolve over time after severe trauma.’’ To summarize, this study investigated spatiotemporal patterns of brain activation when 24 Chinese subjects (two community samples) performed the classical Stroop task using scalp and dipole source analysis of ERPs. Scalp ERP data showed that there was a N400–600 Stroop interference effect in the Chengdu group, while it was not found in the Deyang group. Consistent with previous findings, our result indicated that the anterior systems of executive control focused on the PFC that monitor and

1000

J. Qiu et al.

control cognitive conflict might be deactivated in the subjects who had greater direct personal experiences of the great Sichuan earthquake (i.e., the Deyang group). Our findings suggest that negative life events (e.g., earthquake, hurricane) have significant influence on brain functions in people of this community sample.

That is, after disasters such as the great Sichuan earthquake, it might be important for government and psychologists to consider the known impact of negative life events on psychological distress and disorder in the overall population (Ganzel et al., 2007).

REFERENCES Ardi, R., & Peter, H. (2002). Control of language use: Cognitive modeling of the hemodynamics of Stroop task performance. Cognitive Brain Research, 15, 85–97. Armony, J. L., Corbo, V., Clement, M. H., & Brunet, A. (2005). Amygdala response in patients with acute PTSD to masked and unmasked emotional facial expressions. American Journal of Psychiatry, 162, 1961–1963. Badzakova-Trajkov, G., Barnett, K. J., Waldie, K. E., & Kirk, I. J. (2009). An ERP investigation of the Stroop task: The role of the cingulate in attentional allocation and conflict resolution. Brain Research, 1253, 139–148. Basoglu, M., Salcioglu, E., & Livanou, M. (2007). A randomised controlled study of single-session behavioural treatment of earthquakerelated posttraumatic stress disorder using an earthquake simulator. Psychological Medicine, 37, 203–213. Botvinick, M. M., Braver, T. S., Carter, C. S., & Barch, D. M. (2001). Conflict monitoring and cognitive control. Psychological Review, 108, 624–652. Bremner, J. D., Staib, L. H., Kaloupek, D., Southwick, S. M., Soufer, R., & Charney, D. S. (1999). Neural correlates of exposure to traumatic pictures and sound in Vietnam combat veterans with and without posttraumatic stress disorder: A positron emission tomography study. Biological Psychiatry, 45, 806–816. Bune, S. A., Hazelting, E., & Scanion, M. D. (2002). Dissociable contributions of prefrontal and parietal cortices to response selection. NeuroImage, 17, 1526–1571. Carter, C. S., Mintun, M., & Cohen, J. D. (1995). Interference and facilitation effects during selective attention: An H215O PET study of Stroop task performance. NeuroImage, 2, 264–272. Carter, C. S., Braver, T. S., Barch, D. M., Botvinick, M. M., Noll, D., & Cohen, J. D. (1998). Anterior cingulate cortex, error detection, and the online monitoring of performance. Science, 280, 747–749. Fehr, T., Wiedenmann, P., & Herrmann, M. (2006). Nicotine Stroop and addiction memoryFAn ERP study. International Journal of Psychophysiology, 62, 224–232. Ganzel, B. L., Casey, B. J., Glover, G. H., Voss, H. U., & Temple, E. (2007). The aftermath of 9/11: Effect of intensity and recency of trauma on outcome. Emotion, 7, 227–238. Ganzel, B. L., Kim, P., Glover, G. H., & Temple, E. (2008). Resilience after 9/11: Multimodal neuroimaging evidence for stress-related change in the healthy adult brain. NeuroImage, 40, 788–795. Goenjian, A. K., Yehuda, R., Pynoos, R. S., Steinberg, A. M., Tashjian, M., Yang, R. K., et al. (1996). Basal cortisol, dexamethasone suppression of cortisol, and MHPG in adolescents after the 1988 earthquake in Armenia. American Journal of Psychiatry, 153, 929–934. Goldstein, R. Z., Tomasi, D., Rajaram, S., Cottone, L. A., Zhang, L., Maloney, T., et al. (2007). Role of the anterior cingulate and medial orbitofrontal cortex in processing drug cues in cocaine addiction. Neuroscience, 144, 1153–1159. Gratton, G., Coles, M. G. H., & Donchin, E. (1983). A new method for off-line removal of ocular artifact. Electroencephalography and Clinical Neurophysiology, 55, 468–484. Han, S., Yund, E. W., & Woods, D. L. (2003). An ERP study of the global precedence effect: The role of spatial frequency. Clinical Neurophysiology, 114, 1850–1865. Hillyard, S. A., & Anllo-Vento, L. (1998). Event-related brain potentials in the study of visual selective attention. Proceedings of the National Academy of Sciences, USA, 95, 781–787. Kanagaratnam, P., & Asbjornsen, A. E. (2007). Executive deficits in chronic PTSD related to political violence. Journal of Anxiety Disorder, 21, 510–525. Kern, J. G., Cohen, J. D., Stenger, V. A., & Carter, C. S. (2004). Anterior cingulated conflict monitoring and adjustments in control. Science, 303, 1023–1026.

Liotti, M., Woldorff, M. G., Perez, R., & Mayberg, H. S. (2000). An ERP study of the temporal course of the Stroop colorFWord interference effect. Neuropsychologia, 38, 701–711. Liston, C., Miller, M. M., Goldwater, D. S., Radley, J. J., Rocher, A. B., Hof, P. R., et al. (2006). Stress-induced alterations in prefrontal cortical dendritic morphology predict selective impairments in perceptual attentional set-shifting. The Journal of Neuroscience, 26, 7870–7874. Liston, C., McEwen, B. S., & Casey, B. J. (2009). Psychosocial stress reversibly disrupts prefrontal processing and attentional control. Proceedings of the National Academy of Sciences, USA, 106, 912–917. Louchart-de la Chapelle, S., Nkam, I., Houy, E., Belmont, A., Me´nard, J. F., Roussignol, A. C., et al. (2005). A concordance study of three electrophysiological measures in schizophrenia. The American Journal of Psychiatry, 162, 466–474. Lui, S., Huang, X. Q., Chen, L., Tang, H. H., Zhang, T. J., Li, X. L., et al. (2009). High-field MRI reveals an acute impact on brain function in survivors of the magnitude 8.0 earthquake in China. Proceedings of the National Academy of Sciences, USA, 106, 15412–15417. Mangun, G. R. (1998). Sustained visual-spatial attention produces costs and benefits in response time and evoked neural activity. Neuropsychlologia, 36, 189–200. Markela-Lerenc, J., Ille, N., Kaiser, S., Fiedler, P., Mundt, C., & Weisbrod, M. (2004). Prefrontal-cingulate activation during executive control: Which comes first? Cognitive Brain Research, 18, 278–287. MacDonald, A. W., Cohen, J. D., Stenger, V. A., & Carter, C. S. (2000). Dissociating the role of the dorsolateral prefrontal and anterior cingulated cortex in cognitive control. Science, 288, 1835–1838. McEwen, B. S. (2005). Glucocorticoids, depression and mood disorders: Structural remodeling in the brain. Metabolism, 54, 20–23. Milham, M. P., Banich, M. T., & Barad, V. (2003). Competition for priority in processing increases prefrontal cortex’s involvement in topdown control: An event-related fMRI study of the Stroop task. Cognitive Brain Research, 17, 212–222. Milham, M. P., & Banich, M. T. (2005). Anterior cingulate cortex: An fMRI analysis of conflict specificity and functional differentiation. Human Brain Mapping, 25, 328–335. Mitra, R., Jadhav, S., McEwen, B. S., & Chattarji, S. (2005). Stress duration modulates the spatiotemporal patterns of spine formation in the basolateral amygdala. Proceedings of the National Academy of Sciences, USA, 102, 9371–9376. Morgan, C. A., & Grillon, C. (1999). Abnormal mismatch negativity in women with sexual assault-related posttraumatic stress disorder. Biological Psychiatry, 45, 827–832. Naatanen, R., Paavilainen, P., & Reinikainen, K. 1989. Do event-related potentials to infrequent decrements in duration of auditory stimuli demonstrate a memory trace in man? Neuroscience Letter, 107, 347– 352. Peterson, B. S., Skudlarski, P., Gatenby, J. C., Zhang, H., Anderson, A. W., & Gore, J. C. (1999). An fMRI study of Stroop wordcolor interference: Evidence for cingulate subregions subserving multiple distributed attentional systems. Biological Psychiatry, 45, 1237–1258. Polo, M. D., Escera, C., Yago, E., Alho, K., Gual, A., & Grau, C. (2003). Electrophysiological evidence of abnormal activation of the cerebral network of involuntary attention in alcoholism. Clinical Neurophysiology, 114, 134–146. Qiu, J., Luo, Y. J., Wang, Q. H., & Zhang, Q. L. (2006). Brain mechanism of Stroop interference effect in Chinese characters. Brain Research, 1072, 186–193. Rauch, S., Whalen, P., Shin, L., McInerney, S., Macklin, M., & Lasko, N. (2000). Exaggerated amygdala response to masked facial stimuli in posttraumatic stress disorder: A functional MRI study. Biological Psychiatry, 47, 769–776.

The Stroop interference effect Ritter, W., Simson, R., & Vaughan, H. G. (1983). Event-related potential correlates of two stages of information processing in physical and semantic discrimination task. Psychophysiology, 20, 168–179. Rowe, J. B., Toni, I., & Josephs, O. (2000). The prefrontal cortex: Response selection or maintenance within working memory? Science, 288, 1656–1660. Sharot, T., Martorella, E. A., Delgado, M. R., & Phelps, E. A. (2007). How personal experience modulates the neural circuitry of memories of September 11. Proceedings of the National Academy of Sciences, USA, 104, 389–394. Shin, L. M., Wright, C. I., Cannistraro, P. A., Wedig, M. M., McMullin, K., Martis, B., et al. (2005). A functional magnetic resonance imaging study of amygdala and medial prefrontal cortex responses to overtly presented fearful faces in posttraumatic stress disorder. Archives of General Psychiatry, 62, 273–281. Stroop, J. R. (1935). Studies of interference in serial verbal reactions. Journal of Experimental Psychology, 18, 643–662. Vogel, E.K, & Luck, S. J. (2000). The visual N1 component as an index of a discrimination process. Psychophysiology, 37, 190–203.

1001 Vyas, A., Mitra, R., Shankaranarayana Rao, B. S., & Chattarji, S. (2002). Chronic stress induces contrasting patterns of dendritic remodeling in hippocampal and amygdaloid neurons. Journal of Neuroscience, 22, 6810–6818. Wang, L., Zhang, Y. Q., Wang, W. Z., Shi, Z. B., Shen, J. H., Li, M., & Xin, Y. (2009). Symptoms of posttraumatic stress disorder among adult survivors three months after the Sichuan earthquake in China. Journal of Traumatic Stress, 22, 444–450. West, R., & Alain, C. (1999). Event-related neural activity associated with the Stroop task. Cognitive Brain Research, 8, 157–164. Williams, L. M., Kemp, A. H., Felmingham, K., Barton, M., Olivieri, G., Peduto, A., et al. (2006). Trauma modulates amygdala and medial prefrontal responses to consciously attended fear. NeuroImage, 29, 347–357.

(Received August 4, 2009; Accepted November 17, 2009)

Psychophysiology, 47 (2010), 1002–1010. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01024.x

New P300-based protocol to detect concealed information: Resistance to mental countermeasures against only half the irrelevant stimuli and a possible ERP indicator of countermeasures

J. PETER ROSENFELD AND ELENA LABKOVSKY Department of Psychology, Northwestern University, Evanston, Illinois, USA

Abstract Here, a rare probe or frequent irrelevant stimulus (S1) appeared in the first part of the trial, followed by either a target or nontarget (S2) in the second. Subjects randomly pressed one of five buttons to S1 to signal seeing it. Then they pressed one of two buttons for nontargets or targets. We tested three groups: simple guilty (SG), in which one stimulus was the subject’s birth date (Probe); innocent (IN) in which all date stimuli were irrelevant; and Countermeasure (CM), like SG but subjects performed mental CMs to 2 of 4 irrelevants. Bootstrapped-based hit rates in the SG group 5 100%, based on probe versus all four averaged irrelevants (Iall), or based on probe versus RT-screened maximum irrelevant (Imax). In the IN group there was one false positive (8%, Probe vs. Iall) or none (0%, Probe vs. Imax). In the CM group, 100% were detected based on Probe versus Iall (92% based on Probe vs Imax). A new eventrelated potential at Fz and Cz at 900 ms indexed CM use. Descriptors: Psychophysiological detection of deception, P300, Event-related potentials, Guilty knowledge tests, Concealed information tests, Lie detection, Credibility assessment

as to remove the dual-task competition and thereby allow larger probe P300s with more resistance to CMs. The CTP was consistently accurate (490% detections) and it resisted physical CMs applied to all four irrelevant stimuli used. Increased reaction time (RT) to each countered S1 irrelevant always allowed identification of CM use. The present version of the CTP contains a novel R1: Instead of pressing a single button as R1 (as in Rosenfeld et al., 2008), the participants had a box with five buttons and were instructed to randomly select one of the five to press on each trial to indicate their having seen S1. There was still no stimulus classification decision to be made; the new R1 button press was randomly selected. We reasoned that executing the more complex R1 would make it more difficult to execute a timely, true covert CM response prior and in addition to the complex R1 in a CM condition. We thus expected even greater RT delays in CM trials, leading to even clearer RT indicators of CM use on these trials. However, in the present test of the CTP, in order to make it easier for participants to use true, covert CMs less detectably, that is, more successfully, we also had them make exclusively mental CMs in the form of silent, specific name recalls. The previous studies by Rosenfeld et al. (2004, 2008) involved three physical CMs and one mental one. We reasoned that mental CMs could be executed more rapidly than physical responses and thus pose a greater challenge to the protocol in terms of identification of CM use via RT analysis.

Previous P300-based concealed information tests (CITs; Allen, Iacono, & Danielson, 1992; Farwell & Donchin, 1986; Rosenfeld et al., 1988) were variably accurate (48%–87% detections of guilty subjects), but vulnerable to countermeasures (CMs; Mertens & Allen, 2008; Rosenfeld, Soskins, Bosh, & Ryan, 2004). We suspected this vulnerability was because, on each trial of the old protocol, the subject had two simultaneous tasks: (a) the explicit task of discriminating targets (irrelevant stimuli assigned a unique response) from all other stimuli, including probes (key ‘‘guilty knowledge’’ items) and irrelevants and (b) the implicit task of recognizing probe stimuli because of their personal meaningfulness. The task simultaneity was that in each trial, a subject was presented unpredictably with one of either a probe, target, or irrelevant and was alert for either target or probe appearance. We reasoned that this dual task competition (Donchin, Kramer, & Wickens, 1986) would reduce probe P300s, creating the CM vulnerability. We recently introduced a novel, P300-based protocol for detecting concealed information that we called the complex trial protocol (CTP; Rosenfeld et al., 2008). We designed the CTP to separate in time the probe or irrelevant (S1) recognition (R1) from a delayed target or nontarget (S2) discrimination (R2), so Address correspondence to: Department of Psychology, Northwestern University, Evanston, IL 60208, USA. E-mail: jp-rosenfeld@ northwestern.edu 1002

P300 CTP resists two mental countermeasures Finally, there was one further change in the present experiment. In the previous study, we required participants to make CMs to each of the four irrelevant stimuli seen, as we did in Rosenfeld et al. (2004, 2008) in which the CMs were most effective against the older P300-based CIT protocols. In the present study, we required CMs to just two of the four irrelevant stimuli. Since the submission of the 2008 report, we became concerned that when the participant countered all irrelevants but not, of course, the probe, the latter acquired salience beyond that associated with its crime or personal relevance: It became the only noncountered S1 stimulus. Although the subject in the CM condition might have been prepared in this block to execute a CM most of the time (as there were 80% irrelevant trials), she would have to inhibit a CM response uniquely when the S1 was a probe. Our concern was that the special salience the probe thus acquired was responsible at least in part for the enhanced probe P300 in the CM condition seen previously. Indeed, we have recently confirmed this (‘‘OMIT’’) effect (Meixner & Rosenfeld, 2010). Thus, having a participant counter four of four irrelevants but not the sole probe stimulus gives an ‘‘unfair’’ advantage to the experimenter/operator. It is more of a challenge to the CTP to have participants counter less than all available irrelevants, so here only two of four irrelevants are countered.

Method Participants Participants were 38 students (20 men and 18 women) from the introductory psychology class pool. All had normal or corrected vision and ranged in age from 18 to 23 (mean 5 19.89, SD 5 1.91). The subjects were randomly assigned to three groups: (a) a simple guilty (SG) group, for which one of the five S1 stimuli was a subject’s birth date (probe) and the other stimuli were irrelevant to that subject; (b) an innocent (IN) group, where all five S1 date stimuli were irrelevant to subjects; and (c) a countermeasure (CM) group, similar to the SG group except that, in this group, subjects performed mental CMs to two of the four irrelevants. Procedures In the SG group (n 5 13, 5 women, mean age 5 19.3 years, SD 5 1.65), a subject first saw either a probe (subject’s birthdate, p 5 1/5) or one of four irrelevant dates (p 5 4/5), presented in random order. If a subject stated before testing that an irrelevant was personally meaningful, it was replaced. Then, after about a 1-s delay, S2, a string of numbers, was presented. Participants were asked to make two responses within a trial, one to each type of stimulus. For R1, each was asked to randomly press one of the five buttons on a five-button response box with the left hand immediately after seeing a date (Figure 1). The second response (R2) was to be made when S2, the string of numbers, appeared on the screen. This response had to be made on a two-button response box with the right hand. The string of numbers was either ‘‘111111,’’ ‘‘222222,’’ ‘‘333333,’’ ‘‘444444,’’ or ‘‘555555.’’ These were also presented in a random order. A participant had to press the right target button following ‘‘11111’’ and the left nontarget button following any other number string. In the IN group (n 5 13, 4 women, mean age 5 19.69 years, SD 5 0.63) subjects were presented with all irrelevant S1s. The second part of each trial was exactly as in SG participants. R1 and R2 were as in SG group.

1003

Figure 1. Structure of each trial. On each trial, subjects viewed two stimuli: one date (probe or irrelevant) and one string of numbers (target or nontarget). Using the left hand, subjects pressed a randomly selected response button on a five-button box. When the string of numbers later appeared, subjects used their right hand to press a right button on another box if the string was all 1s (target) and the left button if the string was a series of any other numbers (nontarget).

In the CM group (n 5 12, 9 women, mean age 5 20.75 years, SD 5 2.77) the stimuli were exactly the same as in SG group, but subjects were additionally instructed to ‘‘make a secret additional response prior to the first random button [R1] response to two of the four irrelevant dates.’’ These two CM responses were the participant’s silent, mental imaging of his or her first name and last name. In each of the three groups, prior to the actual run, participants were given 20–30 practice trials. Detailed Trial Structure (Figure 1) Each trial began with a 100-ms baseline period during which prestimulus elelctroencephalogram (EEG) was recorded. Then, as EEG recording continued, a 1-cm-tall S1 (e.g., ‘‘Feb 9’’) was presented for 300 ms in white font on a computer display about 1 m from the subject’s eyes. Subjects were instructed that immediately after they saw the stimulus to press a randomly selected button on the left-hand five-button response box, thus signaling their having seen the first stimulus. The first stimulus was followed by a randomly varying interval with a black screen (Figure 1) that randomly endured for 1400 to 1650 ms. At the expiration of this interval, the second stimulus (target or nontarget) was presented for 300 ms. Subjects were instructed to press a right button for a rare target and a left button for a nontarget, both on a right-hand, two-button box. Both probes and irrelevants could be followed by targets or nontargets. Targets were used here to maintain attention, but we also forced attention to the first stimulus by interrupting the run unpredictably every 20–40 trials and requiring the subject to identify it by speaking it aloud. Prior to the run we told each participant that he or she would be asked at various times throughout the experiment which date he or she saw last. This also tends to discourage simple CMs such as vision blurring. Subjects failing to correctly name S1 on more than two tests were to be dropped, but there were none. Table 1a shows the numbers and probabilities of probe/target, probe/nontarget, irrelevant/target, and irrelevant/nontarget trials. It is noted that probe targets and nontargets have equal probabilities whereas irrelevant nontargets are much more probable than irrelevant targets. This was done in the first study (Rosenfeld et al., 2008) because we wanted to confirm that irrelevant targets would evoke P300s to the targets, so we kept their probability rare. We retained this probability matrix here for comparative purposes. A possible confounding problem

1004

J.P. Rosenfeld and E. Labkovsky

Table 1a. Stimulus Probabilities Trial type: S1–S2 Probe-target Probe-nontarget Irrelevant-target Irrelevant-nontarget All probes All irrelevants

Number

Probability

33 33 33 250 66 283

.09 .09 .09 .72 .19 .81

Note: A ‘‘probe-target’’ trial is one in which a probe is followed by a target, an ‘‘irrelevant-target’’ trial is one in which an irrelevant is followed by a target, and so on.

results: Probes could have become much more salient than irrelevants because they were much more likely to be followed by a target, that is, the conditional probability of a target following a probe was much greater than the conditional probability of a target following an irrelevant. In the previous study, and in the present study, we controlled for this confound by using innocent control groups. If the confound was operating, then SG and CM results could have shown high detection rates due only to probe salience associated with relatively high probability of pairing with targets. If this were so, then IN subjects should also have shown high false positive rates, which were, in fact, only 8% in the previous study. Moreover in Rosenfeld, Tang, Meixner, Winograd, and Labkovsky (2009), we found no differences between symmetric and asymmetric conditional probability groups regarding hit rates. Data Acquisition EEG was recorded with Ag/AgCl electrodes attached to sites Fz, Cz, and Pz. The scalp electrodes were referenced to linked mastoids. Electrooculogram (EOG) was recorded with Ag/AgCl electrodes above and below the right eye. The diagonal placement of the eye electrodes ensured that both vertical and horizontal eye movements would be picked up, as verified in a pilot study and in Rosenfeld et al. (2004, 2008). The artifact rejection criterion was 40 mV. The EEG electrodes were referentially recorded, but the EOG electrodes were differentially amplified. The forehead was connected to the chassis of the isolated side of the amplifier system (‘‘ground’’). Signals were passed through Grass P511K amplifiers with a 30-Hz low-pass filter setting and high-pass filters set (3 db) at 0.3 Hz. Amplifier output was passed to a 16-bit National Instruments A/D converter sampling at 500 Hz. For all analyses and displays, single sweeps and averages were digitally filtered off-line to remove higher frequencies; the digital filter was set up to pass frequencies from 0 to 6 Hz using a ‘‘Kaiser’’ filtering algorithm. P300 at Pz was measured using the peak–peak (p–p) method, which, as repeatedly confirmed in our previous studies, is the most sensitive in P300-based deception investigations (e.g., Soskins, Rosenfeld, & Niendam, 2001): The algorithm searched from 300 to 650 ms for the maximally positive 100-ms segment average. The midpoint of the segment defined P300 latency. Then it searched from this P300 latency to 1300 ms for the maximum 100-ms negativity. The difference between the maximum positivity and negativity defined the p–p measure. Analyses, Error Handling To determine group effects analyses of variance (ANOVAs) were run. Any within-subject tests with 41 df resulted in our use of the Greenhouse–Geisser (GG) corrected value of probability,

p(GG), and the associated epsilon value. All artifact trials were discarded so that analyses were done only on artifact-free trials. To ensure that subjects were cooperating with instructions, we monitored S1 random responses during recording. During the recording we verified that subjects were randomizing the choices, avoiding pressing the same button twice in a row, and not developing any other pattern of button presses. In terms of the S2, the computer monitored accuracy for target and nontarget buttons (results below). Within Individual Analysis: Bootstrapped Amplitude Difference Method To determine whether or not the P300 evoked by one stimulus is greater than that evoked by another within an individual, the bootstrap method (Wasserman & Bockenholt, 1989) was used on the Pz site where P300 is typically largest. This will be illustrated with an example of a probe response being compared with an irrelevant response. The type of question answered by the bootstrap method is: ‘‘Is the probability more than 90 in 100 that the true difference between the average probe P300 and the average irrelevant P300 is greater than zero?’’ For each subject, however, one has available only one average probe P300 and one average irrelevant P300. Answering the statistical question requires distributions of average P300 waves, and these actual distributions are not available. One thus bootstraps these distributions, in the bootstrap variation used here, as follows: A computer program goes through the combined probe–target and probe–nontarget set (all single sweeps) and draws at random, with replacement, a set of n1 waveforms. It averages these and calculates P300 amplitude from this single average using the maximum segment selection method as described above for the p–p index. Then a set of n2 waveforms is drawn randomly with replacement from the irrelevant set, from which an average P300 amplitude is calculated. The number n1 is the actual number of accepted probe (target and nontarget) sweeps for that subject, and n2 is the actual number of accepted irrelevant sweeps for that subject multiplied by a fraction (about .25 on average across subjects in the present report), which reduces the number of irrelevant trials to within one trial of the number of probe trials. The calculated irrelevant mean P300 is then subtracted from the comparable probe value, and one thus obtains a difference value to place in a distribution that will contain 100 values after 100 iterations of the process just described. Multiple iterations will yield differing (variable) means and mean differences due to the sampling-withreplacement process. To state with 90% confidence (the criterion used in preceding studies, e.g., Farwell & Donchin, 1991; Rosenfeld, Angell, Johnson, & Qian, 1991; Rosenfeld et al., 2004; Soskins et al., 2001) that probe and irrelevant evoked ERPs are indeed different, we required that the value of zero difference or less (a negative difference) not be ! 1.29 SDs below the mean of the distribution of differences. In other words, the lower boundary of the 90% confidence interval for the difference would be greater than 0. It is further noted that a one-tailed 1.29 criterion yields a po.1 confidence level within the block because the hypothesis that the probe evoked P300 is greater than the irrelevant evoked P300 is rejected either if the two are not found significantly different or if the irrelevant P300 is found larger. (T tests on single sweeps are too insensitive to use to compare mean probe and irrelevant P300s within individuals; see Rosenfeld et al., 1991.) In the present study we additionally used a more rigorous test, introduced in Rosenfeld et al. (2008): We compared the probe to

P300 CTP resists two mental countermeasures the largest irrelevant P300 (Imax), which, if probe4Imax, justifies the inference that it is larger than all irrelevant P300s. This is because, in comparing the probe P300 against the average of all four irrelevant P300s (Iall) combined in the bootstrap, as we and others did in all previous studies, it is possible to obtain a positive outcome, even though one or more irrelevant P300s may be as large as or larger than the probe P300. We chose this bootstrap method of comparing the Imax P300 to the probe P300 within a subject because it is uses the same approximate number of trials for each member of the comparison. However, it was also confirmed that the maximum irrelevant P300 amplitude was not associated with a statistically confirmed, unusual reaction time to the first (‘‘I saw it’’) stimulus. As we show below and as already shown in Rosenfeld et al. (2004, 2008), significant reaction time increases were associated strongly with CM use. (A detailed diagnostic algorithm will be provided on request to the senior author.) Finally, in describing diagnostic accuracy results of experiments, we made use of the signal detection theoretical parameter, A’, based on Grier (1971). This is a function of the distance between a ROC curve and the main diagonal of a ROC plot of hits and false alarms. It makes no assumptions about the shape or variances of the distributions of the key variables (such as PI P300 amplitude differences). A’ varies from .5 (null effect) to 1.0 (maximum effect). A’ 5 1/21((y ! x)(11y ! x)/(4y(1 ! x))); y 5 hit rate, x 5 false alarm rate.

Results Behavioral; error rates and reaction time; S1 Regarding S1 error rate (failing to correctly identify S1 on the six or seven random test occasions when a subject was interrupted with the S1 identity question), the mean percent over groups and subjects was 8.5% " 10%; the maximum number of such errors seen across all groups/blocks was two (of six or seven tests per subject). A one-way ANOVA on error rates across groups was not significant (p4.9). RTs for the probe and each of the four irrelevant S1 stimuli are shown in Figure 2. It is noted in this figure that in the CM group, the first two irrelevants (I1 and I2) are the non countered ones and I3 and I4 are countered. These were different randomly designated stimuli for each subject in the CM group.

1005 Table 1b. Percent Use of Each of Five Buttons in Each Group Group SG IN CM

Button 1

Button 2

Button 3

Button 4

Button 5

18.7 17.4 15.4

21.5 23.1 22.3

20.4 18.6 19.7

23.8 23.5 24.6

15.6 17.3 17.9

It is evident from Figure 2 that, as in Rosenfeld et al. (2004, 2008), CMs produce elevated RTs for all S1 stimuli but mostly for countered stimuli (I3 and I4 in the CM group). It is also a typical finding (e.g., Farwell & Donchin, 1991; Rosenfeld et al., 2004, 2008; Seymour, Seifert, Mosmann, & Shafto, 2000) that in a SG group, the probe produces the slowest RT. It also appears that in the IN group, because all stimuli are irrelevant and there is no decision or salient stimulus, all stimuli appear to have the same RT. As a first analytic step, we ran a 3 # 2 (Group # Stimulus Type: probe vs. all irrelevants combined) mixed ANOVA. The results revealed a main effect of group (SG, IN, and CM), F(2,32) 5 12.2, po.001, obviously due to the elevated RTs in the CM group as previously reported (Rosenfeld et al., 2004, 2008), and a significant Group # Stimulus type interaction, F(2,32) 5 6.9, po.004. We attributed the lack of main effect of stimulus type, F(1,32) 5 0.5, p4.4, to the interaction and did follow-up tests within groups and particular stimulus types. A post hoc t test of interest on RT for Iall, SG versus IN, yielded t(21) 5 0.57, p4.5, meaning that the mean difference of about 30 ms (IN faster than SG) seen in the figure is not reliable. Moreover, in the IN group, there was no difference between the probe RTand the Iall RT, as expected and suggested in the figure, t(10) 5 0.98, p4.3. In contrast, and as expected, the same comparison within the SG group yielded probe (545.6 ms)4Iall (509.1 ms), t(11) 5 2.72, po.03. A final, repeated measures t test of interest within the CM group showed, as expected, that reaction times to countered irrelevant stimuli (825.8 ms) were greater than to noncountered irrelevant stimuli (689.7 ms), t(11) 5 3.2, po.009. Table 1b shows the percent button presses by groups and button numbers. It appears that the thumb and little fingers were used least in all groups. Average presses per button were at least 40, at most 77. Behavioral; Error Rates and Reaction Time; S2 Table 2a illustrates error rates for the second (target vs. nontarget) response. (An error for the second response means responding with a wrong button press either to a target or nontarget.) There were higher error rates to the targets in all three groups. This is as seen previously with the CTP (Rosenfeld et al., 2008) and will be discussed in terms of response perseveration. A mixed 3 # 2 # 2 ANOVA was applied to these S2 error data. Table 2a. S2 Error Rates Sorted by Groups and Stimulus 2 Types

Figure 2. RTs in milliseconds as a function of stimulus type for each of the three groups. In the CM group, Probe, I1, and I2 were uncountered and I3 and I4 are countered.

Group

PT

PN

IT

IN

SG IN CM

.032 .036 .039

.018 .016 .009

.157 .034 .082

.011 .008 .012

Note: PT refers to a target stimulus preceded by a probe, PN is a nontarget preceded by a probe, IT is a target preceded by an irrelevant, and so on.

1006

J.P. Rosenfeld and E. Labkovsky

Table 2b. S2 Reaction Times (in Milliseconds) Sorted by Groups and Stimulus 2 Types Group SG IN CM

PT

PN

IT

IN

486.3 575.4 603.0

495.6 540.3 545.9

569.2 598.0 633.5

473.8 537.4 549.0

Note: PT refers to a target stimulus preceded by a probe, PN is a nontarget preceded by a probe, IT is a target preceded by an irrelevant, and so on for IN.

The independent variable (factor) was group: SG, IN, and CM (three levels). One within-subject independent variable was stimulus type: probe versus Iall. The other was target versus nontarget. The results revealed no difference between groups, F(2,35) 5 1.656, p4.2. There was an effect of stimulus type, F(1,35) 5 9.97, po.004 (irrelevant4probe) and an interaction between stimulus type and group, F(2,35) 5 5.4, po.01. There was a large effect of target versus nontarget, F(1,35) 5 24.1, po.001 (target4nontarget). The interaction of the target effect and group was not significant, F(2,35) 5 2.52, p 5 .095. There was an interaction of probe versus irrelevant ! target versus non target, F(1,35) 5 7.8, po.009. The triple interaction was also significant, F(2,35) 5 3.29, po.05. Table 2b illustrates RTs for the second (target vs. nontarget) response. There were generally higher RTs to the targets in all three groups. This is as seen previously with the CTP (Rosenfeld et al., 2008) and will be discussed in terms of response perseveration. It also appears that the subjects in the simple guilty group

were the fastest, but this was not statistically reliable (see below). A mixed 3 ! 2 ! 2 ANOVA was applied to these S2 RT data. The independent variable (factor) was group: SG, IN, and CM (three levels). One within-subject independent variable was stimulus type: probe versus Iall. The other was target versus nontarget. The results revealed no difference between groups, F(2,35) 5 0.5, p4.6. There was an effect of stimulus type, F(1,35) 5 10.2, po.004 (irrelevant4probe) but no interaction between stimulus type and group, F(2,35) 5 1.1, p4.3. There was a large effect of target versus nontarget, F(1,35) 5 26.97, po.001 (target4nontarget). The interaction of the target effect and group was not significant, F(2,35) 5 0.66, p4.5. There was a large interaction of probe versus irrelevant ! target versus non target, F(1,35) 5 17.9, po.001. The triple interaction was also significant, F(2,35) 5 4.47, po.02. The possibly interesting finding in these S2 data concerns the significant probe versus irrelevant effects, because they could yield a new diagnostic index. Unfortunately, in the SG group, there were 5 of 13 exceptions (cases where probe error rate4irrelevant) to the mean findings (of probe error rateoirrelevant) with respect to error rate. There were 3 of 13 such exceptions (probe RT4irrelevant RT) with respect to RT. In the CM group, there were 6 of 12 exceptions regarding error rate and 4 of 12 exceptions regarding RT. Exceptions were common regarding error rate, as there were so many individual values of 0.0 errors. ERPs: Qualitative Figure 3A shows the grand averaged ERPs at Pz for each group and S1 type, with all four irrelevants combined (Iall) and with

Figure 3. (A) Superimposed Pz probe and Iall ERPs for each of three groups. Positive is down and each tic–tic interval on the y-axis represents 2 mV, and on the x-axis it represents 200 ms. (B) Superimposed Pz ERPs in the CM group. The left graph shows superimposed probes and both averaged countered irrelevants (Call). The center graph shows probes superimposed on uncountered irrelevants (NCall). The right graph shows superimposed countered and uncountered irrelevants. Polarity and tic-tic intervals are as in A.

P300 CTP resists two mental countermeasures

1007 Table 3. Diagnoses of Guilt within Each Group Based on Four Bootstrap Tests Group SG IN CM

Figure 4. Computed peak to peak Pz amplitudes for probes and irrelevant (Iall) in each group.

target and nontarget trials combined. In the SG group, the probe P300 (p–p) average towers over the Iall average, whereas in the IN group, there is no apparent difference. In the CM group, the probe average 4 the Iall average, but the difference is smaller than in the SG group. Figure 3b breaks down this difference in the CM group. The left panel of Figure 3b shows superimposed probe and countered irrelevant averages (both countered irrelevants combined), which are similar. The center panel of Figure 3b shows the probe average clearly larger than the average of the two noncountered irrelevants. Although the probe is clearly larger in this central panel, it is still not as large as in the SG group or in the SG and CM groups of Rosenfeld et al. (2008), in which all four irelevants were countered. The right panel of Figure 3b shows countered and noncountered irrelevant averages. Figure 4 shows line graphs of computer-calculated mean (p–p) P300 amplitudes for probes (targets and nontargets combined) and all four irrelevants combined (also targets and nontargets combined) in the three groups. The figure shows more clearly the relationships shown in Figure 3A and, in particular, illustrates the fact that the group differences are mainly in probe size. ERPs: Quantitative Group Data It seemed reasonable to keep the group comparison restricted to SG versus CM, because all we wanted to show in the IN group was that there was no difference between probe and Iall, which was confirmed with a repeated measures value of t(12)o0.5, p4.6. A 2 (Group; SG vs. CM) ! 2 (S1 type) ANOVA in these two groups (which both had concealed probe information) yielded F(1,23)o0.8, p4.3 for the group effect. The stimulus type effect was significant, F(1,23) 5 115, po.001, as was the 2 ! 2 interaction, F(1,23) 5 8.89, po.008. Figure 4 suggests that the latter interaction represents that the probeIall difference is greater in the SG group than in the CM group. A post hoc t test comparing the probe P300s in the CM versus SG groups yielded a marginally significant t(23) 5 1.81, po.09, but only t(23) 5 0.5, p4.6 for the Iall P300s. ERPs: Quantitative Individual Table 3 gives the detection rates within each group, based on the bootstrapping procedures described above. Within the CM group, there are four sets of proportions tabulated. First, there is Iall, exactly as in the SG and IN groups. There is also an Imax, as in the other groups, but for the CM group, we used a RT-based screening procedure (because CMs elevate RTs) to determine

Iall

Imax

Ic

In

13/13 1/13 12/12

13/13 1/13 11/12

F F 5/12

F F 11/12

Imax: We first looked for the numerically largest irrelevant P300 amplitude among I1–I4. If the probe P300 was greater than this Imax, the subject was considered detected. If not, we checked the associated RT: If, say, this unscreened Imax was I3, we compared (t test) the RTs of the probe and I3. If the RT for I3 was not significantly greater than the probe RT, then the subject was called a miss. If the RT for I3 was significantly greater than the probe RT, then we assumed CM use with I3 and did the bootstrap test on probe versus the next largest irrelevant as the RTscreened Imax test. This procedure could have been repeated with another irrelevant, but in the present study, it was never necessary to look beyond the second largest irrelevant to diagnose guilt despite CM use. There are two extra row values in Table 3 for the CM subjects only, Ic and In. These are based on bootstrap comparisons of the probe P300 with the two irrelevant stimuli (a) specified to the subjects as to be countered (Ic), and (b) not to be countered (In), respectively. These values are not relevant to a field situation in which an operator does not know which or how many irrelevant stimuli a subject might choose to counter. The screened Imax value in the second column of Table 3, discussed above, serves that purpose, as it allows the operator to take advantage of a known relationship between CMs and RT in screening irrelevant P300s for comparison with the probe P300. The critical summary observations here are that both Iall and screened Imax values that are available in a field situation yield 100% and 92% accuracy, respectively, in the CM group, and in the SG group accuracy is 100% overall. With the low 8% false positive values in the IN group, detection efficiency as measured by Grier’s (1971) A’ values remain high at .96 and .98 for CM and SG groups, respectively. ERPs: Serendipitous Putative New ERP Figure 5 shows the average ERPs for all three groups and three sites, probes superimposed on Ialls. What was not apparent at the CM group’s Pz site (Figure 3A) is the late positivity in the probe ERP, which can now be prominently seen in the probe responses at Fz and Cz (but not in Pz) and restricted to the CM group. This putative new component is labeled ‘‘P900,’’ as its mean probe latency was 912 ms across sites. For analytic purposes, we measured this component from the average prestimulus 100 ms baseline to the maximum positivity between 700 and 1100 ms so as to obtain the P900, baseline to peak. We chose the most positive 100-ms average segment between 700 and 1100 ms. To explore this component, we first did a 3 (site) ! 3 (group) ANOVA on the probes and found a main effect of both variables. For site, F(2,70) 5 7.8, p(GG)o.005, e 5 .69. For group, F(2,35) 5 3.94, po.03. The interaction appeared significant in Figure 6, but was not significant, F(4,70) 5 1.75, p(GG) 5 .17, e 5 .69. To establish that the P900 effect is in the CM group and not in SG or IN groups, we first did a 2 (groups: SG vs. IN) ! 3 (site) ANOVA. There was no group effect (showing SG 5 IN), F(1,24) 5 1.74, p4.2. There remained a marginal site effect,

1008

J.P. Rosenfeld and E. Labkovsky

Figure 5. Same as Figure 3A for all three scalp sites.

F(2,48) 5 3.01, p(GG) o.08, e 5 .76, but no interaction, F(2,48) 5 .49, p(GG)4.5, e 5 .76. We then did a 2 (groups, combined SG and IN, vs. CM) ! 3 (sites) ANOVA. This time,

the group effect was significant, F(1,36) 5 6.22, po.02, confirming a P900 difference between the CM and other two groups. The site effect remained significant, F(2,74) 5 9.61, p(GG)o.001, e 5 .69. The marginal interaction was F(2,72) 5 3.11, p(GG)o.08, e 5 .69. Finally, to help confirm that P900 in the CM group is largest in Fz and Cz and reduced in Pz, we first did a repeated measures t test within the CM group only, Fz versus Cz, so as to establish that P900 is the same at both Fz and Cz. The result was confirmatory: t(11) 5 .964, p4.3. Then we averaged Fz and Cz and compared these P900 values with Pz values. The result was t(11) 5 6.17, po.001. This confirmed what is seen in Figure 5, that the P900 is largest in Fz and Cz.

Discussion

Figure 6. Baseline to peak, P900 probe amplitudes as a function of site and group.

The major difference between the present report and Rosenfeld et al. (2008) was the use of two of four countered irrelevants in the CM group here versus four of four countered irrelevants previously. The latter CM strategy may be a very likely CM method for subjects to choose in field situations, because the aim would

P300 CTP resists two mental countermeasures

1009

be to convert all the irrelevant stimuli into covert targets that produce P300s as large as the probe’s. However, as already noted, we have shown (Meixner & Rosenfeld, 2010) that countering all S1 stimuli except the probe stimulus engages the effect of salience due to unique CM omission, resulting in a greatly enhanced probe P300 that easily exceeds the countered irrelevant P300s, resulting in detection. Countering a smaller fraction of all irrelevants would then seem to be a better CM strategy for defeat of the CTP. Thus, we believed it critical to test the CTP against such a CM strategy in the present report. The result was still most encouraging: If one uses the bootstrap test of probe against Iall, one still detects 100% of the CM users with only an 8% false positive rate. Even though the countered irrelevants produced averaged P300s almost as large as the probe’s (Figure 3b, left), the combined countered and noncountered irrelevant average (i.e., Iall) is still smaller than the probe P300 in all cases, yielding detections. If one uses the rigorous probe versus Imax bootstrap test, the false positive rate remains at 8%. If one applies the RT screening approach, in which one tests the probe P300 against the maximum irrelevant P300 not associated with a significantly elevated RTFa finding that clearly indicates CM useFthe detection rate is near perfect at 92%. Thus the CTP is robust to countering a few as well as all irrelevant stimuli. Another novel demonstrated attribute of the present CTP protocol is its robustness even to all mental CMs. We had assumed that mental CMs would be more challenging to the CTP because we reasoned that they would involve less time to execute and that would threaten the RT index of CM use and thus also the RT-based screening procedure described in the previous paragraph. Our reasoning was based on the obvious fact that a command for a mental CM response would be less time-consuming than a command for a physical CM response (such as those used in Rosenfeld et al., 2004, 2008) that would have to traverse the motor pathways. This may still be the case, but apparently, the bulk of the increase in RTassociated with CM use is not related to command and execution processes but to response selection processes that must precede execution commands. It is evident from Figure 2 that there are actually two response selection and decision processes associated with CM use: First, upon seeing S1, the user of two CMs must decide if the stimulus is or is not a to-be-countered irrelevant. Thus, the RT to

all stimuli is significantly increased in CM users in comparison to the RT of the other groups not using CMs. Second, the CM user must detect one of the two (of four) irrelevants to be countered and then select which of the two CMs to use. Thus Figure 2 shows that countered irrelevants produce significantly longer RTs than uncountered irrelevants, but, nevertheless, mental CM use remains detectable. The promise of the CTP for field use is tempered by the issue of probe and especially irrelevant development in real-world scenarios: Assumed irrelevant items could prove to have significance for a given subject. This was controlled here by screening irrelevant dates with each subject prior to testing and making replacements as needed. In field situations, using multiple blocks with different item categories in each block should alleviate the problem, as with any concealed information test (Lykken, 1998). Indeed, these tests are in field use in Japan, and this issue has not posed a problem (Hira & Furumitsu, 2009). Finally, we briefly comment on the putatively novel ‘‘P900’’ ERP. We think this component is novel, as we are unaware of any previous report of a similar component, unless one regards it as an ersatz later P300. This is contraindicated by its frontal-central scalp distribution, which differs from that of the posteriorly maximal P300. Because the P900 component appears only in the CM group (mainly at Fz and Cz), one may indeed have in it a new CM indicator to supplement enhanced RT. This begs the question of why it was not seen in the CM condition of Rosenfeld et al. (2008). The explanation may depend upon what cognitive process P900 signifies. The complete answer will take further research. We do have a preliminary hypothesis that the presentation of the probe in the present CM group signals the subject that his task for the trial is done, as there is no CM to select, and that P900 represents processing of that signal. In contrast, if an irrelevant is presented in the present CM condition, the subject has the further task of deciding whether a CM is required for this stimulus and, if so, which of the two assigned CMs to execute. In the 2008 paper, a probe may have had the same signal value to a CM subject, but because the P300 in the 2008 paper was greatly enhanced by the omit effect (described above), the large negative recovery that follows P300 (that was studied in Soskins et al., 2001, and is obvious here in the SG group of Figure 3A) probably obscured any possible P900. Meixner and Rosenfeld (2010) confirmed this effect.

REFERENCES Allen, J., Iacono, W. G., & Danielson, K. D. (1992). The identification of concealed memories using the event-related potential and implicit behavioral measures: A methodology for prediction in the face of individual differences. Psychophysiology, 29, 504–522. Donchin, E., Kramer, A., & Wickens, C. (1986). Applications of brain event related potentials to problems in engineering psychology. In M. Coles, S. Porges, & E. Donchin (Eds.), Psychophysiology: Systems, Processes and Applications (pp. 702–710). New York: Guilford. Farwell, L. A., & Donchin, E. (1986). The brain detector: P300 in the detection of deception [abstract]. Psychophysiology, 24, S34. Farwell, L. A., & Donchin, E. (1991). The truth will out: Interrogative polygraphy (‘‘lie detection’’) with event-related potentials. Psychophysiology, 28, 531–547. Grier, J. B. (1971). Non-parametric indexes for sensitivity and bias: Computing formulas. Psychology Bulletin, 75, 424–429. Hira, S., & Furumitsu, I. (2009). Tonic arousal during field polygraph tests in guilty vs. innocent suspects in Japan. Applied Psychophysiology and Biofeedback, 34, 173–176.

Lykken, D. T. (1998). A tremor in the blood. Reading, MA: Perseus Books. Meixner, J. B., & Rosenfeld, J. P. (2010). Countermeasure mechanisms in a P300-based concealed information test. Psychophysiology, 47, 57–65. Mertens, R., & Allen, J. J. (2008). The role of psychophysiology in forensic assessments: Deception detection, ERPs, and virtual reality mock crime scenarios. Psychophysiology, 45, 286–298. Rosenfeld, J. P., Angell, A., Johnson, M., & Qian, J. (1991). An ERPbased, control-question lie detector analog: Algorithms for discriminating effects within individuals’ average waveforms. Psychophysiology, 38, 319–335. Rosenfeld, J. P., Cantwell, G., Nasman, V. T., Wojdac, V., Ivanov, S., & Mazzeri, L. (1988). A modified, event-related potential-based guilty knowledge test. International Journal of Neuroscience, 24, 157–161. Rosenfeld, J. P., Labkovsky, E., Winograd, M., Lui, M. A., Vandenboom, C., & Chedid, E. (2008). The Complex Trial Protocol (CTP): A new, countermeasure-resistant, accurate P300-based method for detection of concealed information. Psychophysiology, 45, 906–919.

1010 Rosenfeld, J. P., Soskins, M., Bosh, G., & Ryan, A. (2004). Simple effective countermeasures to P300-based tests of detection of concealed information. Psychophysiology, 41, 205–219. Rosenfeld, J. P., Tang, M., Meixner, J. B., Winograd, M. R., & Labkovsky, E. (2009). The effects of asymmetric versus symmetric probability of targets following probe and irrelevant stimuli in the Complex Trial Protocol for detection of concealed information with P300. Physiology and Behavior, 98, 10–16. Seymour, T. L., Seifert, C. M., Mosmann, A. M., & Shafto, M. G. (2000). Using response time measures to assess ‘‘guilty’’ knowledge. Journal of Applied Psychology, 85, 30–37.

J.P. Rosenfeld and E. Labkovsky Soskins, M., Rosenfeld, J. P., & Niendam, T. (2001). The case for peakto-peak measurement of P300 recorded at .3 Hz high pass filter settings in detection of deception. International Journal of Psychophysiology, 40, 173–180. Wasserman, S., & Bockenholt, U. (1989). Bootstrapping: Applications to psychophysiology. Psychophysiology, 26, 208–221.

(Received May 12, 2009; Accepted November 18, 2009)

Psychophysiology, 47 (2010), 1011–1018. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01026.x

BRIEF REPORT

Top-down control of visual sensory processing during an ocular motor response inhibition task

BRETT A. CLEMENTZ,a YUAN GAO,a JENNIFER E. McDOWELL,a STEPHAN MORATTI,b SARAH K. KEEDY,c and JOHN A. SWEENEYc a

Departments of Psychology and Neuroscience, BioImaging Research Center, University of Georgia, Athens, Georgia, USA Centro de Magnetoencefalografı´ a Dr. Perez Modrego, Universidad Complutense de Madrid, Madrid, Spain c Center for Cognitive Medicine, University of Illinois, Chicago, Illinois, USA b

Abstract The study addressed whether top-down control of visual cortex supports volitional behavioral control in a novel antisaccade task. The hypothesis was that anticipatory modulations of visual cortex activity would differentiate trials on which subjects knew an anti- versus a pro-saccade response was required. Trials consisted of flickering checkerboards in both peripheral visual fields, followed by brightening of one checkerboard (target) while both kept flickering. Neural activation related to checkerboards before target onset (bias signal) was assessed using electroencephalography. Pretarget visual cortex responses to checkerboards were strongly modulated by task demands (significantly lower on antisaccade trials), an effect that may reduce the predisposition to saccade generation instigated by visual capture. The results illustrate how top-down sensory regulation can complement motor preparation to facilitate adaptive voluntary behavioral control. Descriptors: Attention, Bias signal, EEG, Saccade, Antisaccade, Visual steady state

Goldberg, 1999). Neurophysiology studies indicate that suppression of reflexive saccades to novel visual targets during antisaccade tasks requires an anticipatory reduction of neural activity in saccade motor circuitry prior to stimulus appearance (Everling & DeSouza, 2005), an effect of top-down control mediated by PFC (Johnston & Everling, 2006). This process reduces the predisposition to move instigated by visual capture and contributes to successful suppression of context-inappropriate saccades to peripheral targets on antisaccade trials (Munoz & Everling, 2004). Although preparatory effects in the motor system are established in neurophysiology studies, possible analogous effects in the human visual system have been less systematically investigated (but see McDowell et al., 2005). Reduced activity in visual cortex to antisaccade targets could complement anticipatory top-down inhibition in saccadic motor circuitry to support successful performance. The notion that topdown bias signals (what Desimone & Duncan, 1995, called the ‘‘attentional template’’) begin their influence on perception in a preparatory manner has a history in the visual selective attention literature (see, e.g., Desimone & Duncan, 1995; Kastner & Ungerleider, 2000; Maunsell & Treue, 2006). There are various theories about the neural source(s) of such bias signals (i.e., which brain region or regions control early visual cortical responses by ‘‘informing’’ visual cortex of preferred stimulus features and/or locations) and their importance for optimizing behavioral performance by influencing the course of information flow early in visual processing (e.g., Miller & Cohen, 2001).

Volitional control over gaze is critical for successfully navigating the environment. Paradigms like the antisaccade task require volitional cognitive control over otherwise prepotent responses (Munoz & Everling, 2004). Compared to a prosaccade task, where a glance is made to a peripheral target, antisaccade tasks require withholding a glance to a peripheral target and looking to that target’s mirror image location. The ability to perform antisaccade tasks depends on prefrontal cortex (PFC) mediated top-down control (McDowell, Dyckman, Austin, & Clementz, 2008; Pierrot-Deseilligny et al., 2003; Sweeney, Luna, Keedy, McDowell, & Clementz, 2007), which makes antisaccade paradigms excellent probes of the neural substrates of flexible behavior (Miller & Cohen, 2001). Demonstrating relationships between behavioral performance and the neural dynamics of sensory processing and motor planning are important steps toward discerning how top-down control supports flexible behavioral regulation. The present study addressed this issue by measuring neural activity with high temporal resolution electroencephalography (EEG; Nunez & Srinivasan, 2006). Abruptly appearing visual stimuli capture attention (Yantis & Jonides, 1996), manifest as an increased activity in extrastriate neurons tuned for that particular spatial location (Colby & This works was supported by grants from the United States Public Health Service (MH51129, MH001852). Address correspondence to: Brett A. Clementz, Psychology Department, Psychology Building, Baldwin Street, University of Georgia, Athens, GA 30602, USA. E-mail: [email protected] 1011

1012 In the present study, the effects of top-down bias signals, defined as modifications of sustained sensory responses in neural mass activity in visual cortex, were assessed prior to peripheral target onset during anti- and prosaccade trials using dense-array EEG. Use of the steady-state visual evoked potential (ssVEP) allowed for assessment of cortical facilitation/suppression of sensory processing in relation to possible peripheral target locations. The ssVEP is an electrocortical response to flickering stimuli, coming primarily from striate/extrastriate cortex (e.g., Clementz, Wang, & Keil, 2008; Di Russo, Taddei, Apnile, & Spinelli, 2006; Pastor, Artieda, Arbizu, Valencia, & Masdeu, 2003), where the frequency of brain activity equals the stimulus flicker rate. The ssVEP occurs at specific frequencies set by the experimenter via manipulating the frequency of the flickering visual stimuli, an advantage compared with visual evoked potentials (VEPs), which measure changes over a wider frequency range (Makeig et al., 2002). This type of ‘‘frequency tagging’’ (Mu¨ller, Malinowski, Gruber, & Hillyard, 2003) allowed by ssVEPs facilitates identification of neural activity related to specific stimuli, and can be used to track neural activity in relation to multiple stimuli simultaneously (Clementz et al., 2008; Mu¨ller & Hubner, 2002; Mu¨ller et al., 2003; Regan & Regan, 2003; Wang, Clementz, & Keil, 2007). Using a ssVEP paradigm, the present experiment reveals neural modulation of visual cortex activity preceding correct antisaccades and illustrates the importance of PFC mediated top-down control of sensory cortices for the successful control of context-appropriate behaviors.

Materials and Methods Participants Fifteen healthy right-handed individuals (age range: 19–27 years; 8 women) recruited from the student population at the University of Georgia participated after providing written informed consent. Participants had normal or corrected-to-normal vision, had no evidence of neurological impairment, were free of psychiatric or substance use disorders (by self-report), and were given course credit for their participation. This project was approved by the Institutional Review Board at the University of Georgia. Stimuli and Experimental Procedures Stimuli were presented on a 21-in. flat-surface high resolution color monitor (60 Hz vertical refresh) situated 100 cm from the participant’s nasion. Central fixation was a 21 white cross (see Figure 1). Peripheral stimuli were 41 square checkerboards that were centered at 91 left and right of central fixation. The checkerboards were composed of alternating gray and black boxes (each box was 11 square). The gray boxes were of equal luminance (5 cd/m2) and were presented against a dark background (o0.2 cd/m2). Steady-state visual evoked potential. The ssVEP was used to assess sustained brain activity in relation to peripheral target locations during the period preceding target onset. At trial initiation (see Figure 1), participants fixated the central cross for 5000 ms (the fixation period; the long duration allows for settling of the ssVEP neural generators). Following the fixation period, peripheral checkerboards were presented that were square-wave

B.A. Clementz et al.

Figure 1. An example trial. All trials began with subjects fixating a central stimulus for 5000 ms (the fixation period). After this interval, checkerboards began simultaneous flickering left (12 Hz) and right (15 Hz) for 5360 ms while participants maintained fixation on the central stimulus (the preparatory period). At the end of this interval, one of the peripheral checkerboards increased in luminance for 1500 ms (the target; began the response period), which cued the participants to make the required response (toward the target on pro trials and in the opposite direction on anti trials).

luminance modulated (100% modulation depth) at two flickering frequencies (12 Hz in the left visual field, 15 Hz in the right visual field) for 5360 ms (the first common multiple of 12- and 15-Hz cycles after 5000 ms) so that possible lateralized changes in electrocortical facilitation and suppression could be tracked via the ssVEP. Subjects were instructed to remain fixated on the central cross during this preparatory period. At the end of the preparatory period, one of the peripheral checkerboards (randomly determined) increased in luminance to 20 cd/m2 (defining target onset), signaling the subject to make a saccade response. The two checkerboards continued flickering at their respective frequencies for an additional 1500 ms (the saccadic response period), after which the peripheral checkerboards disappeared, leaving only the white cross on the screen to initiate the fixation period marking the start of a new trial. Behavioral task requirements. Participants were presented with two blocks of stimuli. During one block they were required to generate a saccade as quickly and accurately as possible to the middle of the peripheral target checkerboard (prosaccade block). During the other block, participants were asked to generate a saccade as quickly and accurately as possible to the middle of the nontarget peripheral checkerboard without first making a saccade to the target checkerboard (antisaccade block). There were 50 trials to both the left and right checkerboards in each block; the order of block presentation was counterbalanced. Electrophysiological Recording and Data Preprocessing Data collection. EEG data were vertex referenced and recorded using a 256-sensor Geodesic Sensor Net and NetAmps 200 amplifiers (EGI, Eugene, OR). Individual sensor impedances were kept below 50 kO (see Ferree, Luu, Russell, & Tucker, 2001). In addition, an electrolyte bridge test was conducted between all pairs of sensors prior to recording (Tenke & Kayser, 2001), and if there was evidence of bridging, sensors were adjusted until bridging was no longer evident (this was rarely required). Data were sampled at 500 Hz with an analog filter bandpass of 0.1–200 Hz. Sensors located at the outer canthi of each eye and below and above both eyes recorded horizontal and

Neural modulation preceding saccades vertical eye movements and eyeblinks. Following data collection, the three-dimensional locations of sensors were acquired using a photogrammetry rig (EGI, Eugene, OR). The horizontal EEG eye sensors were used to measure saccadic response for each participant. Data preprocessing. EEG sensors located at the neck and cheeks were excluded, leaving 207 sensors for data analyses. Raw data were visually inspected off-line for bad sensor recordings (BESA 5.1, MEGIS Software, Gra¨felfing, Germany). Bad sensors (no more than 5% of sensors for any subject) were interpolated using BESA’s spherical spline interpolation method. Trials (6860 ms in length) with EEG signals greater than 100 mV and/or other artifacts were automatically eliminated from further processing (fewer than 25% of trials were eliminated for any subject for both pro- and antisaccade conditions). The artifactfree data were then transformed to an average reference. The position data from the horizontal eye sensors for individual trials were visually inspected and scored for correct saccadic behavioral performance as in Dyckman and McDowell (2005). Saccadic latencies also were calculated (time from peripheral target onset to saccade initiation; see Dyckman & McDowell, 2005). Subjects made correct eye movements on 98% of prosaccade trials (reaction time M 5 375.0 ms, SD 5 19.1) and 92% of antisaccade trials (reaction time M 5 422.1 ms, SD 5 23.2). Given the low error rates, only correctly performed trials were analyzed. Data Analyses Two approaches were used to quantify brain activity. First, standard VEP measures were used to evaluate the amplitude and spatial distribution of brain activity at the onset of visual stimulation (at the beginning of the preparatory period and before initiation of the steady-state response). Second, spectral measures were used to assess the phase stability and power of the ssVEP over time during the preparatory period. VEP measures. The VEP data were used to assess for differences in brain activations, at the beginning of visual processing, between the pro- and antisaccade conditions that could be related to similar effects observed in previous publications (Clementz, Brahmbhatt, McDowell, Brown, & Sweeney, 2007; McDowell et al., 2005). This was done as a validity check on the paradigm used here, which was different in style from previous EEG antisaccade studies. As the overwhelming frequency composition of early VEPs is below 10 Hz (Moratti, Clementz, Gao, Ortiz, & Keil, 2007), the EEG data were digitally low-pass filtered at 10 Hz (12 dB/octave rolloff), which also served to reduce possible confusion between the early VEPs and the initiation of the ssVEP (at 12 and 15 Hz). Next, VEPs were averaged time-locked to flicker onset separately for the pro- and antisaccade trials. The VEP data were baseline corrected using the 200-ms prestimulus period, and VEP peaks occurring during the first 300 ms after stimulus onset (clearly prior to stabilization of the ssVEP) were compared between pro- and antisaccade trials. Latencies of these peaks were determined by inspection of global field power plots (global field power is the root mean square of voltage over all EEG recording sensors at each data sampling point, which can yield an initial assessment of the presence and latency of abovebaseline VEPs; Lehmann & Skrandies, 1984). Comparisons of scalp potential amplitudes between pro- and antisaccade trials were then conducted for each VEP peak (! 4 ms; see Figure 3, below). For each comparison, a paired t test was conducted separately for each EEG recording sensor. For sig-

1013 nificance thresholding, a clustering method (e.g., Forman et al., 1995) was used to take account of the nonindependence of data from adjacent EEG sensors, with significance levels determined based on the noise level of the data (estimated from the prestimulus period; see Krusemark, Campbell, & Clementz, 2008, for an example) and Monte Carlo simulations calculated using AlphaSim (Cox, 1996). To maintain the familywise alpha lower than .01, individual t tests for a given sensor required at least six neighboring sensors with effects statistically significant at po.035. After VEP analyses calculated on voltage data at the sensors, we used standardized low-resolution brain electromagnetic tomography (sLORETA; Pascual-Marqui, 2002) to estimate brain regions involved in determining the between-condition differences on VEPs observed in the sensor space data (see Clementz et al., 2008; Krusemark et al., 2008). sLORETA is a modification of minimum norm least squares (Ha¨ma¨la¨inen & Ilmoniemi, 1994) that uses the standardization of the minimum norm inverse solution to infer high probability regions of brain activation given the measured EEG data. sLORETA solutions yield pseudo-statistics that are not appropriate for determining strength of activity, but they provide accurate information about the regions of activity that can account for the voltage pattern recorded at the sensors (e.g., Soufflet & Boeijinga, 2005). The sLORETA calculations were performed using CURRY (Version 5.0, Neuroscan, Inc.). An averaged magnetic resonance (MR) image from the Montreal Neurological Institute (Collins, Neelin, Peters, & Evans, 1994) was used to construct a realistic head model (Fuchs, Kastner, Wagner, Hawes, & Ebersole, 2002) prior to source localization. The MR images were segmented into skin surface, inside of the skull, and cortex. A three compartment Boundary Element Method (BEM) model was then constructed; standard homogeneous conductivities were assumed for the skin, skull, and brain. For this BEM model, the average triangle edge lengths were 7.5 mm for the skin, 5.1 mm for the skull, and 3.1 mm for the brain compartment. Prior to source analysis calculations, the fiducial locations from the EEG data collection session were matched to the fiducial locations on the averaged segmented skin surface (using a least squares fitting procedure in CURRY). The sLORETA solutions were projected to the cortical surface. Spectral measures. Previous work has demonstrated that, under typical circumstances, the ssVEP is determined primarily by increased intertrial phase alignment (increased across-trial phase similarity of the EEG signals in relation to frequency of the visual flickering stimuli) without substantial changes in single trial power (Ding, Sperling, & Srinivasan, 2006; Moratti et al., 2007). In the present study, however, subjects were required to impose sustained top-down control as a function of pro- versus antisaccade task demands. The effect of this top-down control could be manifest on changes in either single trial power or intertrial phase alignment, so both aspects of the ssVEP response were quantified for between-condition comparisons. Single trial power of the steady-state response across time was estimated by complex demodulation at the flickering rates (12 and 15 Hz) of the checkerboards for each trial (Regan & Regan, 2003). Complex demodulation is a time series analysis method for quantifying the power of oscillatory biosignals (similar to wavelet analysis). Complex demodulation yields power as a function of time for oscillations at frequencies of interest. It can be conceptualized as a band-pass filter that eliminates all other frequencies except for the selected one (Draganova & Popivanov,

1014 1999; Haenschel, Baldeweg, Croft, Whittington, & Gruzelier, 2000), allowing for quantification of time-dependent changes in power of a particular frequency component. Data were multiplied with 12- and 15-Hz sine and cosine functions. A Butterworth zero-phase low pass filter of 1 Hz was applied to the resulting time series before we obtained the vector length of the sine and cosine parts (separately for 12- and 15-Hz signals) as a measure of time-varying power on single trials. The narrow low-pass filter was chosen to obtain sufficient frequency resolution by separating the driving stimulus frequencies from each other and from background alpha activity. Finally, the single trial 12- and 15-Hz power estimates were averaged across trials for each subject. To calculate intertrial phase-locking (ITPL) of the steadystate response, as in our previous work (Moratti et al., 2007), each trial was submitted to the same complex demodulation analysis as described above. ITPL was determined by normalizing the phase vectors spanned by the sine and cosine parts of the complex demodulation components for each flicker frequency and time step by the corresponding length of the vectors. For each time point and flicker frequency, the normalized phase vectors were added across trials of each condition and subject. Thereafter, the sum of the phase vectors was divided by the corresponding number of trials, resulting in the Rayleigh statistic R (Jammalamadaka & SenGupta, 2001). The R value is bound between 0 and 1. The higher the ITPL of an oscillatory response, the closer the R value will be to 1. EEG data from 91 sensors over posterior cortex that captured the ssVEP (see Figure 2 and Clementz et al., 2008; Wang et al., 2007) were separately analyzed to obtain single trial power and ITPL values. Results were then averaged over these sensors to obtain single estimates of these dependent variables. The single trial power and ITPL values were standardized (mean 0, unit variance) across time for each subject. This was done by initially combining all the time points (from ! 500 to 5000 ms) for the

Figure 2. Head surface map (looking at the back of the head) of evoked visual response power to the steady-state stimuli (averaged over both frequencies and conditions). For display purposes, brain responses to the 12- and 15-Hz stimuli were placed on the same scale (maximum of 1) to adjust for between-frequency ssVEP power differences.

B.A. Clementz et al. pro- and antisaccade conditions for each driving frequency individually (because 12 Hz power is normally larger than 15 Hz power, but we wanted these metrics to be on the same relative scale). Combining the pro- and antisaccade conditions data into this standardization also allowed for a direct comparison of differences between conditions in standard units. Averaged values for the spectral measures were then calculated for 500-ms intervals, resulting in 11 time points for statistical analyses. Fivehundred-millisecond time bins were used because, given the filter type, low-pass filter frequency, and filter order (fifth), the fullwidth half-maximum (FWHM) of the impulse response was estimated at 502 ms. No baseline adjustments were applied to these data because differences in baseline activity were possibly an important component of the analyses. Results VEPs to Peripheral Stimulus Onset At the initiation of the peripheral checkerboards flickering, there were three above baseline VEP peaks that were clearly identified from the global field power plots during both pro- and antisaccade blocks (see Figure 3) prior to the onset of the ssVEPs. The average latency of these peaks was 135 ms (SE 5 0.9), 196 ms (SE 5 3.1), and 227 ms (SE 5 2.0); the latencies did not differ as a function of task condition. Paired t tests comparing brain activity for pro- and antisaccade trials at the 135-ms peak revealed groups of sensors over superior parieto-frontal regions (see Figure 3), where participants had more extreme VEP amplitudes during antisaccade (average voltage in cluster 5 ! 1.52 mV, SD 5 0.31) than prosaccade trials (average voltage in cluster 5 ! 0.91 mV, SD 5 0.30). The sLORETA solution on this between-condition voltage difference indicated increased activity in bilateral visual cortices and PFC during antisaccade compared to prosaccade trials. Paired t tests comparing brain activity for pro- and antisaccade trials at the 196-ms peak revealed groups of sensors over bilateral inferior parieto-occipital regions (see Figure 3), where participants had more extreme VEP amplitudes during prosaccade (average voltage across both clusters 5 ! 1.15 mV, SD 5 0.12) compared to antisaccade trials (average voltage across both clusters 5 ! 0.94 mV, SD 5 0.18). The sLORETA solution on this between-condition voltage difference indicated increased activity in bilateral middle occipital gyrus during prosaccade compared to antisaccade trials. Paired t tests comparing brain activity for pro- and antisaccade trials at the 227-ms peak revealed groups of sensors over left and right superior and inferior parietal regions (see Figure 3) that had more extreme VEP amplitudes during antisaccade (average voltage across both clusters 5 1.10 mV, SD 5 0.1) than prosaccade trials (average voltage across both clusters 5 0.52 mV, SD 5 0.1). The sLORETA solution on this between-condition voltage difference indicated increased activity in bilateral visual cortex and superior parietal lobe during antisaccade compared to prosaccade trials. ssVEP response to peripheral stimuli. We used saccade type (pro vs. anti) by stimulus location (left vs. right) by time interval (eleven 500-ms intervals beginning 500 ms before steady-state stimulus onset) repeated measures analyses of variance (ANOVAs) to analyze (a) ITPL and (b) single trial spectral power in relation to the peripheral flickering checkerboards during the sustained preparatory period. For ITPL there was only a significant main effect of time interval, F(10,140) 5 13.6, po.001, indicating that synchronization of neuronal firing increased from

Neural modulation preceding saccades

1015

Figure 3. Visual evoked potentials (VEPs) to onset of the flickering stimuli at the beginning of the preparatory period. The VEP waveforms on the left illustrate the evoked response effects for sensors close to the regions of significant voltage differences (time 0 indicates stimuli onset). The colored topdown view topographies (voltage maps, with scales of ! 2 mV peak to peak) at the three VEP peaks are shown to the right of the waveforms for anti- and prosaccades (black dots show the locations of Fz, Cz, and POz). Clusters of sensors with significant between-condition effects for the three peaks are indicated in gray shaded sensor distributions to the right of the topography plots (red dots show the locations of Fz, Cz, and POz, from top to bottom). The sLORETA solutions for the VEP voltages differences between anti- and prosaccades are shown at the far right, with the label indicating which condition had stronger activity.

the beginning to the end of steady-state stimulation in both saccade conditions (see Figure 4). For single trial spectral power, there was a significant main effect of saccade type F(1,14) 5 9.9, p 5 .007, indicating that participants had differences in neural gain control of visual cortex signals between pro- and antisaccade trials (see Figure 4). Four follow-up analyses were also conducted on single trial power: (1) activity in the 500-ms prestimulus period did not differ significantly between pro- and antisaccade conditions, t(14) 5 1.83, p 5 .089, although the antisaccade condition tended to have lower power than the prosaccade condition even before stimuli onset; (2) during prosaccade trials, prestimulus activity did not differ significantly from averaged poststimulus activity, t(14) 5 0.61, although power was higher after stimuli onset; (3) during antisaccade trials, prestimulus activity did not differ significantly from averaged poststimulus activity, t(14) 5 1.69, p 5 .113, although power was lower after stimuli onset; and (4) averaged pro- and antisaccade activities during the poststimulus period were significantly different even after adjusting for differences in prestimulus activities, t(14) 5 2.92, p 5 .011. Discussion Executive control involves using learned rules to guide contextappropriate behavioral responses and is supported by PFC (for reviews, see Bunge, 2004; Miller & Cohen, 2001). PFC ostensibly accomplishes this task by influencing activity of cortical and subcortical neurons involved in behavioral regulation via topdown control. Research on cognitive operations that require motor responses for their successful completion typically focuses on manifestations of top-down control in frontal cortical and

subcortical motor output structures. This literature parallels a largely separate one documenting top-down influences on sensory perceptual systems in the control of visual spatial attention (Desimone & Duncan, 1995). The interplay between the topdown regulation of motor processes and external perceptual stimulus registration has been infrequently examined (but see Buschman & Miller, 2007) despite its potential importance for efficiently influencing behavior guided by sensory cues (e.g., Yantis & Jonides, 1996). The present study expands the literature on top-down regulatory control of motor preparation by measuring the neural correlates of visual sensory registration in preparation for antisaccade responses. Antisaccade tasks are well suited for assessing the manifestations of top-down control on both sensory and motor structures in humans, given an extensive understanding of the supporting neurophysiology through primate and human studies, because of their established life span developmental profile (e.g., Luna, Garver, Urban, Lazar, & Sweeney, 2004; Sweeney, Rosano, Berman, & Luna, 2001), their abnormality in psychiatric patients with executive control problems (Harris, Reilly, Keshavan, & Sweeney, 2006; Hutton et al., 2004; McDowell, Myles-Worsley, Coon, Byerley, & Clementz, 1999), and their deficits in patients with focal damage to PFC (Hamilton & Martin, 2005; Pierrot-Deseilligny et al., 2003; Ploner, Gaymard, Rivaud-Pechoux, & Pierrot-Deseilligny, 2005). Primate neurophysiology studies indicate that successful antisaccade performance requires a sustained anticipatory reduction of neural activity in specific regions of saccade motor circuitry prior to stimulus appearance (Everling & DeSouza, 2005), an effect related to top-down control imposed by PFC (Johnston & Ever-

1016

B.A. Clementz et al. side of visual cortex are involved (e.g., Clementz et al., 2007; McDowell et al., 2005) and stronger responses during prosaccade trials when only extrastriate cortex is involved, presumably because of the need to immediately prepare a saccadic response to peripheral visual stimulus (e.g., Dyckman et al., 2007). Increased neural activity in parieto-frontal circuitry during the initial phases of stimulus processing may be related to the extra cognitive and accompanying neurophysiological requirements associated with correct antisaccade performance. On correct antisaccade trials, subjects must suppress a response to a peripheral stimulus, plan a response to an opposite visual field location, and then execute that response. The greater complexity of anti- versus prosaccade tasks can yield greater initial involvement of dorsal visual stream circuitry (e.g., Curtis & D’Esposito, 2003; Dyckman et al., 2007; Ettinger et al., 2008; Ford, Goltz, Brown, & Everling, 2005; Luna et al., 2001; Reynolds & Chelazzi, 2004). In addition, increased activity in PFC, a brain region supporting higher cognitive functions such as attention, planning, spatial orientation, and behavioral restraint (Goldman-Rakic, 1995; Miller & Cohen, 2001), is also frequently reported on correct antisaccade trials (Dyckman et al., 2007; Ettinger et al., 2008; Ford et al., 2005).

Figure 4. Intertrial phase locking (ITPL; top plot) and single trial power (bottom plot) in relation to the flickering checkerboards during the preparatory period prior to the target onset. Time is on the x-axis and standardized ITPL and single trial power are on the y-axis for the top and bottom plots, respectively. The mean (SE) values are shown for 500-ms bins for both pro- (solid) and antisaccade (dashed) trials. The checkerboards began flickering at 0 ms (indicated by the stimuli screen insert and ‘‘Flicker onset’’ label) and flickered through the entire poststimuli onset time illustrated here.

ling, 2006). The results of the present experiment add complementary and unique information to this literature by demonstrating an analogous effect in the human visual system on correct antisaccade trials. Stimulus Onset Effects: Manifestation of Cognitive Complexity When subjects began processing the peripheral flickering stimuli that identified possible peripheral target locations, their initial visual responses (VEPs) were stronger during anti- than prosaccade trials (at 135 and 227 ms) in visual and medial extrastriate, superior parietal, and mesial prefrontal brain regions and stronger during pro- than antisaccade trials (at 196 ms) in lateral extrastriate cortex (in the vicinity of middle occipital gyrus). These cortical activation differences early in sensory stimulus processing as a function of saccadic task demands are consistent with previous studies (e.g., Clementz et al., 2007; Dyckman, Camchong, Clementz, & McDowell, 2007; Everling, Matthews, & Flohr, 2001; McDowell et al., 2005). Indeed, it is typical to find stronger responses on antisaccade trials when brain regions out-

Preparatory Period Effects: Manifestation of Top-Down Control on Visual Cortex After the initial neural response to peripheral stimuli onset, single trial ssVEP power from visual cortex was also strongly modulated as a function of task demands. On prosaccade trials, the amount of neural activity in relation to the peripheral flickering stimuli was high throughout the preparatory interval and tended to increase as time of target onset neared. On antisaccade trials, however, amount of power was consistently lower than during prosaccade trials in relation to the peripheral flickering stimuli, and tended to decrease from the prestimulus period through the poststimulus period. The ssVEP power difference between proand antisaccade trials was especially striking given that (a) physical properties of sensory stimuli at the peripheral target locations were the same throughout the preparatory period and (b) this difference on single trial power was manifest in the context of highly similar ITPL over the course of the preparatory period. Results of the initial response to stimuli onset (VEP) and from the sustained effects captured by ssVEPs provide an illustration that these different approaches to assessing brain activity using EEG data may yield independent and complimentary information (e.g., Clementz et al., 2008; Mu¨ller & Hillyard, 2000). The phasic VEP largely indexed the increased cognitive control requirements associated with antisaccade tasks whereas the tonic ssVEP indexed the manifestation of contextually specific sensory bias signals necessary to support proper antisaccade performance. In the present study, sustained top-down control in relation to the peripheral flickering checkerboards during correct antisaccade performance was manifest as electrocortical suppression of visual cortex activity (reduced single trial power), not as modification of intertrial phase alignment. Alterations in synchronization of neural activity is theoretically a ‘‘gating’’ and/or filtering mechanism for modifying information flow through sensory systems (Hillyard & Anllo-Vento, 1998; Steinmetz et al., 2000) and perhaps for coordinating timing between local and distant neural populations (Sauseng & Klimesch, 2008). During antisaccade tasks, accurately and precisely localizing the peripheral cue is critical to successful performance. Gating inflow of

Neural modulation preceding saccades

1017

relevant sensory signals via altering neural synchronization, thus disrupting fidelity of sensory cue registration, would be counterproductive. Rather, modifying neural gain control seemed to be employed, manifest as changes in the power of neural activity in visual cortex in relation to peripheral stimuli. This strategy may help reduce the probability that neural activity in motor output circuitry will reach the threshold for movement generation in relation to the peripheral cue on an antisaccade trial (e.g., Munoz & Everling, 2004), but still allow for reasonable fidelity of stimulus registration via phase synchronization. This reduction of visual cortex activity, as measured by reduced ssVEP power on antisaccade trials, is reminiscent of similar preparatory reductions seen in saccade motor circuitry (Everling & DeSouza, 2005), an effect attributed to top-down control imposed by PFC (Johnston & Everling, 2006). This attenuation during the preparatory period theoretically reduces the predisposition to movement generation instigated by visual capture and contributes to successful suppression of contextually inappropriate saccades (i.e., error responses) to peripheral targets (Munoz & Everling, 2004). Anticipatory reductions in visual cortex activity based on expected behavioral response demands also parallel results in other literatures demonstrating visual

spatial attentional modulation of context-dependent stimulus processing (Desimone & Duncan, 1995; Kastner & Ungerleider, 2000). This similar finding during an antisaccade task is consistent with a special functional significance associated with topdown regulation of sensory input to the motor systems as a complementary strategy for voluntarily controlling behavioral response generation in a context appropriate manner (see also Buschman & Miller, 2007). In addition, visual cortex (including striate and extrastriate visual regions) has access to the brain stem saccade generators through the superior colliculus (Collins, Lyon, & Kaas, 2005; Lock, Baizer, & Bender, 2003), so it might be expected that the response of neurons in striate and extrastriate regions could facilitate prediction of subsequent saccadic responses. Such effects have been demonstrated during sustained attention-related tasks (e.g. Chawla, Rees, & Friston, 1999; Driver & Frith, 2000; Pessoa, Kastner, & Ungerleider, 2003). It would be important in future studies, by virtue of either selecting for or using manipulations that increase behavior performance variance, to determine whether a quantified indicator of strength of top-down control (e.g., PFC activity) mediates the relationship between the observed anticipatory reductions in visual cortex activity and correct antisaccade responding.

REFERENCES Bunge, S. A. (2004). How we use rules to select actions: A review of evidence from cognitive neuroscience. Cognitive, Affective & Behavioral Neuroscience, 4, 564–579. Buschman, T. J., & Miller, E. K. (2007). Top-down versus bottom-up control of attention in the prefrontal and posterior parietal cortices. Science, 315, 1860–1862. Chawla, D., Rees, G., & Friston, K. J. (1999). The physiological basis of attentional modulation in extrastriate visual areas. Nature Neuroscience, 2, 671–676. Clementz, B. A., Brahmbhatt, S. B., McDowell, J. E., Brown, R., & Sweeney, J. A. (2007). When does the brain inform the eyes whether and where to move? An EEG study in humans. Cerebral Cortex, 17, 2634–2643. Clementz, B. A., Wang, J., & Keil, A. (2008). Normal electrocortical facilitation but abnormal target identification during visual sustained attention in schizophrenia. Journal of Neuroscience, 28, 13411–13418. Colby, C.L, & Goldberg, M. E. (1999). Space and attention in parietal cortex. Annual Review of Neuroscience, 22, 319–349. Collins, C. E., Lyon, D. C., & Kaas, J. H. (2005). Distribution across cortical areas of neurons projecting to the superior colliculus in new world monkeys. Anatomical Record. Part A, Discoveries in Molecular, Cellular, and Evolutionary Biology, 285, 619–627. Collins, D. L., Neelin, P., Peters, T. M., & Evans, A. C. (1994). Automatic 3D intersubject registration of MR volumetric data in standardized Talairach space. Journal of Computer Assisted Tomography, 18, 192–205. Cox, R. W. (1996). AFNI: Software for analysis and visualization of functional magnetic resonance neuroimages. Computers and Biomedical Research, 29, 162–173. Curtis, C. E., & D’Esposito, M. (2003). Persistent activity in the prefrontal cortex during working memory. Trends in Cognitive Sciences, 7, 415–423. Desimone, R., & Duncan, J. (1995). Neural mechanisms of selective visual attention. Annual Review of Neuroscience, 18, 193–222. Ding, J., Sperling, G., & Srinivasan, R. (2006). Attentional modulation of SSVEP power depends on the network tagged by the flicker frequency. Cerebral Cortex, 16, 1016–1029. Di Russo, F., Taddei, F., Apnile, T., & Spinelli, D. (2006). Neural correlates of fast stimulus discrimination and response selection in toplevel fencers. Neuroscience Letters, 408, 113–118. Draganova, R., & Popivanov, D. (1999). Assessment of EEG frequency dynamics using complex demodulation. Physiology Research, 48, 157–165.

Driver, J., & Frith, C. (2000). Shifting baselines in attention research. Nature Reviews. Neuroscience, 1, 147–148. Dyckman, K. A., Camchong, J., Clementz, B. A., & McDowell, J. E. (2007). An effect of context on saccade-related behavior and brain activity. NeuroImage, 36, 774–784. Dyckman, K. A., & McDowell, J. E. (2005). Behavioral plasticity of antisaccade performance following daily practice. Experimental Brain Research, 162, 63–69. Ettinger, U., Ffytche, D. H., Kumari, V., Kathmann, N., Reuter, B., Zelaya, F., et al. (2008). Decomposing the neural correlates of antisaccade eye movements using event-related FMRI. Cerebral Cortex, 18, 1148–1159. Everling, S., & DeSouza, J. F. (2005). Rule-dependent activity for prosaccades and antisaccades in the primate prefrontal cortex. Journal of Cognitive Neuroscience, 17, 1483–1496. Everling, S., Matthews, A., & Flohr, H. (2001). Prestimulus cortical potentials predict the performance in a saccadic distractor paradigm. Clinical Neurophysiology, 112, 1088–1095. Ferree, T. C., Luu, P., Russell, G. S., & Tucker, D. M. (2001). Scalp electrode impedance, infection risk, and EEG data quality. Clinical Neurophysiology, 112, 536–544. Ford, K. A., Goltz, H. C., Brown, M. R., & Everling, S. (2005). Neural processes associated with antisaccade task performance investigated with event-related FMRI. Journal of Neurophysiology, 94, 429–440. Forman, S. D., Cohen, J. D., Fitzgerald, M., Eddy, W. F., Mintun, M. A., & Noll, D. C. (1995). Improved assessment of significant activation in functional magnetic resonance imaging (fMRI): Use of a cluster-size threshold. Magnetic Resonance in Medicine, 33, 636–647. Fuchs, M., Kastner, J., Wagner, M., Hawes, S., & Ebersole, J. S. (2002). A standardized boundary element method volume conductor model. Clinical Neurophysiology, 113, 702–712. Goldman-Rakic, P. S. (1995). Architecture of the prefrontal cortex and the central executive. Annals of the New York Academy of Sciences, 769, 71–83. Haenschel, C., Baldeweg, T., Croft, R. J., Whittington, M., & Gruzelier, J. (2000). Gamma and beta frequency oscillations in response to novel auditory stimuli: A comparison of human electroencephalogram (EEG) data with in vitro models. Proceedings of the National Academy of Sciences, USA, 97, 7645–7650. Ha¨ma¨la¨inen, M. S., & Ilmoniemi, R. J. (1994). Interpreting magnetic fields of the brain: Minimum norm estimates. Medical and Biological Engineering and Computing, 32, 35–42.

1018 Hamilton, A. C., & Martin, R. C. (2005). Dissociations among tasks involving inhibition: A single-case study. Cognitive, Affective and Behavioral Neuroscience, 5, 1–13. Harris, M. S., Reilly, J. L., Keshavan, M. S., & Sweeney, J. A. (2006). Longitudinal studies of saccades in antipsychotic-naive first-episode schizophrenia. Psychological Medicine, 36, 485–494. Hillyard, S. A., & Anllo-Vento, L. (1998). Event-related brain potentials in the study of visual selective attention. Proceedings of the National Academy of Sciences, USA, 95, 781–787. Hutton, S. B., Huddy, V., Barnes, T. R., Robbins, T. W., Crawford, T. J., Kennard, C., et al. (2004). The relationship between antisaccades, smooth pursuit, and executive dysfunction in first-episode schizophrenia. Biological Psychiatry, 56, 553–559. Jammalamadaka, S. R., & SenGupta, A. (2001). Topics in circular statistics. Singapore: World Scientific Press. Johnston, K., & Everling, S. (2006). Neural activity in monkey prefrontal cortex is modulated by task context and behavioral instruction during delayed-match-to-sample and conditional prosaccade-antisaccade tasks. Journal of Cognitive Neuroscience, 18, 749–765. Kastner, S., & Ungerleider, L. G. (2000). Mechanisms of visual attention in the human cortex. Annual Review of Neuroscience, 23, 315–341. Krusemark, E. A., Campbell, W. K., & Clementz, B. A. (2008). Attributions, deception, and event related potentials: An investigation of the self-serving bias. Psychophysiology, 45, 511–515. Lehmann, D., & Skrandies, W. (1984). Spatial analysis of evoked potentials in manFA review. Progress in Neurobiology, 23, 227–250. Lock, T. M., Baizer, J. S., & Bender, D. B. (2003). Distribution of corticotectal cells in macaque. Experimental Brain Research, 151, 455–470. Luna, B., Garver, K. E., Urban, T. A., Lazar, N. A., & Sweeney, J. A. (2004). Maturation of cognitive processes from late childhood to adulthood. Child Development, 75, 1357–1372. Luna, B., Thulborn, K. R., Munoz, D. P., Merriam, E. P., Garver, K. E., Minshew, N. J., et al. (2001). Maturation of widely distributed brain function subserves cognitive development. NeuroImage, 13, 786–793. Makeig, S., Westerfield, M., Jung, T. P., Enghoff, S., Townsend, J., Courchesne, E., et al. (2002). Dynamic brain sources of visual evoked responses. Science, 295, 690–694. Maunsell, J. H., & Treue, S. (2006). Feature-based attention in visual cortex. Trends in Neurosciences, 29, 317–322. McDowell, J. E., Dyckman, K. A., Austin, B. P., & Clementz, B. A. (2008). Neurophysiology and neuroanatomy of reflexive and volitional saccades: Evidence from studies of humans. Brain and Cognition, 68, 255–270. McDowell, J. E., Kissler, J. M., Berg, P., Dyckman, K. A., Gao, Y., Rockstroh, B., et al. (2005). Electroencephalography/magnetoencephalography study of cortical activities preceding prosaccades and antisaccades. NeuroReport, 16, 663–668. McDowell, J. E., Myles-Worsley, M., Coon, H., Byerley, W., & Clementz, B. A. (1999). Measuring liability for schizophrenia using optimized antisaccade stimulus parameters. Psychophysiology, 36, 138–141. Miller, E. K., & Cohen, J. D. (2001). An integrative theory of prefrontal cortex function. Annual Review of Neuroscience, 24, 167–202. Moratti, S., Clementz, B. A., Gao, Y., Ortiz, T., & Keil, A. (2007). Neural mechanisms of evoked oscillations: Stability and interaction with transient events. Human Brain Mapping, 28, 1318–1333. Mu¨ller, M. M., & Hillyard, S. (2000). Concurrent recording of steadystate and transient event-related potentials as indices of visual-spatial selective attention. Clinical Neurophysiology, 111, 1544–1552. Mu¨ller, M. M., & Hubner, R. (2002). Can the spotlight of attention be shaped like a doughnut? Evidence from steady-state visual evoked potentials. Psychological Science, 13, 119–124.

B.A. Clementz et al. Mu¨ller, M. M., Malinowski, P., Gruber, T., & Hillyard, S. A. (2003). Sustained division of the attentional spotlight. Nature, 424, 309–312. Munoz, D. P., & Everling, S. (2004). Look away: The anti-saccade task and the voluntary control of eye movement. Nature Reviews. Neuroscience, 5, 218–228. Nunez, P. L., & Srinivasan, R. (2006). A theoretical basis for standing and traveling brain waves measured with human EEG with implications for an integrated consciousness. Clinical Neurophysiology, 117, 2424–2435. Pascual-Marqui, R. D. (2002). Standardized low-resolution brain electromagnetic tomography (sLORETA): Technical details. Methods and Findings in Experimental and Clinical Pharmacology, 24(Suppl. D), 5–12. Pastor, M. A., Artieda, J., Arbizu, J., Valencia, M., & Masdeu, J. C. (2003). Human cerebral activation during steady-state visual-evoked responses. Journal of Neuroscience, 23, 11621–11627. Pessoa, L., Kastner, S., & Ungerleider, L. G. (2003). Neuroimaging studies of attention: From modulation of sensory processing to topdown control. Journal of Neuroscience, 23, 3990–3998. Pierrot-Deseilligny, C., Muri, R. M., Ploner, C. J., Gaymard, B., Demeret, S., & Rivaud-Pechoux, S. (2003). Decisional role of the dorsolateral prefrontal cortex in ocular motor behaviour. Brain, 126, 1460– 1473. Ploner, C. J., Gaymard, B. M., Rivaud-Pechoux, S., & PierrotDeseilligny, C. (2005). The prefrontal substrate of reflexive saccade inhibition in humans. Biological Psychiatry, 57, 1159–1165. Regan, M. P., & Regan, D. (2003). Techniques for investigating and exploiting nonlinearities in brain processes by recording responses evoked by sensory stimuli. In Z. Lu & L. Kaufman (Eds.), Magnetic source imaging of the human brain (pp. 135–157). Mahwah, NJ: Erlbaum. Reynolds, J. H., & Chelazzi, L. (2004). Attentional modulation of visual processing. Annual Review of Neuroscience, 27, 611–647. Sauseng, P., & Klimesch, W. (2008). What does phase information of oscillatory brain activity tell us about cognitive processes? Neuroscience and Biobehavioral Reviews, 32, 1001–1013. Soufflet, L., & Boeijinga, P. H. (2005). Linear inverse solutions: Simulations from a realistic head model in MEG. Brain Topography, 18, 87–99. Steinmetz, P. N., Roy, A., Fitzgerald, P. J., Hsiao, S. S., Johnson, K. O., & Niebur, E. (2000). Attention modulates synchronized neuronal firing in primate somatosensory cortex. Nature, 404, 187–190. Sweeney, J. A., Luna, B., Keedy, S. K., McDowell, J. E., & Clementz, B. A. (2007). fMRI studies of eye movement control: Investigating the interaction of cognitive and sensorimotor brain systems. NeuroImage, 36(Suppl 2), T54–60. Sweeney, J. A., Rosano, C., Berman, R. A., & Luna, B. (2001). Inhibitory control of attention declines more than working memory during normal aging. Neurobiology of Aging, 22, 39–47. Tenke, C. E., & Kayser, J. (2001). A convenient method for detecting electrolyte bridges in multichannel electroencephalogram and eventrelated potential recordings. Clinical Neurophysiology, 112, 545–550. Wang, J., Clementz, B. A., & Keil, A. (2007). The neural correlates of feature-based selective attention when viewing spatially and temporally overlapping images. Neuropsychologia, 45, 1393–1399. Yantis, S., & Jonides, J. (1996). Attentional capture by abrupt onsets: New perceptual objects or visual masking? Journal of Experimental Psychology. Human Perception and Performance, 22, 1505–1513.

(Received February 10, 2009; Accepted November 5, 2009)

Psychophysiology, 47 (2010), 1019–1027. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01027.x

BRIEF REPORT

Anticipating the consequences of action: An fMRI study of intention-based task preparation

HANNES RUGE,a,b SVEN C. MU¨LLER,c and TODD S. BRAVERd a

Neuroimaging Center, Department of Psychology, Technische Universita¨t Dresden, Dresden, Germany Institute of General Psychology, Biopsychology, and Methods of Psychology, Department of Psychology, Technische Universita¨t Dresden, Dresden, Germany c Section of Developmental and Affective Neuroscience, National Institute of Mental Health, National Institutes of Health, Bethesda, Maryland, USA d Department of Psychology, Washington University in St. Louis, St. Louis, Missouri, USA b

Abstract A key component of task preparation may be to anticipate the consequences of task-appropriate actions. This task switching study examined whether such type of ‘‘intentional’’ preparatory control relies on the presentation of explicit action effects. Preparatory BOLD activation in a condition with task-specific motion effect feedback was compared to identical task conditions with accuracy feedback only. Switch-related activation was found selectively in the effect feedback condition in the middle mid-frontal gyrus and in the anterior intraparietal sulcus. Consistent with research on attentional control, the posterior superior parietal lobule exhibited switch-related preparatory activation irrespective of feedback type. To conclude, preparatory control can occur via complementary attentional and intentional neural mechanisms depending on whether meaningful task-specific action effects lead to the formation of explicit effect representations. Descriptors: Action selection, Action effects, Attention, Cognitive control, Task switching

Our central objective was to isolate brain areas involved in preparatory intentional control processes that serve to disambiguate actions associated with task-ambiguous ‘‘meanings’’ in the sense that they entail different consequences (i.e., they are used for different purposes) depending on the current task context. We were interested in this particular issue for two reasons. First, although the issue of preparatory intentional control may be central to understanding human goal-directed behavior, it has not yet been examined extensively using neuroimaging methods. Second, and more specifically, existing research suggests that preparatory control during task switching is solely attentional in nature, at least when concrete target stimuli are not yet available (Brass et al., 2003; Meiran, 2000; Ruge et al., 2005; Ruge, Braver, & Meiran, 2009). The current study challenges this general conclusion by postulating that task preparation might also operate at the level of intentional action representations, but only when the involved action effects are made sufficiently salient to engage a more explicit internal representation of task-specific effects. We examined this hypothesis by use of a modified version of a spatial task switching paradigm. In the original design (Meiran, 1996), participants had to switch between two spatial discrimination tasks regarding a target stimulus that appeared unpredictably in one out of four positions within a 2 ! 2 grid. One task required judgment regarding the horizontal position of the target (left or right within the grid) whereas the alternative task required judgment regarding the vertical position of the target (up or down within the grid). In this situation, attentional control

In almost any given situation, there are multiple different possible ways to interact with the environment. Thus, actions have to be selected by choosing one out of the available options. Two control processes appear to govern this selective interaction between agent and environment. First, attentional control processes serve as a perceptual filter that can constrain the selection of actions to those most strongly associated with the currently relevant stimulus dimension or feature. Second, intentional control processes can constrain the selection of actions in terms of the effects that will result from them. In other words, the conceptual distinction between attention and intention made here is tightly related to the fundamental distinction between (a) actions directly specified by the appropriate antecedent stimulus conditions as mediated by stimulus–response associations and (b) actions specified by their anticipated consequences as mediated by response–effect associations (de Wit & Dickinson, 2009; Dickinson, 1985; Hommel, Musseler, Aschersleben, & Prinz, 2001; Waszak et al., 2005). The basic design of the present study rests on previous research using the task switching paradigm, an experimental approach that is particularly well suited to creating consistently high demands on both selective attention and selective intention as defined above and elaborated further below. Address correspondence to: Hannes Ruge, Technische Universita¨t Dresden, Fakulta¨t Mathematik und Naturwissenschaften, Fachrichtung Psychologie, 01062 Dresden, Germany. E-mail: ruge@psychologie. tu-dresden.de 1019

1020 is thought to be relevant for selectively activating the stimulus– response associations that are appropriate in the current task. For instance, if the target appears in the upper-left position of the grid and the task is to make a horizontal judgment, spatial attentional mechanisms focus perceptual processing on the horizontal position of the stimulus, so that it is this dimension that triggers the (preexperimentally) associated response, rather than the vertical dimension. One advantage of this paradigm is that it has been used to not only demonstrate the role of attentional control mechanisms during task-switching but also to show that intentional control mechanisms can play a role in response selection and execution as well. Specifically, the relevance of intentional action representations has been revealed by contrasting two conditions that differed with regard to the presence of task-related ambiguity on the level of response meanings (Meiran, 2000). In the ambiguous condition, the same responses were used in both tasks (i.e., left and down stimulus positions both required a left button press response and right and up stimulus positions both required a right button press response). Such a mapping of four stimulus positions onto two responses implies that a given response, for example, a right button press, is ambiguously associated with two task-dependent intentions: Either it can be used to achieve the goal of indicating ‘‘right’’ or it can be used to achieve the goal of indicating ‘‘up.’’ In contrast, the nonambiguous condition was characterized by a unique one-to-one mapping between the four stimulus positions and four distinct responses. Thus, each response was unambiguously associated with one distinct intention. Importantly, it has been shown that response ambiguity is associated with specific behavioral performance costs (Meiran, 2000) as well as with neuroanatomically specific brain activations within the lateral prefrontal cortex (Brass et al., 2003). Yet, in these previous studies, it was assumed that task-specific disambiguation of response meanings only occurs after presentation of the target stimulus and not during the preparation period (in which the upcoming task is known, but the target has not yet appeared). The primary hypothesis of the present study is that task-related disambiguation of response meanings (i.e., intentional control) can occur during preparatory as well as target (imperative) periods. However, intentional action representations might only be engaged for preparatory purposes under conditions in which the meaning of actions are made sufficiently salient. We reasoned that one way to increase salience would be to have task responses result in task-specific and highly plausible perceptual effects. To this end, we modified the task so that correct responses were immediately followed by perceptual motion effects in the direction of the intended target location (cf. Ansorge, 2002; Kiesel & Hoffmann, 2004). We expected that this coupling of responses with a salient and plausible perceptual effect would lead to a stronger and more explicit action–effect associative representation to be formed. Moreover, if action effects were task unique (i.e., upward/downward motions only occurred during the vertical task, and leftward/rightward motions only occurred during the horizontal task), they could be invoked during the preparatory period as a means of reducing interference related to the otherwise task-ambiguous responses (i.e., because the same two responses were used in both tasks). To enable the formation of such unique associations between motion effects and respective tasks, the original spatial target arrangement was slightly modified. For instance, instead of presenting one target square in the upper-left corner of the grid, two target squares were presented,

H. Ruge et al. Accuracy Feedback (correct/incorrect)

Effect Feedback (moving square)

Task cue

Target Response Feedback

Figure 1 Experimental design. Participants performed two blocked task switching conditions involving either ‘‘accuracy feedback’’ or ‘‘effect feedback’’ after responding. On each trial the currently relevant task was indicated by a centrally displayed task cue (‘‘H’’ for horizontal discrimination and ‘‘V’’ for vertical discrimination). For further details, see the Methods.

one left and one up (see Figure 1). After a correct response was made, a central red square appeared to ‘‘jump’’ to the task-appropriate square indicated by the response (e.g., for the horizontal task the central square would jump to the left target square position). To directly test our hypothesis, we compared preparatory brain activation in the novel ‘‘effect feedback’’ condition with preparatory brain activation in the standard control condition presenting ‘‘accuracy feedback’’ only. Critically, both conditions were physically identical until after response execution. Thus, differences in preparatory brain activation can unambiguously be attributed to strategy differences during task preparation. On the basis of prior research, we expected that two brain regions within lateral prefrontal and parietal cortexFthe anterior intraparietal sulcus (aIPS) and the middle mid-frontal gyrus (mMFG)Fwould show selective involvement in preparatory intentional control (i.e., increased activity in the effect feedback relative to accuracy feedback conditions). These two regions have been implicated in intentional control processes under accuracy feedback conditions (Brass et al., 2003), but only when concrete target stimuli were present (i.e., when concrete actions can be planned) and not when advance task information was available for preparation (Ruge et al., 2005, 2009). In contrast to these previous results, we hypothesized that under effect feedback conditions, aIPS and mMFG should be engaged even during the preparation period, before actual task implementation (i.e., active prior to target presentation). Furthermore, we hypothesized that brain regions involved in preparatory attentional control (i.e., selecting the task-appropriate stimulus dimension) should be similarly engaged in both feedback conditions (because there was no difference between conditions with regard to the upcoming target stimuli). Such attentional control regions were expected to be located most prominently in the posterior superior parietal lobule (pSPL; Wager, Jonides, & Reading, 2004). Finally, we were also interested in evaluating the impact of feedback type on behavioral performance. In particular, the theoretical considerations outlined above directly imply the prediction that advance task preparation involving the usage of taskspecific action effect representations in the effect feedback condition should reduce, if not eliminate, residual switch cost as

Anticipating the consequences of action compared to the standard accuracy feedback condition (Meiran, 2000).

Methods Participants Eighteen human participants took part in the functional magnetic resonance imaging (fMRI) study (mean age 5 22 years; age range: 19–29 years; 12 women, 6 men). An additional 30 participants (mean age 5 23 years; age range 20–31 years; 18 women, 12 men) were recruited to perform the behavioral task, but outside of the scanner. All participants gave written informed consent prior to taking part in the experiment. Experimental Design: fMRI Study The fMRI experiment consisted of two different blocked task switching conditions, including (a) a standard control condition in which responses were followed by accuracy feedback and (b) a novel condition designed to increase the salience of task-specific response meanings by presenting task-dependent motion effect feedback. The order of blocks was counterbalanced across participants. In each condition, a practice block of 20 trials was performed before the experimental block started. Except for the use of different types of response feedback, the two blocked conditions were identical in terms of performance demands, as described next. On each trial, participants were presented with a task-ambiguous target stimulus comprised of two empty squares, one located on the horizontal axis of a 2 ! 2 grid and the other one located on the vertical axis (for an exemplary target, see Figure 1). Participants had to indicate the position of only one of these squares, depending on whether they were instructed to perform a horizontal or a vertical discrimination task. In the horizontal discrimination task participants had to indicate whether the target square located on the horizontal axis appeared to the left or right of center by responding with the left or right index finger, respectively. In the vertical discrimination task participants had to indicate whether the target square located on the vertical axis appeared above or below center by responding with the left or right index finger, respectively. Importantly, within this setup the same two manual responses were involved in both tasks; thus, response meanings (i.e., intentional action representations) were task ambiguous. The currently relevant task was indicated by a task cue displayed at the beginning of each trial (‘‘H’’ for horizontal task; ‘‘V’’ for vertical task; centrally displayed on a red square). The two tasks occurred in a pseudorandom and unpredictable sequence, constrained so that the number of task switch trials and task repetition trials was equal. Task sequences were generated using the SeqGen2008 algorithm (Remillard, 2008) so that the number of task switch trials and task repetition trials was equal. The preparation interval between task cue and target stimulus (CTI) varied randomly between 2.5 s and 3.75 s. The task cue remained on screen during the entire preparation interval. Participants had to respond within a window of 1.25 s. Following a response, feedback was displayed immediately for 700 ms. In the accuracy feedback condition, in case of a correct response the central red square turned green, and a check symbol was superimposed. In case of an incorrect or late response, the central red square remained red, and an ‘‘X’’ was superimposed. In the motion effect feedback condition, in case of a correct

1021 response, the central red square disappeared and then reappeared in one of the two peripheral target squares according to the currently relevant task (see Figure 1 for an example). Perceptually, these actions gave the appearance of the red square ‘‘jumping’’ to the location indicated by the participant’s response. In case of an incorrect or late response, the central red square remained stationary with an ‘‘X’’ superimposed. Thus, the only difference between the two conditions was the nature of postresponse feedback on correct response trials. The intertrial interval varied between 2.5 s and 12.0 s, with exponentially decreasing probability of longer intervals (Hagberg, Zito, Patria, & Sanes, 2001). The actual trial onset was randomly jittered by TR/2 (i.e., 1.25 s) relative to the start of fMRI acquisition cycles to double the sampling rate of the trial-related BOLD response (Josephs, Turner, & Friston, 1997). Because the study aimed at comparing preparatory brain activation in the context of accuracy feedback versus motion-effect feedback, we included partial cue-only trials to decorrelate cuerelated and target-related BOLD activation components. Thereby, we were able to obtain separate BOLD response estimates for cue-related and target-related activation (Ollinger, Shulman, & Corbetta, 2001; Shulman et al., 1999). Note that the targetrelated BOLD response estimate also comprises possible activation components elicited by the response or by the feedback. Yet, because we were specifically interested in cue-related preparatory brain activation, disentangling these target-related BOLD response subcomponents was not important, here. Each blocked condition comprised 144 trials, including 96 full cue–target trials and 48 partial cue-only trials. Because responses were only to be made following the target, S1-only trials had no associated task response. The order of condition blocks was counterbalanced across participants. Experimental Design: Behavioral Pilot Study Because the fMRI experiment comprised only 18 participants and the expected behavioral effects are rather weak (i.e., residual switch cost following task preparation), we decided to increase statistical power by also including data from an additional 30 participants who performed the behavioral task outside of the scanner. The task design was the same as that for the scanned participants except for the following differences in procedure. First, there was one short CTI of 100 ms and one long CTI of 1500 ms, instead of two long CTIs of 2500 ms and 3750 ms as realized in the fMRI experiment. Thus, different from the fMRI study in which there was always sufficient time to prepare for the upcoming task, the 100-ms CTI condition prevented participants from full advance task preparation. Thus, we used only the trials with the 1500-ms CTI for analysis. Second, there was a constant intertrial interval of only 300 ms instead of a jittered ITI. Third, there were no partial cue-only trials. Fourth, in the standard feedback condition, the correct–incorrect feedback was provided by presenting the written German words for ‘‘correct’’ and ‘‘incorrect’’ in the center of the screen instead of symbols. Fifth, the pilot experiment was controlled by the ERTS software (BeriSoft) instead of Eprime 1.2. The participants completed 170 trials in each feedback condition, and the order of feedback blocks was counterbalanced across subjects. Imaging Procedure Whole-brain images were acquired on a Siemens 3 Tesla wholebody Trio System (Erlangen, Germany) with a 16-channel circularly polarized head coil. Headphones dampened scanner noise

1022 and enabled communication with the participants. Both structural and functional images were acquired for each participant. Structural images (1.25 mm ! 1 mm ! 1 mm) were acquired using an MP-RAGE T1-weighted sequence (TR 5 9.7 ms, TE 5 4 ms, flip 5 121, TI 5 300 ms). Functional images were acquired using a gradient echo planar sequence (TR 5 2500 ms, TE 5 30 ms, flip 5 901, interleaved slice acquisition, slice gap 5 0). Each volume contained thirty-two 4.0-mm-thick slices (in-plane resolution 3.0 mm ! 3.0 mm). Participants performed a total of eight functional scanning runs, which were separated into two blocks of four runs of each blocked condition (accuracy feedback, motion effect feedback). Each scanning run consisted of 36 trials (in total 144 trials per blocked condition) and lasted approximately 6 min. The experiment was controlled by Eprime 1.2 software (Psychology Software Tools, Inc., Pittsburgh, PA) running on a Windows-XP PC. Stimuli were projected to participants via Visuastim digital goggles (Resonance Technology, Inc., Northridge, CA) simulating a viewing distance of 100 cm. A fiber-optic, light-sensitive key press was used to record participants’ behavioral responses. Data Analysis Preprocessing. The empirical data set was analyzed with SPM5 running under MATLAB 7.1. The preprocessing included slice-time correction, rigid body movement correction (three translation and three rotation parameters), normalization of the functional images by directly registering the mean functional image to the standard MNI EPI template image provided by SPM5 (the resulting interpolated spatial resolution was 3 ! 3 ! 3 mm), and smoothing of the functional images (Gaussian Kernel, FWHM 5 8 mm). Event-related analysis. The preprocessed imaging data were analyzed using the General Linear Model (GLM) approach as implemented in the SPM5 software package. Model regressors were created by convolving neural input functions for the different event types with the assumed canonical hemodynamic response function used by SPM5, including both derivatives. For each condition block, the GLM included two regressors for cuerelated activation separately for task switch and task repetition trials and two regressors for target-related activation separately for task switch and task repetition trials. Thus, a total of eight event-related BOLD responses were to be estimated (plus regressors for the two derivatives). The actual data analysis focused on the four cue-related BOLD estimates (cue–repeat [accuracy feedback], cue–switch [accuracy feedback], cue–repeat [effect feedback], and cue–switch [effect feedback]). These four BOLD estimates were used to compute two whole-brain images that captured the patterns of preparatory brain activation associated with either attentional control or intentional control. Preparatory control demands were expected to be especially high in task switch trials as compared to task repetition trials, reflecting enhanced reconfiguration demands due to proactive interference resulting from implementing the alternative task in the previous trial. Although results from previous fMRI studies on task switching are rather heterogeneous with regard to enhanced switch-related preparatory BOLD activation, we nevertheless focused our primary analysis on this switchrepetition contrast because it allows for a more specific interpretation in terms of task-related preparatory disambiguation processes. Furthermore, from a methodological point of view, by comparing relative BOLD activation differences (switch vs. repeat) across the

H. Ruge et al. two blocked conditions (accuracy vs. effect feedback), we could circumvent the potential problem that baseline differences between blocks (resulting from differential activation during the intertrial interval) might cause apparent differences between conditions that are unrelated to the relevant preparatory processes within the trial itself. Intentional preparatory control, that is, the advance activation of the currently task-relevant response meanings, was hypothesized to be involved specifically in the motion effect feedback condition. We therefore expected enhanced preparatory BOLD activation for switch trials as compared to repetition trials, especially in the effect-feedback condition. Thus, we specifically isolated voxels that exhibited enhanced switch-related preparatory activation in the effect feedback condition more than in the accuracy feedback condition. To this end, we used a twostage procedure in which voxels were first identified at the group level based on the the switchrepetition contrast for the effect feedback condition with po.001 and a minimum of 30 contiguous above threshold voxels. Next, voxel clusters were only included for further analysis if switch-related activity in the effect feedback condition was significantly greater than that in the accuracy feedback condition. This constraint was imposed by applying an inclusive mask based on the interaction contrast ([switch " repetition] effect feedback " [switch " repetition] accuracy feedback) with an intermediate threshold of po.01. In contrast to intentional preparatory control, preparatory attentional control, that is, the advance activation of the currently task-relevant perceptual dimension, was expected for both feedback conditions. Thus, we specifically isolated voxels that exhibited enhanced switch-related preparatory activation irrespective of the feedback condition. To this end, we again used a two-stage masking procedure. First, voxels were identified based on the switchrepetition contrast collapsed across both feedback conditions with po.001 and a minimum of 30 contiguous above threshold voxels. Second, voxel clusters were only included for further analysis if there was no effect of feedback condition. This constraint was imposed by applying an exclusive mask based on the same interaction contrast used above. This masking procedure effectively excluded voxels for which the switch – repetition difference was modulated by the type of feedback using a very lenient whole-brain threshold of po.05, so that voxels showing even subtle effects of feedback type were masked out. The minimum of 30 contiguous above threshold voxels was chosen arbitrarily, after considering (a) the objectively defined cluster size threshold of 44 contiguous voxels as determined based on Gaussian Random Field theory implemented within SPM5 and (b) the often much more lenient, but rather subjectively defined, cluster sizes found in the literature. In order not to ignore potentially relevant activation clusters comprising less than 44 contiguously activated voxels, we arbitrarily lowered the threshold down to 30. When reporting the fMRI results (Table 1 and Table 2), we explicitly indicate the clusters that did not reach the objectively defined 44 voxel threshold.

Results Behavioral Performance Data As described above in the experimental design section, in addition to the behavioral data obtained from the 18 fMRI participants, we also included data from additional 30 subjects who performed the task under unscanned conditions in order to

Anticipating the consequences of action

1023

Table 1. Preparatory Intentional Task Control MNI coordinate Brain region L mMFG L pACC L dPMC L aIPS R aIPS L OCC L OCC R OCC R OCC

Statistics Switch4repetition for effect feedback

x

y

Z

z value

Number of voxels

" 30 "6 " 39 " 51 48 " 36 " 24 39 33

39 18 "6 " 27 " 27 " 63 " 87 " 60 " 75

30 45 51 45 54 " 12 "9 "9 27

3.50 4.31 5.04 4.18 4.29 3.86 4.38 4.33 4.03

52 48 58 159 284 33a 142 338 39a

Note: a: anterior, IPS: intra-parietal sulcus, L: left, m: mid, OCC occipital cortex, p: posterior, MFG: middle frontal gyrus, PMC: premotor cortex, R: right. a Denotes regions that failed to reach the Gaussian random field cluster size threshold of 44 contiguously activated voxels, but reached a more liberal threshold of 30 contiguous voxels.

increase statistical power for identifying relevant behavioral effects of feedback type. Mean response times (RT) and error rates of each subject were entered into two separate four-way analyses of variance (ANOVAs). These ANOVAs included the two factors of primary interest, that is, task transition (task repetition vs. task switch) and feedback type (accuracy vs. effect). To check for possible modulatory effects we included the two additional factors response congruency (the two target squares associated with either the same response or different responses depending on task) and response transition (same vs. different response as compared to preceding trial). The analysis of response times revealed a significant main effect of task transition, F(1,47) 5 16.69, po.001, reflecting the standard residual task-switch cost. More importantly, there was also a significant interaction effect of Task Transition ! Feedback Type, F(1,47) 5 4.30, po.05, reflecting slightly larger residual switch cost for the accuracy feedback condition (repeat 5 496 ms; switch 5 512 ms) relative to the effect feedback condition (repeat 5 502 ms; switch 5 508 ms). The main

Table 2. Preparatory Attentional Task Control MNI coordinate Brain region L aIFG L pre-SMA L/R pSPL L OCC

Statistics Switch4repetition irrespective of feedback type

x

y

Z

z value

Number of voxels

" 36 "9 " 15 "6 24 "6

48 0 " 72 " 60 " 63 " 75

12 60 39 66 57 "6

3.96 3.79 3.98 3.98 3.95 4.42

34a 32a 541

effect of feedback type failed to reach significance, F(1,47) 5 0.02, n.s.). Follow-up tests revealed that residual switch cost in the accuracy feedback condition were highly significant, F(1,47) 5 24.34, po.001, whereas the residual switch cost in the effect feedback condition failed to reach significance, F(1,47) 5 2.40, n.s. This finding confirms the prediction that the presentation of effect feedback would encourage the advance activation of task-specific action effect representations and, thus, reduce residual switch cost. This effect was not significantly modulated by response congruency or response transition. Also, there were no such significant effects regarding error rates (overall error rate was 4%). Notably, the Task Transition ! Feedback Type interaction effect failed to reach significance when evaluated separately for each the two experiments (pilot and fMRI). Yet, importantly, the relevant RT pattern was numerically similar for both experiments. Specifically, in the pilot experiment the residual switch cost was reduced from 18 ms in the accuracy feedback condition to 8 ms in the effect feedback condition. Similarly, in the fMRI experiment, the residual switch cost was reduced from 12 ms in the accuracy feedback condition to 3 ms in the effect feedback condition. These descriptive results indicate that the fMRI-related modifications of the experimental procedure did not alter in a qualitative way the cognitive processes of interest (as reflected by response times). Imaging Data Figure 2 depicts (in yellow) regions exhibiting an activation pattern consistent with intentional preparatory control, that is, stronger switch-related preparatory activation in the effect feedback condition compared to the standard accuracy feedback condition (see Table 1). As predicted, we found activation clusters in the aIPS bilaterally and in the left mMFG. A homologous right mMFG activation cluster comprised only 18 out of the 30 required contiguous above threshold voxels (peak voxel MNI coordinates are 36, 36, and 33; z 5 3.42). Additionally, we found activation clusters including the dorsal premotor cortex (dPMC), the posterior portion of the anterior cingulate cortex (pACC), and regions within the occipital cortex. An important follow-up question concerning this effect-feedback-specific activation pattern is whether it is influenced by the order of condition blocks. In particular, one might suspect that participants who performed the accuracy feedback block first might be less inclined to engage in intentional preparation in the subsequent effect feedback block as compared to participants who started with the effect feedback block. Yet, as Figure 3 shows for two representative brain regions, none of the previously indentified intention-related brain regions was modulated by the order of condition blocks. We also found voxels (Figure 2, colored in pink) that exhibited an activation pattern consistent with attentional preparatory control, that is, switch-related preparatory activation not affected by feedback type (see Table 2). As predicted, we found activation clusters in the pSPL and in the pre-SMA. Additionally, we found activation in the occipital cortex.

48

Note: a: anterior, IFG: inferior frontal gyrus, L: left, OCC: occipital cortex, p: posterior, R: right, SMA: supplementary motor area, SPL: superior parietal lobule. a Denotes regions that failed to reach the Gaussian random field cluster size threshold of 44 contiguously activated voxels, but reached a more liberal threshold of 30 contiguous voxels.

Discussion The aim of the present study was to identify brain areas specifically involved in intention-based task preparation, that is, preparatory control processes that serve to disambiguate taskambiguous action effect representations (i.e., ‘‘response mean-

1024

H. Ruge et al. Switch-related preparatory BOLD activation x=−9

z=51

pre-SMA

pSPL

dPMC

pACC

pSPL aIPS

z=39

x=−31 mMFG dPMC

pACC

mMFG

aIPS

0

2

4

0

6

effect feedback > accuracy feedback

2

4

6

effect feedback = accuracy feedback

Figure 2 Two types of preparatory BOLD activation were identified, associated either with ‘‘intentional preparatory control’’ processes (stronger switch-related activation in the effect feedback condition than in the accuracy feedback condition) colored in red/yellow or with ‘‘attentional preparatory control’’ processes (similar switch-related activation for both feedback types) colored in blue/pink. The depicted brain sections were created with SPM5 within MNI coordinate space.

ings’’) prior to the presentation of concrete target stimuli that trigger response selection and generation processes. We reasoned that intentional preparatory task control is an optional process that may preferentially occur when actions are immediately followed by effects that are task-specific, plausible, and directly perceivable. As explained in detail in the introduction, this is not the case in standard task switching procedures involving only accuracy feedback that is task nonspecific. Thus, we compared this standard control condition with a novel condition in which actions were followed by task-specific motion effects. Importantly, these two conditions were physically identical with regard to the task cue, the preparation interval, and the target stimulus.

Thus, differences in preparatory (i.e., cue-related) brain activation could only be due to differences in the mental representation and anticipation of the upcoming response feedback. Both, the behavioral performance results and the imaging results confirmed our hypothesis. First, the residual switch cost (i.e., switch cost even after ample time to prepare for the current task) was statistically eliminated in the effect feedback condition (6 ms) and was significantly reduced relative to the standard accuracy feedback condition (16 ms). This finding nicely supports a hypothesis suggested earlier (Meiran, 2000), namely, that a substantial portion of the residual switch cost under standard accuracy feedback conditions might indeed be due to insufficiently prepared task-ambiguous response meanings (i.e., representations of task-specific action effects). By contrast, when response meanings are made sufficiently salient as in the present effect feedback condition, task-ambiguous response meanings seem to be disambiguated during the preparation interval, as indicated by much smaller residual switch cost. Although the performance data seem to reflect the impact of intentional task preparation in terms of reduced residual switch cost, the imaging data might directly reflect the upstream preparatory disambiguation of response meanings, which entails the reduced behavioral switch cost. Several brain regions exhibited switch-related preparatory BOLD activation specifically in the effect feedback condition but not in the standard accuracy feedback condition. Importantly, this included the predicted brain areas within lateral prefrontal and parietal cortex: aIPS and mMFG. Furthermore, our results confirmed the role of other fronto-parietal brain regions in attentional control, notably the pSPL and pre-SMA. In line with such an interpretation, in the present study, these regions exhibited switch-related preparatory activation irrespective of the type of feedback. Whereas the involvement of mMFG and aIPS in intentional preparatory control was hypothesized, there were three additional brain regions that exhibited the same activation pattern (pACC, dPMC, and occipital cortex) but were not expected from the outset. Yet, from a broader perspective, the involvement of these regions seems quite plausible. First, both pACC and dPMC have been suggested to be related to action-related processes rather than stimulus-directed (i.e., attentional) ones (Passingham, 1995; Picard & Strick, 1996, 2001). Thus, if task preparation in the effect feedback condition really leads to the activation of action codes via their associated effects, it seems

6

accuracy feedback effect feedback

BOLD activation [arbitray units]

BOLD activation [arbitray units]

No influence of block order on switch-related activation Left mMFG [−30 39 30]

4

2

0

−2

effect fbk. 1st

accuracy fbk. 1st

block order

6

Left aIPS [−51 −27 45] accuracy feedback effect feedback

4

2

0

−2

effect fbk. 1st

accuracy fbk. 1st

block order

Figure 3 Results of a follow-up analysis confirming that effect-feedback-specific switch-related preparatory activation was not influenced by the order of blocks (i.e., effect feedback, then accuracy feedback vs. accuracy feedback, then effect feedback).

Anticipating the consequences of action plausible that this also implicates the engagement of areas that are involved in action planning processes on a more generic level. Notably, there is evidence that specifically the pACC might be related to the coding of actions with regard their consequences, particularly the incentive values of action effects (Rushworth, Buckley, Behrens, Walton, & Bannerman, 2007; Rushworth, Walton, Kennerley, & Bannerman, 2004), suggesting that the pACC might be involved in motivating the execution of actions. Such a view suggests that the task-specific anticipation of action effects, even when these do not involve explicit incentive value, might nevertheless reflect a motivational drive to prepare the currently task-appropriate response options, that is, coding the task-appropriate set of responses with higher motivational priority than the task-inappropriate set of responses even before the upcoming target stimulus enables the ultimate selection and execution of one specific action. Finally, the activation of visual cortex might be related to the visual imagery of the anticipated motion effects. This latter aspect suggests an alternative explanation, namely, that what we called ‘‘intentional preparatory control’’ might in fact be nothing but visual imagery of the expected sensory events without any connection to action-related processes. Yet, we believe that the involvement of the other regions discussed above does, in fact, strongly indicate the engagement of truly action-related preparatory processes in the effect feedback condition as compared to the accuracy feedback condition. Also, the finding that intentionbased preparation reduced behavioral switch cost speaks against the interpretation that we are solely dealing with an epiphenomenon of pure visual expectation without any relationship to action-related preparation. Otherwise, our experimental manipulation should not have been expressed in the observed modulation of behavioral performance. Of course, this does not exclude the possibility that the activation observed in visual cortex by itself might indeed be solely due to sensory (rather than action) expectation. It is worth noting that the observed preparatory effects were characterized by increased activity on task-switch relative to task-repeat trials. Such an activation pattern appears highly plausible, as switch trials seem to demand stronger engagement of cognitive control to overcome the task representations established in the preceding trial. This argument holds for both attentional control engagement with regard to ambiguous stimulus representations and intentional control engagement with regard to ambiguous response meanings. In line with such reasoning, enhanced switch-related preparatory activation has consistently been observed in event-related brain-electrical recordings (Karayanidis, Coltheart, Michie, & Murphy, 2003; Kieffaber & Hetrick, 2005; Rushworth, Hadland, Paus, & Sipila, 2002). By contrast, event-related fMRI study results have been rather heterogeneous with regard to preparatory switch-related activation effects (Badre & Wagner, 2006; Brass & von Cramon, 2002; Braver, Reynolds, & Donaldson, 2003; Bunge, Kahn, Wallis, Miller, & Wagner, 2003; Ruge et al., 2005; Wylie, Javitt, & Foxe, 2006; Yeung, Nystrom, Aronson, & Cohen, 2006). Yet, the studies that do report significant switch-related preparatory BOLD activation have most reliably revealed effects in the pSPL overlapping with the parietal cortex region associated with attentional preparatory control in the present study and consistent with the broader literature on flexible attentional control (Wager et al., 2004). As previous task switching studies used designs that seem to rather discourage the engagement of intentional preparatory control (no explicitly perceivable action effects), in the light

1025 of the present results it does not seem surprising that those previous studies did not reliably report switch-related preparatory activation in the aIPS and the mMFG (i.e., the regions that we found to be specifically associated with intentional preparatory control). From a broader perspective, the involvement of aIPS and mMFG in intentional preparatory control during task switching is consistent with results from two conceptually related research fields. First, studies examining action observation and imitation processes, which tap into action planning processes triggered by the observed action effects caused by other agents, typically discuss the aIPS as one important region (Arbib, 2005; Hamilton & Grafton, 2006; Rizzolatti & Craighero, 2004). Second, the significance of the dorsolateral PFC (i.e., mMFG) has been emphasized in various types of paradigms involving nonroutine action planning processes (Genovesio, Brasted, Mitz, & Wise, 2005; Pochon et al., 2001; Rowe, Toni, Josephs, Frackowiak, & Passingham, 2000; Ruge et al., 2009), especially when actions are ‘‘freely’’ determined by participants without external selection criteria (Frith, Friston, Liddle, & Frackowiak, 1991; Jahanshahi & Dirnberger, 1999; Lau, Rogers, Ramnani, & Passingham, 2004). The present study extends such previous findings by demonstrating that the engagement of these regions (a) can be triggered already during task preparation before the concrete target for action is known, (b) is specifically linked to the availability of explicit action effect information, and (c) preferentially occurs when task-ambiguous action effect representations need to be disambiguated (i.e., under task-switching conditions). Thus, the mMFG and aIPS act together to disambiguate and select the currently appropriate actions based on representations of action goals as defined in the original sense, that is, in terms of the anticipated consequences expected to be achieved by acting in a particular, currently appropriate way. Furthermore, the current study extends and further clarifies the interpretation of results from previous cued task-switching studies involving accuracy feedback only (Brass et al., 2003; Ruge et al., 2009). These studies revealed an enhanced engagement of mMFG, aIPS, or both related to target processing that was even demonstrated to occur in a preparatory manner (Ruge et al., 2009), but when only accuracy feedback was available. These activations related to target presentation were also interpreted as reflecting intention-based as compared to attention-based control. In the light of the present study results, which imply mMFG and aIPS in intention-based preparatory processes specifically in the effect feedback condition, it might appear unclear why these areas should be engaged following target presentation with accuracy feedback only. These apparent contradictions can be resolved when we consider that action effects are implicitly involved also during target processing in these previous studies, but in a relatively implicit and not directly perceivable form. For instance, in Ruge et al. (2009), the implicit effect associated with a left response for the target letter ‘‘N’’ in the ‘‘consonant-vowel’’ task is that the presence of a consonant letter was correctly identified (but not the presence of the concurrently displayed odd digit ‘‘3,’’ as would have been indicated by the same response in case of the ‘‘odd-even’’ task). The implicit nature of such action effects implies that the internal representation of effects would only be (automatically) activated after perceiving the respective target stimulus (e.g., the consonant letter ‘‘N’’) as a given target becomes associated with its effect irrespective of the level of awareness with regard to this effect. By contrast, a task cue that is not directly associated with a specific

1026

H. Ruge et al.

response and a corresponding effect would not automatically activate the task-related set of action effects unless the involved effects are more explicitly represented. Consequently, the rationale behind the present study was to encourage the preparatory engagement of intentional action effect representations upon task cue presentation by making action effects more salient and, thereby, more likely be explicitly used during cuebased task preparation. We believe that this present study designFby explicitly manipulating the type of task-related action effectsFenables a stronger interpretation of mMFG and aIPS activation as being related to intention-based task preparation. By contrast, the previous studies have confounded intentional and attentional task preparation with target- versus cue-related processing. Finally, another critical implication of the current results is that they appear to broaden the notion of intentional control. In particular, the term intentional control is often used to refer to basic motor planning processes when a specific action and the

respective outcome are known in advance (Andersen & Cui, 2009). By contrast, in the present study, the task cue indicated the dimension, or the set of appropriate effects (e.g., leftward or rightward movement in the horizontal direction), to be achieved from a set of possible actions (left or right button presses), rather than one particular action–effect association. We speculate that the role of the mMFG might be to internally represent these higher level associations between action–effect associative sets, rather than between particular actions and particular outcomes. In this way, the intentional control system might operate hierarchically, along a posterior–anterior axis within the lateral PFC, with posterior regions (i.e., dPMC) representing specific actioneffect pairings, whereas more anterior regions (i.e., mMFG) represent action–effect relationships at the set or dimensional level. Thus, the intentional control system might be organized analogously to the types of posterior–anterior hierarchies that have been postulated within attentional control (e.g., Koechlin & Summerfield, 2007).

REFERENCES Andersen, R. A., & Cui, H. (2009). Intention, action planning, and decision making in parietal-frontal circuits. Neuron, 63, 568–583. Ansorge, U. (2002). Spatial intention-response compatibility. Acta Psychologica, 109, 285–299. Arbib, M. A. (2005). From monkey-like action recognition to human language: An evolutionary framework for neurolinguistics. Behavioral and Brain Sciences, 28, 105–124; discussion 125–167. Badre, D., & Wagner, A. D. (2006). Computational and neurobiological mechanisms underlying cognitive flexibility. Proceedings of the National Academy of Sciences, USA, 103, 7186–7191. Brass, M., Ruge, H., Meiran, N., Koch, I., Rubin, O., Prinz, W., et al. (2003). When the same response has different meanings: Recoding the response meaning in the lateral prefrontal cortex. NeuroImage, 20, 1026–1031. Brass, M., & von Cramon, D. Y. (2002). The role of the frontal cortex in task preparation. Cerebral Cortex, 12, 908–914. Braver, T. S., Reynolds, J. R., & Donaldson, D. I. (2003). Neural mechanisms of transient and sustained cognitive control during task switching. Neuron, 39, 713–726. Bunge, S. A., Kahn, I., Wallis, J. D., Miller, E. K., & Wagner, A. D. (2003). Neural circuits subserving the retrieval and maintenance of abstract rules. Journal of Neurophysiology, 90, 3419–3428. de Wit, S., & Dickinson, A. (2009). Associative theories of goal-directed behaviour: A case for animal-human translational models. Psychology Research, 73, 463–476. Dickinson, A. (1985). Actions and habits: The development of behavioural autonomy. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 308, 67–78. Frith, C. D., Friston, K., Liddle, P. F., & Frackowiak, R. S. (1991). Willed action and the prefrontal cortex in man: A study with PET. Proceedings. Biological Sciences, 244, 241–246. Genovesio, A., Brasted, P. J., Mitz, A. R., & Wise, S. P. (2005). Prefrontal cortex activity related to abstract response strategies. Neuron, 47, 307–320. Hagberg, G. E., Zito, G., Patria, F., & Sanes, J. N. (2001). Improved detection of event-related functional MRI signals using probability functions. NeuroImage, 14, 1193–1205. Hamilton, A. F., & Grafton, S. T. (2006). Goal representation in human anterior intraparietal sulcus. Journal of Neuroscience, 26, 1133–1137. Hommel, B., Musseler, J., Aschersleben, G., & Prinz, W. (2001). The theory of event coding (TEC): A framework for perception and action planning. Behavioral and Brain Sciences, 24, 849–878; discussion 878– 937. Jahanshahi, M., & Dirnberger, G. (1999). The left dorsolateral prefrontal cortex and random generation of responses: Studies with transcranial magnetic stimulation. Neuropsychologia, 37, 181–190. Josephs, O., Turner, R., & Friston, K. J. (1997). Event-related fMRI. Human Brain Mapping, 5, 243–248.

Karayanidis, F., Coltheart, M., Michie, P. T., & Murphy, K. (2003). Electrophysiological correlates of anticipatory and poststimulus components of task switching. Psychophysiology, 40, 329–348. Kieffaber, P. D., & Hetrick, W. P. (2005). Event-related potential correlates of task switching and switch costs. Psychophysiology, 42, 56– 71. Kiesel, A., & Hoffmann, J. (2004). Variable action effects: Response control by context-specific effect anticipations. Psychological Research, 68, 155–162. Koechlin, E., & Summerfield, C. (2007). An information theoretical approach to prefrontal executive function. Trends in Cognitive Sciences, 11, 229–235. Lau, H. C., Rogers, R. D., Ramnani, N., & Passingham, R. E. (2004). Willed action and attention to the selection of action. NeuroImage, 21, 1407–1415. Meiran, N. (1996). Reconfiguration of processing mode prior to task performance. Journal of Experimental Psychology: Learning, Memory, and Cognition, 22, 1423–1442. Meiran, N. (2000). Reconfiguration of stimulus task-sets and response task-sets during task-switching. In S. Monsell & J. Driver (Eds.), Control of cognitive processes: Attention and performance XVIII (pp. 377–400). Cambridge, MA: MIT Press. Ollinger, J. M., Shulman, G. L., & Corbetta, M. (2001). Separating processes within a trial in event-related functional MRI I. NeuroImage, 13, 210–217. Passingham, R. E. (1995). The frontal lobes and voluntary action. Oxford, England: Oxford University Press. Picard, N., & Strick, P. L. (1996). Motor areas of the medial wall: A review of their location and functional activation. Cerebral Cortex, 6, 342–353. Picard, N., & Strick, P. L. (2001). Imaging the premotor areas. Current Opinion in Neurobiology, 11, 663–672. Pochon, J. B., Levy, R., Poline, J. B., Crozier, S., Lehericy, S., Pillon, B., et al. (2001). The role of dorsolateral prefrontal cortex in the preparation of forthcoming actions: An fMRI study. Cerebral Cortex, 11, 260–266. Remillard, G. (2008). A program for generating randomized simple and context-sensitive sequences. Behavior Research Methods, 40, 484–492. Rizzolatti, G., & Craighero, L. (2004). The mirror-neuron system. Annual Review of Neuroscience, 27, 169–192. Rowe, J. B., Toni, I., Josephs, O., Frackowiak, R. S., & Passingham, R. E. (2000). The prefrontal cortex: Response selection or maintenance within working memory? Science, 288, 1656–1660. Ruge, H., Brass, M., Koch, I., Rubin, O., Meiran, N., & von Cramon, D. Y. (2005). Advance preparation and stimulus-induced interference in cued task switching: Further insights from BOLD fMRI. Neuropsychologia, 43, 340–355.

Anticipating the consequences of action Ruge, H., Braver, T., & Meiran, N. (2009). Attention, intention, and strategy in preparatory control. Neuropsychologia, 47, 1670–1685. Rushworth, M. F., Buckley, M. J., Behrens, T. E., Walton, M. E., & Bannerman, D. M. (2007). Functional organization of the medial frontal cortex. Current Opinion in Neurobiology, 17, 220–227. Rushworth, M. F., Hadland, K. A., Paus, T., & Sipila, P. K. (2002). Role of the human medial frontal cortex in task switching: A combined fMRI and TMS study. Journal of Neurophysiology, 87, 2577–2592. Rushworth, M. F., Walton, M. E., Kennerley, S. W., & Bannerman, D. M. (2004). Action sets and decisions in the medial frontal cortex. Trends in Cognitive Sciences, 8, 410–417. Shulman, G. L., Ollinger, J. M., Akbudak, E., Conturo, T. E., Snyder, A. Z., Petersen, S. E., et al. (1999). Areas involved in encoding and applying directional expectations to moving objects. Journal of Neuroscience, 19, 9480–9496.

1027 Wager, T. D., Jonides, J., & Reading, S. (2004). Neuroimaging studies of shifting attention: A meta-analysis. NeuroImage, 22, 1679–1693. Waszak, F., Wascher, E., Keller, P., Koch, I., Aschersleben, G., Rosenbaum, D. A., et al. (2005). Intention-based and stimulus-based mechanisms in action selection. Experimental Brain Research, 162, 346–356. Wylie, G., Javitt, D. C., & Foxe, J. J. (2006). Jumping the gun: Is effective preparation contingent upon anticipatory activation in task-relevant neural circuitry? Cerebral Cortex, 16, 394–404. Yeung, N., Nystrom, L. E., Aronson, J. A., & Cohen, J. D. (2006). Between-task competition and cognitive control in task switching. Journal of Neuroscience, 26, 1429–1438. (Received August 18, 2009; Accepted December 16, 2009)

Psychophysiology, 47 (2010), 1028–1039. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01031.x

The involvement of emotion recognition in affective theory of mind

DANIELA MIER,a,b,c STEFANIE LIS,b KERSTIN NEUTHE,b CARINA SAUER,b,c CHRISTINE ESSLINGER,a,b BERND GALLHOFER,b and PETER KIRSCHa,b,c,d a

Division for Imaging in Psychiatry, Central Institute of Mental Health, Mannheim, Germany Centre for Psychiatry and Psychotherapy, University of Giessen, Giessen, Germany Department of Clinical Psychology, Central Institute of Mental Health, Mannheim, Germany d Mannheim School of Medicine, University of Heidelberg, Heidelberg, Germany b c

Abstract This study was conducted to explore the relationship between emotion recognition and affective Theory of Mind (ToM). Forty subjects performed a facial emotion recognition and an emotional intention recognition task (affective ToM) in an event-related fMRI study. Conjunction analysis revealed overlapping activation during both tasks. Activation in some of these conjunctly activated regions was even stronger during affective ToM than during emotion recognition, namely in the inferior frontal gyrus, the superior temporal sulcus, the temporal pole, and the amygdala. In contrast to previous studies investigating ToM, we found no activation in the anterior cingulate, commonly assumed as the key region for ToM. The results point to a close relationship of emotion recognition and affective ToM and can be interpreted as evidence for the assumption that at least basal forms of ToM occur by an embodied, non-cognitive process. Descriptors: Intention recognition, Emotion recognition, Simulation, Mirror neuron system, Embodied perception, Amygdala

direction should culminate in the recognition of intentions. Thus, Brothers’ (1990) concept of social cognition is very close to the original ToM concept developed by Premack and Woodruff (1978), who assume that ToM can be used to make predictions about the behavior of others, which can also be conceptualized as the recognition of intentions. Following this conceptualization of social cognition, emotion recognition in turn can be described as a part of intention recognition or ToM. From an ontogenetic point of view, it has been shown that emotion recognition develops earlier than the ability to mentalize (Montague & Walker-Andrews, 2001; Saxe, Carey, & Kanwisher, 2004). There is also evidence that the ability to recognize emotions is an important component for the development of an understanding of intentionality (Phillips, Wellman, & Spelke, 2002). However, there are only few studies on the relation between emotion recognition and ToM. Most, but not all (see Langdon, Coltheart, & Ward, 2006 and Phillips, MacLean, & Allen, 2002 for divergent results) of these studies, support the idea of such a close relation between emotion recognition and ToM showing a correlation between the performance in emotion recognition and ToM in children as well as in adults (Bora, Vahip, Gonul, Akdeniz, Alkan et al., 2005; Bru¨ne, 2005a; Buitelaar & van der Wees, 1997; Dyck, Pieck, Smith, & Hallmayer, 2006; Henry, Phillips, Crawford, Ietswaart, & Summers, 2006). These results are in line with the model of Coricelli (2005), who defines ToM as a two component process: First, an unconscious automatic process consisting of the basic aspects of intention

The study of social cognition is a rather new but rapidly growing field in the neurosciences. One of the milestones of the field was the paper by Brothers (1990), who defined social cognition as ‘‘the processing of any information which culminates in the accurate perception of the dispositions and intentions of other individuals.’’ (Brothers, 1990, p. 28; see also Brothers, 2002, p. 367). Part of the information used for the recognition of dispositions and intentions are, according to Brothers, identity, category of posture, direction of movement, quality of vocalization, and facial expression. Following this definition, emotion recognition and Theory of Mind (ToM) are two core components of social cognition. Emotion recognition is the ability to infer an emotional state of another individual, mainly from acoustic and visual features like vocalization and facial expression. ToM and its often used synonyms ‘‘mind reading,’’ ‘‘mentalizing,’’ or ‘‘mental state attribution’’ can be defined as the ability to attribute mental states, such as beliefs, desires, and intentions to oneself and others (Frith & Frith, 2001). According to Brothers (1990), the representation of another person’s emotional state and the knowledge about posture and movement

Daniela Mier was supported by a fellowship for doctoral students from the University of Giessen. Address correspondence to: Daniela Mier, Division for Imaging in Psychiatry, Central Institute of Mental Health, J 5, D- 68159 Mannheim, Germany. E-mail: [email protected] or Peter.Kirsch@ zi-mannheim.de 1028

Emotion recognition and affective ToM recognition such as emotion recognition, emotional contagion, and action recognition; second, a conscious process comprising hypotheses testing, which results in a decision about the most likely intention of the other person. To our knowledge, on a neurobiological level, this interconnection between emotion recognition and ToM has not been studied so far, because existing functional imaging studies have investigated both processes only separately. Therefore, the present study was conducted as a first attempt to fill this gap. According to Haxby, Hoffman, and Gobbini (2002), the neuronal basis for emotion recognition consists of a core system for a primary visual analysis of stimuli and an extended system for the analysis of the emotional content. The core system consists of inferior occipital gyrus, lateral fusiform gyrus, and superior temporal sulcus (STS). The extended system consists of limbic areas like amygdala and insula. Most of these regions were identified in a number of neuroimaging studies using functional magnetic resonance imaging (fMRI) (e.g., Gur, Schroeder, Turner, McGrath, Chan, et al., 2002; Hariri, Bookheimer, & Mazziotta, 2000; Ishai, Schmidt, & Boesiger, 2005). In general, there seems to be consistency that these areas are involved in facial emotion recognition (e.g., Adolphs, 2002; Blair, 2003; Haxby et al., 2002). There are two main theoretical accounts on ToM and its underlying neuronal correlates: The more cognitive ‘‘metarepresentational Theory-Theory’’ and the more perceptual ‘‘Simulation-Theory’’ (see Bru¨ne, 2005b, for an overview). Cognitive theories assume that, based on acquired meta-representations about the world, one recognizes the mental state of another person by building a theory about his state. Gallagher and Frith (2003) propose the medial prefrontal cortex, especially the anterior cingulate cortex (ACC), the temporal poles, and the posterior STS with the adjacent temporo-parietal junction as core structures for the generation of these meta-representations of others’ mental states. All these structures are seen to be associated with several sub-functions of mentalizing. The medial prefrontal cortex seems to be the key structure that encodes mental states decoupled from reality (Gallagher & Frith, 2003). The temporal poles are the region where, based on past experience, a wider semantic and emotional context in terms of scripts is processed. This region therefore promotes the understanding of mental states (Frith & Frith, 2003). Activation in the medial prefrontal cortex (e.g., Castelli, Happe´, Frith, & Frith, 2000; Gallagher, Happe´, Brunswick, Fletcher, Frith & Frith, 2000; Gallagher, Jack, Roepstorff, & Frith, 2002; Gallagher & Frith, 2004; Spiers & Maguire, 2006; Vogeley, Bussfeld, Newen, Herrmann, Happe´, et al., 2001) as well as in the temporal poles (Gallagher & Frith, 2004; Spiers & Maguire, 2006; Vo¨llm, Taylor, Richardson, Corcoran, Stirling, et al., 2006) was found in ToM studies using a broad range of different designs. Simulation-theorists suppose that the mental state of others is recognized by a simulation process (e.g., Gallese & Goldman, 1998; Gallese, Keysers, & Rizzolatti, 2004). The related core structures for such a simulation process are located in areas that belong to the so-called mirror neuron system (Gallese, 2007). Another structure that is associated with simulation processes, but does not belong to the mirror neuron system, is the somatosensory cortex (Adolphs, Damasio, Tranel, Cooper, & Damasio, 2000; Avenanti, Bolognini, Maravita, & Aglioti, 2007; Cheng, Yang, Lin, Lee, & Decety, 2008; Winston, O’Doherty, & Dolan, 2003). The mirror neuron system was originally found to be activated in primates not only when performing an action, but also

1029 when observing a similar action being performed by a fellow (di Pellegrino, Fadiga, Fogassi, Gallese, & Rizzolatti, 1992; Gallese, Fadiga, Fogassi, & Rizzolatti, 1996). In the human brain, mirror neurons can be found in Brodmann area 44 (BA 44) in the inferior prefrontal cortex (Rizzolatti, Fogassi, & Gallese, 2002) and in the inferior parietal cortex (Gallese & Goldman, 1998; Keysers & Gazzola, 2006). Activation in the inferior prefrontal gyrus was found during the observation (Buccino, Binkofski, Fink, Fadiga, Fogassi, et al., 2001; Gazzola, Rizzolatti, Wicker, & Keysers, 2007) and imitation of actions (Filimon, Nelson, Hagler, & Sereno, 2007; Iacoboni, Woods, Brass, Bekkering, Mazziotta, & Rizzolatti, 1999) and an important function of this region is the representation of action goals (Koski, Wohlschla¨ger, Bekkering, Woods, Dubeau, et al., 2002). An area that is strictly related to the human mirror neuron system is the STS (Rizzolatti & Craighero, 2004). Neurons in the STS lack motor attributes (Rizzolatti & Craighero, 2004), but are critically involved in imitation processes (Iacoboni, Koski, Brass, Bekkering, Woods, et al., 2001). The STS is regarded as a key region in both approaches: The Theory-Theory and the Simulation-Theory. It seems to play an important role in the initial analysis of visual social cues (for a review, see Allison, Puce, & McCarthy, 2000), in biological motion recognition (Bonda, Petrides, Ostry, & Evans, 1996), in imitation (Iacoboni et al., 2001), and in the recognition of intentions (Pelphrey, Morris, & McCarty, 2004). The STS was found in many ToM studies irrespective of the task design (Baron-Cohen, Ring, Wheelwright, Bullmore, Brammer, et al., 1999; Calder, Lawrence, Keane, Scott, Owen, et al., 2002; Castelli, et al., 2000; Den Ouden, Frith, Frith, & Blakemore, 2005; Hoffman & Haxby, 2000; Majoram, Job, Whalley, Gountouna, McIntosh, et al., 2006; Wicker, Perret, Baron-Cohen, & Decety, 2003). In contrast, the involvement of the inferior frontal gyrus was only found in few studies explicitly investigating ToM (Baron-Cohen, et al., 1999; Farrow, Zheng, Wilkinson, Spence, Deakin et al., 2001; Hooker, Verosky, Germine, Knight, & D’Esposito, 2008; Kim, Kim, Jeong, Ki, Im, et al., 2005; Russell, Rubia, Bullmore, Soni, Suckling, et al., 2000). However, some studies have given further evidence for an involvement of areas associated with the mirror neuron system in ToM. Areas of the mirror neuron system are involved in the prediction of actions in different contexts (Iacoboni, Molnar-Szakacs, Gallese, Buccino, Mazziotta, & Rizzolatti, 2005), their activation can be modulated by motivation (Cheng, Meltzhoff, & Decety, 2007), and activation in some of these areas signals violation of action sequences (Buccino, Baumgaertner, Colle, Buechel, Rizzolatti, & Binkofski, 2007). Beside differences in their assumption about core regions, the Theory-Theory- and the Simulation-Theorists agree that there are additional brain regions that play a role in ToM providing the necessary information to understand emotional mental states when necessary (Frith & Frith, 2001; Gallese et al., 2004). One of the most important of these areas is the amygdala. Lesion studies provide evidence for an involvement of the amygdala in mental state attribution (Shaw, Lawrence, Radbourne, Bramham, Polkey, & David, 2004; Stone, Baron-Cohen, Calder, Keane, & Young, 2003). However, since most of the studies on ToM investigated cognitive ToM tasks and did not include any emotional components in their design, amygdala activation was not reported in those studies. In contrast, when an emotional component was included into the ToM task, amygdala activation was found, as with the recognition of expressive gestures (Gallagher & Frith, 2004) and the recognition of intentional

1030 behavior (Castelli et al., 2000). This also holds true for the recognition of mental states with the Reading the Mind in the Eyes Test (Baron-Cohen et al., 1999), although this result could not be consistently replicated (Russell et al., 2000). Because of the important role of the amygdala in emotion recognition (Adolphs, 2002; Blair, 2003; Haxby, Hoffman, & Gobbini, 2002, see above), these findings again support the assumption of a close relation between emotion recognition and ToM. Furthermore, there are some interesting functional imaging studies on the observation and imitation of emotional expressions that might provide a link between emotion recognition and ToM. These studies found activation in the amygdala as well as in areas of the mirror neuron system during the observation and imitation of emotions in children (Pfeifer, Iacoboni, Mazziotta, & Dapretto, 2008) and in adults (Carr, Iacoboni, Dubeau, Mazziotta, & Lenzi, 2003; Leslie, Johnson-Frey, & Grafton, 2004; Schulte-Ru¨hter, Markowitsch, Fink, & Pifke, 2007; Van der Gaag, Minderaa, & Keysers, 2007). Leslie et al. (2004) showed a common neural basis for imitation of emotional expressions and finger movements in the mirror neuron system. Carr et al. (2003) assume that facial emotions are identified via an interaction of simulation processes of the expression taking place in the mirror neuron system and the activation of the amygdala, which is responsible for processing the emotional content. Taking these findings together, there is convincing evidence that the medial prefrontal gyrus including the ACC and the STS are involved in ToM. The mirror neuron system is involved in the understanding and prediction of simple movements and in the observation and imitation of emotional expressions, and the amygdala is involved in the imitation and recognition of emotions. Moreover, existing data suggest that at least basal processes of ToM such as the recognition of intentions can be localized in areas of the human mirror neuron system. But until now, only little is known about the interplay of emotion recognition and ToM and its neuronal basis. As a first approach to investigate the relationship of ToM and emotion recognition, the aim of the present study was to analyze the relationship of emotion recognition and affective ToM. For this purpose, a new affective ToM task was developed that investigates a basic aspect of affective ToM: The recognition of emotional intentions. By using varying instructions, processes of emotion recognition and affective ToM were induced within an identical experimental setting using the same stimuli and requiring equal motor responses. In light of the existing literature, we assumed that emotion recognition and affective ToM are two closely related processes, with affective ToM being the higher order social cognitive process that is based on emotion recognition. To confirm this assumption on the behavioral and the neural level, we tested the following hypothesis: (1) The close relationship between emotion recognition and affective ToM is reflected by a significant correlation between performance measures during the emotion recognition and the affective ToM task. (2) The recognition of intentions takes longer than the recognition of emotions. (3) On a neural level, the close relationship between emotion recognition and affective ToM is reflected by overlapping activation in structures known to be associated with emotion processing and intention recognition. Therefore, a conjunction analysis of both tasks should reveal significant activation in the amygdala, the inferior prefrontal gyrus, and the STS.

D. Mier et al. (4) During affective ToM, we expected additional activation in the STS, the inferior prefrontal gyrus, the ACC, and the temporal pole compared to emotion recognition because these areas are known to be involved in intention recognition.

Methods Participants Forty voluntary undergraduate students (20 females) participated in the study. Their mean age was 25.25 years (range 19–32 years; SD 5 3.52). All participants were right-handed and had normal or corrected-to-normal vision. Before participating in the study, subjects gave their written informed consent. The study was approved by the local ethics board of the University of Giessen Medical School and conducted in accordance with the Declaration of Helsinki. Experimental Design Stimulus material consisted of 32 portrait photos of eleven actors. The faces depicted an emotional expression of joy, anger, fear, or disgust or a neutral one. The pictures were taken from our own set of facial stimuli, which were modified by means of a morphing-program in order to obtain emotional facial expressions with different intensities by stepwise morphing a neutral and an emotional face of maximal expression. The stimuli used for this study were pictures with 80% intensity, derived from the morphing program (80% from a picture with an emotional expression and 20% from a picture with a neutral expression) to avoid the use of exaggerated emotional expressions and to assure that the task was not too easy. Pictures of seven out of the eleven actors were available for each emotion and four with neutral expressions, resulting in 3 pictures per person and 32 stimuli overall. In a validation study of the stimulus material, the stimuli were presented to 93 undergraduates (age 22.25; range 19–34 years; SD 5 2.9; 48 females). Subjects had to rate the intensity of each of the emotions: fear, anger, happiness, and disgust on a 5-point scale from not present (1) to full present (5) on a computer. An emotional expression was categorized as recognized if the depicted emotion had the highest rating of all emotions under investigation. The mean recognition rate for the emotional expressions used in the present study was 87.5% (SD 5 10.8). The emotional expressions of fear had a mean recognition rate of 81.1% (SD 5 11.3), the emotional expressions of anger of 92.9% (SD 5 7.2), the emotional expression of disgust of 76.3% (SD 5 19.3), and the emotional expression of happiness of 99.8% (SD 5 0.4). The mean rating for the four neutral expressions was 1.3 (SD 5 0.01). The differences in the mean recognition rate of the different emotional categories should not influence our results, because we used the same stimuli in all conditions and did not analyze the emotions separately. The experimental design of the fMRI study consisted of three conditions implemented by different instructions: affective ToM, emotion recognition, and a control condition. Within these conditions, the four different emotional expressions and a neutral facial expression were depicted. The task was to evaluate whether the preceding statement fitted the actual picture. Each trial started with the presentation of a statement, followed by a facial picture, with the choice ‘yes’ or ‘no’ below. The participants had to signal their decision by a button press. The statements

Emotion recognition and affective ToM

1031

Figure 1. Experimental conditions and design, exemplarily shown for the emotion fear.

described an emotional state, an emotional intention, or a physical feature of the depicted person. For the affective ToM and the emotion recognition conditions, four statements were used belonging to four basic emotions: fear, disgust, anger, and joy, (e.g., Ekman, 1994; for an overview, see Ortony & Turner, 1990). The emotional intention recognition always referred to an explicit future action driven by one of the four basic emotions. The prediction of actions and, with this, the recognition of action intentions can be seen as a basic process of ToM (Premack & Woodruff, 1978). Because all the intentions in our task were driven by an emotional state, we refer to them as affective ToM and not as ToM in general. Each statement for the emotion recognition and for the intention recognition condition was presented four times with a picture of a stimulus person with a matching emotion and four times with a picture of a stimulus person with a non-matching emotion. The statements in the control condition were presented in around half of the trials with a picture of a person with a matching physical feature and in the other half of the conditions with a non-matching physical feature. Each picture of the stimulus person was shown once in each condition, resulting in three presentations altogether. An example for the three experimental conditions is given in Figure 1, and all statements used are listed in Table 1. The three experimental conditions consisted of the same stimulus material and were presented in pseudorandomized or-

der. Each statement was shown eight times, resulting in 32 trials per condition and a total of 96 trials. The statements were presented for 2 s, immediately followed by the picture. When no button was pressed, the picture remained for 3 s. A button press terminated the display of the picture and the fixation cross appeared for the remaining trial. The mean inter-trial interval was 4 s (2.5–5.5 s). The total experimental time was approximately 19 min. The experiment was implemented in the Presentation software, version 9.50 (Neurobehavioral Systems, Albany, CA). The stimuli were projected onto a screen behind the scanner, visible for the participants via a mirror fixed atop of the head coil. The total scanning time was about 30 min, including prescans, anatomical scan, and functional imaging. Based on the fMRI findings, we conducted a post-hoc behavioral experiment to investigate whether we find evidence for a facilitation of the recognition of emotional intentions by the recognition of emotions. We hypothesized that emotional intentions are recognized faster if the same person is presented in the trial before with the according emotion than when the same person is presented in the trial before with a neutral expression. In this experiment, we presented the same trials as in the fMRI study but in a different order to 20 student participants (mean age 5 24.25 years; range 20–29; SD 5 2.4; 9 females) outside the scanner. We varied three conditions, in which the emotional intention recognition trials were preceded by a neutral trial, an implicit emotional trial, or an explicit emotional trial. For the neutral trials, the control statements were followed by pictures with neutral facial expressions. For the implicit emotion trials, the control statements were followed by an emotional facial expression. In the explicit emotional trials, emotion recognition statements were followed by an emotional facial expression (for examples of the three conditions, see Figure 2). Each condition was presented 20 times. To avoid a response choice preference for ‘yes’ answers in 12 trials per condition, the trial proceeding the affective ToM trial was congruent and in 8 trials, the proceeding trial was incongruent showing the same person with another emotional expression or asking for another emotion. To test the facilitation hypothesis, we analyzed only the congruent trials assuming that a facilitation process occurs by the presentation of the emotion in the trial preceding the affective ToM trial. fMRI Data Acquisition Data acquisition was accomplished via a 1.5 Tesla GE Signa whole-body magnetic resonance tomograph (General Electrics, Milwaukee, WI). The functional images were acquired with a T2n-weighted gradient echo planar imaging sequence (TR 5 3000 ms; TA 5 100 ms; TE 5 50 ms; flip angle 901; field of view 5 224 mm; 64 ! 64 matrix). Thirty slices per scan were

Table 1. English Translations of the Statements and the Original German Statements for the Three Conditions. German Statements Are Written in Italics

Anger Fear Disgust Joy

Emotion recognition

Intention recognition

Neutral condition

This person is angry Diese Person a¨rgert sich This person is afraid Diese Person fu¨rchtet sich This person is disgusted Diese Person ekelt sich This person is happy Diese Person freut sich

This person is going to bluster Diese Person schimpft gleich This person is going to run away Diese Person la¨uft gleich weg This person is going to avert her face Diese Person wendet sich gleich ab This person is going to cheer Diese Person jubelt gleich

This person is female Diese Person ist weiblich This person has blonde hair Diese Person ist blond This person is older than 30 Diese Person ist a¨lter als 30 This person weighs more than 70 kilos Diese Person wiegt mehr als 70 kg

1032

Figure 2. Experimental design of the behavioral experiment displaying the three experimental conditions (left side) and the emotional intention recognition trial following each condition (right side).

collected ascending in interleaved order with a slice-thickness of 5 mm (3.5 ! 3.5 ! 5 mm voxel size). A total of 387 scans was collected during the experiment. Data analysis. Data analysis was done with SPM2 (Wellcome Department of Cognitive Neuroscience, London, UK). Data preprocessing consisted of slice time correction to the temporally middle slice, realignment, normalization to the standard space of the Montreal Neurological Institute brain (MNI brain), and spatially smoothing with a 10-mm full width at half maximum kernel. A first-level fixed effects analysis was calculated for each person. Regressors for each condition consisting of all correct responses were defined. To account for the residual variance, a regressor of no interest with all trials with wrong answers was defined. A synthetic hemodynamic response function and its spatial and temporal derivatives were used for response modeling. To minimize the influence of movement-related variance, the 6 movement parameters of the realignment procedure were included as covariates of no interest. In a second-level random effects analysis, individual contrasts from the first-level were taken together for the group analysis. Activation in and between the different conditions was analyzed with one-sample t-tests. The resulting p-values of the t-statistic were corrected for multiple comparisons with the familywise er-

D. Mier et al. ror correction method (FWE-corrected). With the resulting tmaps of the one-sample t-tests for affective ToM4Baseline and Emotion recognition4Baseline, a conjunction-analysis for affective ToM and emotion recognition was conducted to reveal common activation in the two conditions. For this purpose, a tmap was calculated that included only areas with significant activation under both conditions. Furthermore, a within-subject one-way analysis of variance (ANOVA) was conducted to identify areas with increasing activation during an increasing requirement to put oneself in the position of the depicted person (ToM4Emotion recognition4Control condition). Regions of Interest (ROIs) for the ACC and BA 44, the key regions for ToM, and for the amygdala, the key region for emotion processing, were extracted from the WFU Pick-Atlas v2.4 (http://www.fmri.wfubmc.edu/cms/software) and edited with MARINA (Walter, Blecker, Kirsch, Sammer, Schienle, et al., 2003). ROI analyses were conducted with the WFU Pick-Atlas. Significance threshold for the analyses was po.05 familywise error corrected (FWE-corrected), for the ROI analyses po.05 FWE-corrected within a ROI and po.001 FWE-corrected within a ROI for the conjunction analysis. The cluster threshold for all analyses was set to 5 contiguous voxels. Performance data and reaction time of the correct answers were analyzed with SPSS Version 13.0 (SPSS Inc., Chicago, IL). Because reaction times and percent correct answers from the fMRI study were not normally distributed, we used non-parametric tests for the analyses. The main effect was analyzed with the Friedman test for repeated measurement on one factor. Differences between the conditions were analyzed with the Wilcoxon test. Correlation analyses were run using Kendall’s tau. Reaction times from the behavioral study were normally distributed and analyzed with a repeated measures ANOVA. Posthoc tests were done with one-sample t-tests. Moreover, effect sizes were calculated for the ANOVA (Z2) and for the post-hoc tests (Cohen’s d). Results fMRI-Study Behavioral Data There was a significant main effect of condition in reaction time (w2 (2) 5 34.4, po.001) with the longest reaction time for the affective ToM condition and the shortest for the neutral condition (Figure 3, right). Post-hoc tests revealed that reaction times in the affective ToM condition were significantly higher than in the emotion recognition condition (Z 5 3.5, po.001,

Figure 3. Behavioral results from the fMRI study. Mean (1/ " standard error) percentage of correct answers (left) and reaction times (right) for the three experimental conditions. Note: All comparisons between conditions are significant (po0.01).

Emotion recognition and affective ToM

1033

Table 2. Areas with Significant Activation During ToM and Emotion Recognition MNI-coordinates Area Postcentral gyrus Precentral gyrus Inferior prefrontal gyrus Culmen Culmen Lingual gyrus Inferior prefrontal gyrus Middle frontal gyrus Inferior prefrontal gyrus Amygdala Superior frontal gyrus Cingulate gyrus Medial frontal gyrus Globus pallidus Substantia nigra Parahippocampal gyrus/Amygdala Globus pallidus Thalamus Superior temporal gyrus Middle frontal gyrus Middle temporal gyrus Inferior parietal lobule

Brodmann area

Cluster size

x

y

z

t-value

BA 3 BA 6 BA 47

2449

! 42 ! 39 ! 39 36 21 6 33 45 54 21 0 6 !3 ! 12 !9 ! 18 15 12 ! 54 ! 51 51 36

! 24 ! 12 18 ! 51 ! 57 ! 84 24 12 21 !6 15 21 !6 0 ! 21 ! 33 0 ! 15 ! 54 42 ! 39 ! 57

60 66 !9 ! 30 ! 21 ! 12 ! 12 33 !3 ! 18 60 42 57 0 ! 18 !6 3 12 3 !3 0 45

15.66 14.89 13.59 15.44 14.08 12.23 14.34 12.91 10.74 11.62 11.28 10.92 8.82 11.03 10.80 9.97 10.81 7.81 10.04 9.76 9.40 8.69

4015 BA 18 BA 47 BA 9 BA 47 BA 6 BA 32 BA 6

1216 29 779 474

BA 30 118 BA 39 BA 47 BA 22 BA 40

28 39 19 12

Note: Conjunction of affective ToM and emotion recognition; po.001, FWE-corrected within the ROI; t-threshold 6.11. Further activation peaks within the significant activated clusters are inserted.

d 5 0.29) and in the control condition (Z 5 4.97, po.001; d 5 0.7) and that reaction times in the emotion recognition condition were higher than in the control condition (Z 5 4.3, po.001; d 5 0.42). The same pattern occurred for the number of correct answers (main effect condition: w2 (2) 5 40.13, po.001, Figure 3, left). Participants made fewer errors in the control condition than either in the affective ToM (Z 5 4.5, po.001; d 5 1.21) or in the emotion recognition condition (Z 5 4.3, po.001; d 5 1.19) and made fewer errors in the emotion recognition condition than in the affective ToM condition (Z 5 3.34, po.01; d 5 0.48). The number of correct responses for affective ToM and emotion recognition were significantly correlated (r 5 .71, po.01), but neither the correlation between correct responses during affective ToM and the control condition (r 5 ! .016, n.s.) nor the correlation of correct responses during emotion recognition and control condition (r 5 ! .030, n.s.) revealed significance. Functional Imaging Data Common activation during emotion recognition and affective Theory of Mind. The conjunction analysis revealed activation in a large network of areas (see Table 2 and Figure 4A and Figure 5, left) including inferior frontal gyrus, STS, inferior parietal lobe, somatosensory cortex, amygdala, basal ganglia, thalamus, and occipital lobe (po.05 FWE-corrected within the ROI; t-threshold 6.11). Affective Theory of Mind specific activation. Comparisons between affective ToM and emotion recognition were computed to detect areas with stronger activation during mentalizing than during emotion recognition and vice versa. There was stronger activation for affective ToM than for emotion recognition in four clusters (po.05, FWE-corrected; t-threshold 5.15): Left inferior frontal gyrus, right temporal pole and bilateral superior temporal sulci extending into the supramarginal gyrus (Table 3, Figure

4B). ROI analyses further revealed an enhanced BOLD-signal in the left amygdala (MNI coordinates: ! 18 ! 6 ! 4; p 5 .015 FWE-corrected within the ROI, Tmax 5 3.50, Cluster size 23 voxels, t-threshold 2.92; see Figure 5, right) and in BA 44 bilaterally (MNI coordinates: ! 54 21 9; po.001, FWE-corrected within the ROI, Tmax 5 4.82, Cluster size 36 voxels, t-threshold 2.95; MNI coordinates: 54 21 9; p 5 .004, FWE-corrected within the ROI, Tmax 5 4.12; Cluster size 16 voxels, t-threshold 3.01). There was no stronger activation during emotion recognition than during affective ToM, even with a lowered significance level at po.001, uncorrected. Activation in the ACC did not differ significantly between conditions (t-threshold left ACC 3.26 and t-threshold right ACC 3.29). Table 4 displays significantly activated voxels (po.05, FWEcorrected; t-threshold 4.73) with strongest activation during affective ToM, medium activation during emotion recognition, and least activation during the control task as revealed by the one-way ANOVA with condition as within-subject factor. These voxels were located in the superior temporal sulci and the left inferior frontal gyrus (Figure 4C). ROI analyses further revealed activation in the left BA 44 (MNI coordinates: ! 54 21 9, p 5 .003, FWE-corrected within the ROI, Tmax 5 3.90, Cluster size 19 voxels, t-threshold 2.84), but not in the ACC (t-threshold left ACC 3.13 and right ACC 3.16), amygdala (2.82 t-threshold) or right BA 44 (t-threshold 2.68). Post-hoc correlations were done to investigate whether higher activation during affective ToM than during emotion recognition may be caused by stimulus presentation time or by task difficulty. For this purpose, simple regressions were conducted between reaction time of the correct answers and brain activation and between percent correct answers and brain activation. However, simple regression revealed no significant association between reaction time and brain activation during affective ToM and during emotion recognition (po.05, FWE-corrected; t-threshold 5.26), the same was true for the percent correct answers (po.05, FWE-

1034

D. Mier et al.

Figure 4. Cortical structures with significant activation during emotion recognition and affective ToM. A: Results of the conjunction analysis, displaying jointly activated clusters during emotion recognition and ToM (po.001, FWE-corrected within the ROI). B: Results of the ANOVA, displaying clusters with strongest activation during affective ToM, and weakest activation during the control condition (po.05, FWE-corrected). C: Clusters with stronger activation during affective ToM than during emotion recognition (po.05, FWE-corrected).

corrected; t-threshold 5.26). Even when using a less stringent correction method for multiple testing, the false discovery rate (FDR) that controls the number of expected false positives within the significantly activated voxels (Genovese, Lazar, & Nichols, 2002), simple regression analyses revealed no significance. The t-threshold using this method was t 5 2.11.

neutral condition was longer than in the implicit emotional (t(19) 5 2.46, p 5 .024; d 5 0.53) and in the explicit emotional trials (t(19) 5 2.17, p 5 .043; d 5 0.52). Reaction time in the two emotional conditions did not differ (t(19) 5 0.22, p 5 .83; d 5 0.03). Figure 6 displays the reaction times for the three conditions.

Behavioral Experiment Analyses of the results of the behavioral experiment revealed a trend for a main effect of condition (F(2,18) 5 2.91, p 5 .081; Z2 5 .19). Post-hoc tests showed that the reaction time in the

Discussion

Figure 5. Significant activation of the amygdala during emotion recognition and affective ToM. Left: Results of the conjunction analysis, displaying joint activation during emotion recognition and affective ToM (po.001, FWE-corrected within the ROI). Right: Results from the ROI-analysis, displaying voxels with stronger activation during affective ToM than during emotion recognition (po.05, FWE-corrected within the ROI, right).

This study aimed to investigate the effects of emotion recognition and affective ToM on brain activation with a new experimental paradigm that allows the study of emotion recognition and the recognition of emotional intentions. We hypothesized that we would find overlapping brain regions activated during both emotion recognition and affective ToM and additional activation in areas associated with intention recognition during affective ToM. On the behavioral level, we expected a correlation between performance during emotion recognition and affective ToM with longer reaction times for affective ToM than for emotion recognition. In order to differentiate between emotion recognition and affective ToM, the study used identical stimulus material and identical response alternatives for all conditions but different instructions. The emotion recognition task demanded a decision about the emotion, the affective ToM task about the intention, and the control task about a physical feature of the depicted person.

Emotion recognition and affective ToM

1035

Table 3. Areas with Higher Activation During Affective ToM than During Emotion Recognition MNI-coordinates Area Middle temporal gyrus Inferior temporal gyrus Inferior prefrontal gyrus Superior temporal gyrus Angular gyrus Inferior parietal lobe Supramarginal gyrus Middle temporal gyrus

Brodmann area

Cluster size

x

y

z

t-value

BA 21 BA 20 BA 47 BA 39 BA 39 BA 40 BA 40 BA 39

47

54 60 ! 45 57 54 45 ! 57 ! 42

9 !6 24 ! 57 ! 63 ! 66 ! 57 ! 60

! 30 ! 24 !6 21 27 42 21 24

6.60 5.62 6.50 5.93 5.69 5.65 5.81 5.43

81 91 43

Note: (Affective ToM4Emotion recognition; po0.05, FWE-corrected; t-threshold 5.15). Further activation peaks within the significantly activated clusters are inserted.

In line with most of the previous studies (Bora et al., 2005; Bru¨ne, 2005a; Buitelaar & van der Wees, 1997; Dyck et al., 2006; Henry et al., 2006), there was a positive correlation between the amount of correctly recognized emotions and correctly recognized intentions indicating a relationship between these two processes. This relationship between emotion recognition and affective ToM could also be observed in the fMRI data in terms of overlapping brain activity in the two conditions. A conjunction analysis revealed common activation bilaterally in superior temporal sulci, inferior frontal gyri reaching into the insula, in the globus pallidus and the amygdala, right middle frontal gyrus, and left somatosensory cortex and thalamus. These results are in accordance with studies on imitation and observation of emotional expressions, which found activation in the amygdala and in areas of the mirror neuron system during the observation of emotional facial expressions (Carr et al., 2003; Leslie et al., 2004; Schulte-Ru¨hter et al., 2007; Van der Gaag et al., 2007) as well as with studies showing the importance of the somatosensory cortex for the recognition of emotions (Adolphs et al., 2000; Hennenlotter, Schroeder, Erhard, Castrop, Haslinger, et al., 2005; Pitcher, Garrido, Walsh, & Duchaine, 2008; Pourtois, Sander, Andres, Grandjean, Reveret, et al., 2004; Winston et al., 2003). The activation in structures associated with the mirror neuron system, as well as in the insula and amygdala, can be interpreted in terms of action representation as a necessary process for the understanding of emotions in others (Carr et al., 2003). According to this assumption, emotions of others are understood by mental simulation taking place in the mirror neuron system. This information is then relayed by the insula to the amygdala where the emotional content is processed. An explanation that goes beyond this interpretation is that the representation of emotional expressions also preactivates the neuronal network necessary for emotional intention recognition

and thus facilitates the recognition of the emotional intentions. This interpretation is in line with the result that only areas already activated during emotion recognition showed stronger activation during affective ToM. No additional areas were identified as activated during affective ToM compared to emotion recognition. Moreover, in the confirmatory behavioral experiment, we found shorter reaction times during emotional intention recognition when the trials were preceded by a trial with an emotional expression compared to those trials preceded by a trial with a neutral expression. This effect was independent from the emotional expression presented explicitly in an emotion recognition context or implicitly in a neutral context. The results from the behavioral level support our assumption that emotion recognition preactivates the network necessary for emotional intention recognition and thus facilitates affective ToM. Unlike previous studies on ToM, we found no activation of the medial prefrontal cortex/ACC but in the inferior prefrontal gyrus, the STS, and the temporal pole. This holds true for the conjunction analysis, the contrast between affective ToM and emotion recognition, and the comparison of all conditions. Activation in the ACC was linked to cognitive processes related to the presence of conflicting response alternatives (Paus, 2001). In the context of ToM, this process is described as a decoupling mechanism in the ACC, which enables mental states to be rep-

Table 4. Clusters with Strongest Activation During Affective ToM and Weakest Activation During the Control Condition

Area Middle temporal gyrus Inferior prefrontal gyrus Middle temporal gyrus

Brodmann Cluster area size BA 39 BA 47 BA 39

62 46 50

MNIcoordinates x

y

z t-value

! 57 ! 54 9 ! 54 30 0 57 ! 63 9

5.82 5.80 5.69

Note: (Affective ToM4Emotion recognition4Control; po0.05, FWEcorrected; t-threshold 4.73).

Figure 6. Results from the behavioral experiment: Reaction times (means and standard error) during emotional intention recognition for the three preceding conditions.

1036 resented independent of reality (Gallagher & Frith, 2003). This decoupling is especially necessary in classical ToM tasks, such as false belief tasks where one has to recognize that a person has a mental state about the world that differs from the real world. Activation in the inferior prefrontal gyrus and in the STS, however, is associated with the attribution of intentions and intentional movements (Castelli et al., 2000; Pelphrey et al., 2004; Winston, Strange, O’Doherty, & Dolan, 2002), the recognition of biological movements (Bonda et al., 1996), the identification of complex goal-directed motions (Schultz, Imamizu, Kawato, & Frith, 2004), and the representation of action goals (Koski et al., 2002). The involvement of areas associated with the human mirror neuron system and the missing activation of the ACC in the present study are in good accordance with the view of simulation theorists like Gallese (2006), arguing that the attribution of intentions occurs by default via embodied simulation rather than higher cognitive mechanisms. He states that ‘‘. . . action prediction and the ascription of intentionsFat least of simple intentionsFdo not appear to belong to different cognitive realms, but both pertain to embodied simulation mechanisms underpinned by the activation of chains of logically related mirror neurons’’ (Gallese, 2006, p. 18). This interpretation is also in line with the additional activation of the temporal poles during affective ToM in comparison to emotion recognition. Activation in the temporal poles is thought to provide a framework of associated emotional scripts for ToM (Frith & Frith, 2003), which allows the simulation of possible actions based on the knowledge of the emotion. On the basis of the lack of activation in the ACC, but the involvement of areas associated with the mirror neuron system and the longer reaction times for affective ToM than for emotion recognition, we hypothesize that emotional intention recognition occurred by an additive simulation mechanism: The recognition of the emotion by a mental simulation of the emotional expression and the recognition of the emotional intention by the mental simulation of the emotional expression together with the additional mental simulation of possible actions. The STS and the inferior prefrontal gyrus seem to be critically involved in emotion recognition and recognition of emotional intentions, resulting in activation during both processes and additional activation during the recognition of the intentions. Thus, the facilitation of the recognition of emotional intentions by emotion recognition seems to occur by two reasons: (1) the need of the mental simulation of the emotion to infer the correct intention, and with this (2) a pre-activation of those areas necessary for the simulation of prospective actions. However, these conclusions have to remain speculative, because the existence of mirror neurons in the human brain is still controversial and we cannot demonstrate a simulation process, but only assume it. Another interpretation that is possible and does not rely on the assumption of a simulation mechanism, either explicit or implicit, is the idea of direct perception (Gallagher, 2008). Gallagher (2008) proposes that in daily interactions all information necessary to understand emotions as well as the intention of others is visible and can be assessed by direct perception: ‘‘When I see the other’s action or gesture, I see (I immediately perceive) the meaning in the action or gesture. I see the joy or I see the anger, or I see the intention in the face or in the posture or in the gesture or action of the other’’ (Gallagher, 2008, p. 542). In this case, the increased activation and longer reaction times during affective ToM in comparison to emotion recognition can be explained by a more complex and probably also additional perceptional process.

D. Mier et al. Another possibility to explain the activation differences between emotion recognition and affective ToM that is in line with both the direct perception approach as well as the simulation approach is priming by the statements. If direct perception is smart (as Gallagher, 2008 proposes) and has access to concepts of mental states, the statements might prime a perceptual network, e.g., how someone looks who is going to bluster that has to be matched with the following stimulus. On the other hand, simulation theorists propose that simulation processes in the mirror neuron processes can predict action goals (Kilner, Friston, & Frith, 2007). The priming by the statements might have elicited according simulation processes in the mirror neuron system, and the result is matched with a simulation of the actual expression of the shown person. Moreover, direct perception as well as simulation should rely on the amygdala, which enriches the mere action representation/neutral perception with emotional information. Taken together, if we adhere to the literature ascribing more conscious cognitive processes to the ACC and more unconscious automatic processes to STS, inferior prefrontal gyrus, temporal poles, and amygdala, it is insignificant how we interpret the activation: In any case, the activation pattern we found points to the fact that an embodied process occurred during emotion recognition as well as affective ToM. However, these interpretations cannot rule out the possibility that if one has been confronted with an ambiguous complex situation such embodied processes would fail, and an involvement of the ACC would become necessary. An unexpected result was the stronger activation of the left amygdala during affective ToM compared to emotion recognition. Activation in the right amygdala during emotion recognition and affective ToM can be attributed to emotional resonance (Carr et al., 2003), while the stronger activation during affective ToM than during emotion recognition in the left amygdala seems to be related to the active conscious processing of the intention (Castelli et al., 2000). In a study by Shaw et al. (2004), patients with early amygdala damage showed stronger ToM impairment than patients with adulthood amygdala damage. Interestingly, the early damage group mainly consisted of patients with lesions of the left amygdala and the adulthood group of patients with right-sided lesions, pointing to the possibility of lateralization effects. Gre`zes, Berthoz, and Passingham (2006) assume that the amygdala is especially important for the estimation of threat against oneself. This points to the possibility that attributing the intention of another person during the affective ToM task was associated with an increased self-reference in comparison to the attribution of the emotional state. Altogether, these results support the assumption that, in general, the amygdala is involved in mentalizing and that the left amygdala has a specific role for the attribution of emotional intentions. To rule out the possibility that the higher activation during affective ToM is simply due to the longer stimulus presentation, a simple regression with brain activation and reaction time (as a measure of stimulus duration) was conducted. We found no area where the duration of stimulus presentation was significantly correlated with brain activation, neither in the emotion recognition condition nor in the affective ToM condition. Moreover, to pursue the possibility that the differential brain activation between affective ToM and emotion recognition is due to the differences in task difficulty, we calculated a simple regression with brain activation and percent correct answers within the two conditions. This regression also revealed no significant result. Therefore, the stronger activation during the affective ToM task

Emotion recognition and affective ToM

1037

seems not to be caused by stimulus presentation duration or task difficulty, but to the additional processes involved in ToM. The study has several shortcomings. The generalizability of our results is reduced by the fact that we investigated only affective ToM processes. This lack of a non-affective ToM condition is a shortcoming, and it will be an important task for future studies to identify those regions of the social cognition network that are involved in both emotional and non-emotional intention recognition. Furthermore, future studies should not only investigate the relationship of affective and non-affective ToM, but also the relationship with empathy, a closely related process. There are several lines of evidence for an association between empathy and activation in the mirror neuron system (e.g., Carr et al., 2003; Schulte-Ru¨hter et al., 2007). Another shortcoming rising from the experimental design using the same stimuli in all conditions is the use of emotional faces in the neutral control condition, making it an implicit emotion recognition task and thus more difficult to find differences in brain activation, especially in areas of the limbic system like the amygdala. Moreover, conditions differed in their

difficulty, as evident from the reaction times and percent correct answers, with a ceiling effect in the control condition. Therefore, we analyzed the correlation between the behavioral indices of task difficulty and brain activation in the emotion recognition and affective ToM condition and found no relationship. Nevertheless, this cannot rule out the possibility that there were at least subtle influences of task difficulty on brain activation. Due to the ceiling effect in the control condition, we could not examine the relationship between task performance and brain activation for this condition. Moreover, it would be interesting to test in further studies with more trials per emotion condition whether the type of emotion differentially influences activation during affective ToM. In conclusion, with the present study using a novel affective ToM task, we could demonstrate that emotion recognition and affective ToM are two closely related components of social cognition. Both processes share a common neuronal network of areas assumed to be involved in embodied simulation or perception processes as core regions: the amygdala and areas belonging to the mirror neuron system.

REFERENCES Adolphs, R., Damasio, H., Tranel, D., Cooper, G., & Damasio, A. R. (2000). A role for somatosensory cortices in the visual recognition of emotion as revealed by three-dimensional lesion mapping. Journal of Neuroscience, 20, 2683–2690. Adolphs, R. (2002). Neural systems for recognizing emotion. Current Opinion in Neurobiology, 12, 169–177. Allison, T., Puce, A., & McCarthy, G. (2000). Social perception from visual cues: Role of the STS region. Trends in Cognitive Siences, 4, 267–278. Avenanti, A., Bolognini, N., Maravita, A., & Aglioti, S. M. (2007). Somatic and motor components of action simulation. Current Biology, 17, 2129–2135. Baron-Cohen, S., Ring, H. A., Wheelwright, S., Bullmore, E. T., Brammer, M. J., Simmons, A., et al. (1999). Social intelligence in the normal and autistic brain. European Journal of Neuroscience, 11, 1891–1898. Blair, R. J. R. (2003). Facial expressions, their communicatory functions and neuro-cognitive substrates. Philosophical Transactions B, 358, 561–572. Bonda, E., Petrides, M., Ostry, D., & Evans, A. (1996). Specific involvement of human parietal systems and the amygdala in the perception of biological motion. The Journal of Neuroscience, 16, 3737–3744. Bora, E., Vahip, S., Gonul, A. S., Akdeniz, F., Alkan, M., Ogut, M., et al. (2005). Evidence for theory of mind deficits in euthymic patients with bipolar disorder. Acta Psychiatrica Scandinavica, 112, 110–116. Brothers, L. (1990). The social brain: A project for integrating primate behavior and neurophysiology in a new domain. Concepts in Neuroscience, 1, 27–51. Brothers, L. (2002). The social brain: A project for integrating primate behavior and neurophysiology in a new domain. In J. T. Cacioppo, G. G. Berntson, R. Adolphs, C. S. Carter, R. J. Davidson, & M. K. McClintock, et al. (Eds.), Foundations in social neuroscience (pp. 367– 386). Cambridge, MA: MIT Press. Bru¨ne, M. (2005a). Emotion recognition, ‘theory of mind’, and social behavior in schizophrenia. Psychiatry Research, 133, 135–147. Bru¨ne, M. (2005b). ‘‘Theory of mind’’ in schizophrenia: A review of the literature. Schizophrenia Bulletin, 31, 21–42. Buccino, G., Baumgaertner, A., Colle, L., Buechel, C., Rizzolatti, G., & Binkofski, F. (2007). The neural basis for understanding non-intended actions. Neuroimage, 36, T119–T127. Buccino, G., Binkofski, F., Fink, G. R., Fadiga, L., Fogassi, L., Gallese, V., et al. (2001). Action observation activates premotor and parietal areas in a somatotopic manner: An fMRI study. European Journal of Neuroscience, 13, 400–404. Buitelaar, J. K., & van der Wees, M. (1997). Are deficits in the decoding of affective cues and in mentalizing abilities independent? Journal of Autism and Developmental Disorders, 27, 539–556.

Calder, A. J., Lawrence, A. D., Keane, J., Scott, S. K., Owen, A. M., Christoffels, I., et al. (2002). Reading the mind from eye gaze. Neuropsychologia, 40, 1129–1138. Carr, L., Iacoboni, M., Dubeau, M., Mazziotta, J. C., & Lenzi, G. L. (2003). Neural mechanisms of empathy in humans: A relay from neural systems for imitation to limbic areas. PNAS, 100, 5497–5502. Castelli, F., Happe´, F., Frith, U., & Frith, C. (2000). Movement and mind: A functional imaging study of perception and interpretation of complex intentional movement patterns. NeuroImage, 12, 314–325. Cheng, Y., Meltzhoff, A. N., & Decety, J. (2007). Motivation modulates the activity of the human mirror-neuron system. Cerebral Cortex, 17, 1979–1986. Cheng, Y., Yang, C., Lin, C., Lee, P., & Decety, J. (2008). The perception of pain in others suppresses somatosensory oscillations: A magnetoencephalography study. NeuroImage, 40, 1833–1840. Coricelli, G. (2005). Two levels of mental state attribution: From automaticity to voluntariness. Neuropsychologia, 43, 294–300. Den Ouden, H. E. M., Frith, U., Frith, C., & Blakemore, S.-J. (2005). Thinking about intentions. NeuroImage, 28, 787–796. di Pellegrino, G., Fadiga, L., Fogassi, L., Gallese, V., & Rizzolatti, G. (1992). Understanding motor events: A neurophysiological study. Experimental Brain Research, 91, 176–180. Dyck, M. J., Pieck, J. P., Smith, L., & Hallmayer, J. (2006). Are abilities abnormally interdependent in children with autism? Journal of Clinical Child and Adolescent Psychology, 35, 20–33. Ekman, P. (1994). All emotions are basic. In P. Ekman & R. J. Davidson (Eds.), The nature of emotion: Fundamental questions (pp. 15–19). New York: Oxford University Press. Farrow, T. F. D., Zheng, Y., Wilkinson, I. D., Spence, S. A., Deakin, J. F. W., Tarrier, N., et al. (2001). Investigating the functional anatomy of empathy and forgiveness. NeuroReport, 12, 2433–2438. Filimon, F., Nelson, J. D., Hagler, D. J., & Sereno, M. I. (2007). Human cortical representations for reaching: Mirror neurons for execution, observation, and imagery. NeuroImage, 37, 1315–1328. Frith, C. D., & Frith, U. (2001). The biological basis for social interaction. Current Directions in Psychological Science, 10, 151–155. Frith, C. D., & Frith, U. (2003). Development and neurophysiology of mentalizing. Philosophical Transactions of the Royal Society of London, Series B: Biological Sciences, 358, 449–473. Gallagher, S. (2008). Direct perception in the intersubjective context. Consciousness and Cognition, 17, 535–543. Gallagher, H. L., Happe´, F., Brunswick, N., Fletcher, P. C., Frith, U., & Frith, C. D. (2000). Reading the mind in cartoons and stories: An fMRI of ‘theory of mind’ in verbal and nonverbal tasks. Neuropsychologia, 38, 11–21.

1038 Gallagher, H. L., Jack, A. I., Roepstorff, A., & Frith, C. D. (2002). Imaging the intentional stance in a competitive game. NeuroImage, 16, 814–821. Gallagher, H. L., & Frith, C. (2003). Functional imaging of ‘theory of mind’. Trends in Cognitive Sciences, 7, 77–83. Gallagher, H. L., & Frith, C. D. (2004). Dissociable neuronal pathways for the perception and recognition of expressive and instrumental gestures. Neuropsychologia, 42, 1725–1736. Gallese, V., Fadiga, L., Fogassi, L., & Rizzolatti, G. (1996). Action recognition in the premotor cortex. Brain, 199, 593–609. Gallese, V., & Goldman, A. (1998). Mirror neurons and the simulation theory of mind-reading. Trends in Cognitive Sciences, 2, 493–501. Gallese, V., Keysers, C., & Rizzolatti, G. (2004). A unifying view of the basis of social cognition. Trends in Cognitive Sciences, 8, 396–403. Gallese, V. (2006). Intentional attunement: A neurophysiological perspective on social cognition and its disruption in autism. Brain Research, 1079(1), 15–24. Gallese, V. (2007). Before and below ‘theory of mind’: Embodied simulation and the neural correlates of social cognition. Philosophical Transactions of the Royal Society of London, Series B: Biological Sciences, 362, 659–669. Gazzola, V., Rizzolatti, G., Wicker, B., & Keysers, C. (2007). The anthropomorphic brain: The mirror neuron system responds to human and robotic actions. NeuroImage, 35, 1674–1684. Genovese, C. R., Lazar, N. A., & Nichols, T. (2002). Thresholding of statistical maps in functional neuroimaging using the false discovery rate. NeuroImage, 15, 870–878. Gre`zes, J., Berthoz, S., & Passingham, R. E. (2006). Amygdala activation when one is the target of deceit: Did he lie to you or to someone else? NeuroImage, 30, 601–608. Gur, R. C., Schroeder, L., Turner, T., McGrath, C., Chan, R. M., Turetsky, B. I., et al. (2002). Brain activation during facial emotion processing. NeuroImage, 16, 651–662. Hariri, A. R., Bookheimer, S. Y., & Mazziotta, J. C. (2000). Modulating emotional responses: Effect of a neocortical network on the limbic system. NeuroReport, 11, 43–48. Haxby, J. V., Hoffman, E. A., & Gobbini, M. I. (2002). Human neural systems for face recognition and social communication. Biological Psychiatry, 51, 59–67. Hennenlotter, A., Schroeder, U., Erhard, P., Castrop, F., Haslinger, B., Stoecker, D., et al. (2005). A common neural basis for receptive and expressive communication of pleasant facial affect. NeuroImage, 26, 581–591. Henry, J. D., Phillips, L. H., Crawford, J. R., Ietswaart, M., & Summers, F. (2006). Theory of mind following traumatic brain injury: The role of emotion recognition and executive functions. Neuropsychologia, 44, 1623–1628. Hoffman, E. A., & Haxby, J. V. (2000). Distinct representations of eye gaze and identity in the distributed human neural system for face perception. Nature Neuroscience, 3, 80–84. Hooker, C. I., Verosky, S. C., Germine, L. T., Knight, R. T., & D’Esposito, M. (2008). Mentalizing about emotion and its relationship to empathy. Social Cognitive and Affective Neuroscience, 3, 204–217. Iacoboni, M., Koski, L. M., Brass, M., Bekkering, H., Woods, R. P., Dubeau, M., et al. (2001). Reafferent copies of imitated actions in the right superior temporal cortex. PNAS, 98, 13995–13999. Iacoboni, M., Woods, R. P., Brass, M., Bekkering, H., Mazziotta, J. C., & Rizzolatti, G. (1999). Cortical mechanisms of human imitation. Science, 286, 2526–2528. Iacoboni, M., Molnar-Szakacs, I., Gallese, V., Buccino, G., Mazziotta, J. D., & Rizzolatti, G. (2005). Grasping the intentions of others with one‘s own mirror neuron system. PLoS Biology, 3, 529–535. Ishai, A., Schmidt, C. F., & Boesiger, P. (2005). Face perception is mediated by a distributed cortical network. Brain Research Bulletin, 67, 87–93. Keysers, C., & Gazzola, V. (2006). Towards a unifying neural theory of social cognition. Progress in Brain Research, 156, 379–491. Kilner, J. M., Friston, K. J., & Frith, C. D. (2007). The mirror-neuron system: A Bayesian perspective. NeuroReport, 18, 619–623. Kim, J., Kim, J., Jeong, B. S., Ki, S. W., Im, D., Lee, S. J., et al. (2005). Neural mechanism for judging the appropriateness of facial affect. Cognitive Brain Research, 25, 659–667. Koski, L., Wohlschla¨ger, A., Bekkering, H., Woods, R. P., Dubeau, M., Mazziotta, J. C., et al. (2002). Modulation of motor and premotor

D. Mier et al. activity during imitation of target-directed actions. Cerebral Cortex, 12, 847–855. Langdon, R., Coltheart, M., & Ward, P. B. (2006). Empathetic perspective-taking is impaired in schizophrenia: Evidence from a study of emotion attribution and theory of mind. Cognitive Neuropsychiatry, 11(2), 133–155. Leslie, K. R., Johnson-Frey, S. H., & Grafton, S. T. (2004). Functional imaging of face and hand imitation: Towards a motor theory of empathy. NeuroImage, 21, 601–607. Majoram, D., Job, D. E., Whalley, H. C., Gountouna, V., McIntosh, A. M., Simonotto, E., et al. (2006). A visual joke fMRI investigation into Theory of Mind and enhanced risk of schizophrenia. NeuroImage, 31, 1850–1858. Montague, D. F. F., & Walker-Andrews, A. S. (2001). Peekaboo: A new look to infants‘ perception of emotion expression. Developmental Psychology, 37, 826–838. Ortony, A., & Turner, T. J. (1990). What’s basic about basic emotions? Psychological Review, 97, 315–331. Paus, T. (2001). Primate anterior cingulate cortex: Where motor control, drive and cognition interface. Nature Reviews Neuroscience, 2, 417– 424. Pelphrey, K. A., Morris, J. P., & McCarty, G. (2004). Grasping the intentions of others: The perceived intentionality of an action influences activity in the superior temporal sulcus during social perception. Journal of Cognitive Neuroscience, 16, 1706–1716. Pfeifer, J. H., Iacoboni, M., Mazziotta, J. C., & Dapretto, M. (2008). Mirroring others’ emotions relates to empathy and interpersonal competence in children. NeuroImage, 39, 2076–2085. Phillips, L. H., MacLean, R. D., & Allen, R. (2002). Age and the understanding of emotions: Neuropsychological and sociocognitive perspectives. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 57, P526–530. Phillips, A. T., Wellman, H. M., & Spelke, E. S. (2002). Infants’ ability to connect gaze and emotional expression to intentional action. Cognition, 85, 53–78. Pitcher, D., Garrido, L., Walsh, V., & Duchaine, B. C. (2008). Transcranial magnetic stimulation disrupts the perception and embodiment of facial expressions. Journal of Neuroscience, 28, 8929–8933. Pourtois, G., Sander, D., Andres, M., Grandjean, D., Reveret, L., Olivier, E., et al. (2004). Dissociable roles of the human somatosensory and superior temporal cortices for processing social face signals. European Journal of Neuroscience, 20, 3507–3515. Premack, D., & Woodruff, G. (1978). Does the chimpanzee have a theory of mind? The Behavioral and Brain Sciences, 4, 515–526. Rizzolatti, G., Fogassi, L., & Gallese, V. (2002). Motor and cognitive functions of the ventral premotor cortex. Current Opinion in Neurobiology, 12, 149–154. Rizzolatti, G., & Craighero, L. (2004). The mirror-neuron system. Annual Reviews of Neuroscience, 27, 169–192. Russell, T. A., Rubia, K. R., Bullmore, E. T., Soni, W., Suckling, J., Brammer, M. J., et al. (2000). Exploring the social brain in schizophrenia. Left prefrontal underactivation during mental state attribution. American Journal of Psychiatry, 157, 2040–2042. Saxe, R., Carey, S., & Kanwisher, N. (2004). Understanding other minds: Linking developmental psychology and functional neuroimaging. Annual Reviews of Psychology, 55, 87–124. Schulte-Ru¨hter, M., Markowitsch, H. J., Fink, G., & Pifke, M. (2007). Mirror neuron and Theory of Mind mechanisms involved in face-to-face interactions: A functional magnetic resonance imaging approach to empathy. Journal of Cognitive Neuroscience, 19, 1354– 1372. Schultz, J., Imamizu, H., Kawato, M., & Frith, C. D. (2004). Activation of the human superior temporal gyrus during observation of goal attribution by intentional objects. Journal of Cognitive Neuroscience, 16, 1695–1705. Shaw, P., Lawrence, E. J., Radbourne, C., Bramham, J., Polkey, C. E., & David, A. S. (2004). The impact of early and late damage to the human amygdala on ‘theory of mind’ reasoning. Brain, 127, 1535– 1548. Spiers, H., & Maguire, E. A. (2006). Spontaneous mentalizing during an interactive real world task: An fMRI study. Neuropsycholgia, 44, 1674–1682. Stone, V. E., Baron-Cohen, S., Calder, A., Keane, J., & Young, A. (2003). Acquired theory of mind impairments in individuals with bilateral amygdala lesion. Neuropsychologia, 41, 209–220.

Emotion recognition and affective ToM Van der Gaag, C., Minderaa, R., & Keysers, C. (2007). Facial expressions: What the mirror neuron system can and cannot tell us. Social Neuroscience, 2, 179–222. Vogeley, K., Bussfeld, P., Newen, A., Herrmann, S., Happe´, F., Falkai, P., et al. (2001). Mind reading: Neural mechanisms of theory of mind and self-perspective. NeuroImage, 14, 170–181. Vo¨llm, B. A., Taylor, A., Richardson, P., Corcoran, R., Stirling, J., McKie, S., et al. (2006). Neuronal correlates of theory of mind and empathy: A functional magnet resonance imaging study in a nonverbal task. NeuroImage, 28, 90–98. Walter, B., Blecker, C., Kirsch, P., Sammer, G., Schienle, A., Stark, R., et al. (2003). MARINA: An easy to use tool for the creation of MAsks for Region of INterest Analyses [abstract]. Presented at the 9th International Conference on Functional Mapping of the Human Brain,

1039 June 19–22, 2003, New York, NY. Available on CD-Rom in NeuroImage, 19, No. 2. Wicker, B., Perret, D. I., Baron-Cohen, S., & Decety, J. (2003). Being the target of another’s emotion: A PET study. Neuropsychologia, 41, 139– 146. Winston, J. S., Strange, B. A., O‘Doherty, J., & Dolan, R. J. (2002). Automatic and intentional brain responses during evaluation of trustworthiness of faces. Nature Neuroscience, 5, 277–283. Winston, J. S., O’Doherty, J., & Dolan, R. J. (2003). Common and distinct neural responses during direct and incidental processing of multiple facial emotions. NeuroImage, 20, 84–97. (Received July 9, 2009; Accepted December 23, 2009)

Psychophysiology, 47 (2010), 1040–1046. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01012.x

Genetic influence demonstrated for MEG-recorded somatosensory evoked responses

DENNIS VAN ’T ENT,a,b INGE L.C. VAN SOELEN,a,c KEES J. STAM,b,d ECO J.C. DE GEUS,a,b and DORRET I. BOOMSMAa,b a

Department of Biological Psychology, Vrije Universiteit, Amsterdam, The Netherlands Neuroscience Campus Amsterdam, Amsterdam, The Netherlands Rudolf Magnus Institute of Neuroscience, Department of Psychiatry, University Medical Center Utrecht, Utrecht, The Netherlands d Department of Clinical Neurophysiology and MEG Center, VU University Medical Center, Amsterdam, The Netherlands b c

Abstract We tested for a genetic influence on magnetoencephalogram (MEG)-recorded somatosensory evoked fields (SEFs) in 20 monozygotic (MZ) and 14 dizygotic (DZ) twin pairs. Previous electroencephalogram (EEG) studies that demonstrated a genetic contribution to evoked responses generally focused on characteristics of representative brain potentials. Here we demonstrate significantly smaller amplitude differences within MZ compared to DZ twin pairs for the complete SEF time series (across left and right hand SEFs: 0.37 vs. 0.60 pT2 and 0.28 vs. 0.39 pT2 for primary [SI] and secondary [SII] sensory cortex activation) and higher MZ than DZ wave shape correlations (.71 vs. .44 and .52 vs. .35 for SI and SII activation). Our findings indicate a genetic influence on MEG-recorded evoked brain activity and also confirm our recent conclusion (van ’t Ent, van Soelen, Stam, De Geus, & Boomsma, 2009) that higher MZ resemblance for EEG amplitudes is not trivially reflecting greater MZ concordance in intervening biological tissues. Descriptors: Somatosensory evoked response, Magnetoencephalogram, MEG, Similarities, Monozygotic twins, Dizygotic twins

potentials generally compared similarities within MZ versus DZ twin pairs for peak amplitudes, peak latencies, or both of one or more characteristic components of the brain response waveform. Instead of focusing on selected subsections, a more complete picture is obtained if within twin-pair similarities are assessed across the entire brain response time series. In fact, already in an early stage of EEG twin research, Lewis, Dustman, and Beck (1972) tested for a genetic influence on amplitude and wave shape of complete brain response waveforms and found that amplitudes of visual, auditory, and somatosensory evoked response waveforms were significantly more similar within MZ twin pairs compared to DZ twin pairs or pairs of unrelated individuals. The picture was less clear for wave shape, which showed significantly higher MZ correlations for visual- and auditory-evoked responses, but not for somatosensory-evoked responses (although there was a marginal trend toward higher MZ resemblance). Of note, the study was performed with a single EEG electrode over the somatosensory brain region, and therefore, as the authors also noted, the negative finding might have been because of within-twin-pair discrepancies in recording location. In this study, we investigated genetic influences on waveform amplitude and morphology of the entire time series of somatosensory-evoked brain activity in a sample of MZ and DZ twins while controlling for a possible confounding effect of differences

Studies that compared similarities of electroencephalogram (EEG) characteristics between genetically identical, monozygotic (MZ) twins and nonidentical, dizygotic (DZ) twins have indicated a significant genetic contribution to individual differences in electrical brain activity. For example, with regard to power of EEG traces in the classical delta, theta, alpha, and beta frequency bands, high heritabilities (i.e., higher similarities for EEG power within MZ as compared to DZ twin pairs) have been found ranging from 55% up to 90% (Smit, Posthuma, Boomsma, & De Geus, 2005; Smit, Wright, Hansell, Geffen, & Martin, 2006; Van Baal, De Geus, & Boomsma, 1996; van Beijsterveldt, Molenaar, De Geus, & Boomsma, 1996; Zietsch et al., 2007). In addition to ongoing brain activity, there is also evidence from EEG studies for genetic control of primary sensory brain responses to visual, auditory, and somatosensory stimulation as well as brain potentials related to higher order cognitive processing such as the P300 component (van Beijsterveldt, Molenaar, De Geus, & Boomsma, 1998; van Beijsterveldt & Van Baal, 2002; Wright et al., 2001). Studies on heritabilities of evoked/event-related We thank Johan Winnubst, Christianne Vink, and Roderick Bakker for help with contacting the twins and MEG data collection. Scanning costs were, in part, funded by the Amsterdam Brain Imaging Platform, Amsterdam, the Netherlands. Address reprint requests to: D. van ’t Ent, Vrije Universiteit, Department of Biological Psychology, van der Boechorststraat 1, 1081 BT, Amsterdam, The Netherlands. E-mail: [email protected] 1040

Genetic influence on MEG-SEFs in sensor location. To achieve this, we measured brain activity using a magnetoencephalogram (MEG) scanner equipped with a 151-sensor array. The MEG sensors are arranged to cover the whole head, so that the most optimal recording location can be identified separately for each individual. Further, for heritabilities of EEG amplitudes, it has been suggested that MZ twin resemblance may be strongly inflated because of greater MZ similarity of biological tissues between the brain and EEG electrodes, primarily the skull, which are also under genetic control (Kohn, 1991). In a recent study (van ’t Ent, van Soelen, Stam, De Geus, & Boomsma, 2009), we disproved this hypothesis for ongoing resting state brain activity by showing that large MZ twin correlations for power in the classic frequency bands remain if brain activity is recorded with MEG, which is virtually undisturbed by intervening tissues (Okada, Lahteenmaki, & Xu, 1999; Wolters et al., 2006). The use of MEG, rather than EEG, in the present study also allowed us to investigate if this finding applies equally to evoked brain response amplitudes.

Method Participants The sample consisted of 20 healthy right-handed MZ twin pairs (10 male: 18.8 ! 0.6 years; 10 female: 18.8 ! 0.4 years) and 14 healthy right-handed DZ twin pairs (6 male: 20.2 ! 0.4 years; 8 female: 19.9 ! 0.4 years) recruited from the Netherlands Twin Register (Boomsma et al., 2006). Zygosity was based on buccal cell DNA typing. All twins provided informed consent, and the study was approved by the medical ethics committee of the VU University Medical Center. MEG Recordings The twin and co-twin from a twin pair always visited the laboratory on the same day (morning or afternoon). Magnetic brain activity was recorded using a 151-sensor whole-head MEG scanner with axial gradiometers (VSM Medtech Ltd., Canada). Sampling rate was 625 Hz, with low-pass filtering at 265 Hz. Somatosensory-evoked responses were obtained separately for the left hand (left-hand SEF) and right hand (right-hand SEF), in two 2.5-min sessions, by electrically stimulating the median nerve at the wrist. Stimulation was provided by a constant current unit connected to a Grass S48 square-wave stimulator. The electric pulses were of 0.2-ms duration and delivered at 2 Hz. Particularly important factors that determine the sensitivity of the median nerve to the externally applied stimuli are the composition of the nerve and the location of the nerve relative to the ventral surface of the wrist, and it is conceivable that there is a higher anatomical resemblance of the upper limbs in MZ than in DZ twins. Therefore, to ensure that the median nerve was activated with similar strength in both MZ and DZ individuals, we adjusted the stimulus intensity for each individual to just below the threshold of thumb twitch. A post hoc analysis, in fact, did indicate that the applied strength of electrical stimulation tended to be more similar in MZ compared to DZ twins: mean within-twin-pair differences for left hand stimulation: 1.42 ! 0.99 mA (MZ) vs. 2.43 ! 2.28 mA (DZ); mean within-twin-pair differences for right hand stimulation: 1.08 ! 0.86 mA (MZ) vs. 2.34 ! 2.46 mA (DZ); main effect of twin pair type (MZ vs. DZ): F(1,28) 5 4.14, p 5 .051, across left- and right-hand stimulation; data on stimulation strength were incomplete for 1 MZ twin pair and 3 DZ twin pairs.

1041 Before and after each measurement, head position was determined and, to correct for the influence of head position on recorded amplitudes, MEG data for each individual were extrapolated onto new data sets with the same sensor locations corresponding to an average head position across all twins (de Munck, Verbunt, van ’t Ent, & van Dijk, 2001).

Data Processing MEG signals were processed using Fieldtrip software (F.C. Donders Centre for Cognitive Neuroimaging; http://www.ru.nl/ fcdonders/fieldtrip). Raw MEG data were visually inspected for artifacts including eye movements and excessive muscle activity. Subsequently, artifact-free epochs subtending from 100 ms before to 500 ms after stimulation onset were selected and averaged using the first 100 ms as a prestimulus baseline. The resulting SEF waveforms consisted of a number of successive components originating from activity in the primary and secondary somatosensory cortexes. For each individual, characteristic SEF waveform templates were obtained for two consecutive time windows (Figure 1). Time Window 1 covered the initial 90-ms phase of the SEF and Time Window 2 a subsequent, and final, 100-ms phase. To account for individual differences in arm length, the onset of Time Window 1 was set at the peak of the first SEF component. The templates (green colored traces in Figure 1) were obtained by averaging in each time window the SEF response at the sensor with maximum magnetic outflux and the amplitude inverse of the SEF at the sensor with maximum influx. The characteristic SEF in Time Window 1 (see small MEG field map insert in Figure 1) corresponds to a single region of SI activation contralateral to the side of stimulation. For this window, the sensor with maximum magnetic outflux (blue trace in Figure 1) was first selected from all MEG sensors covering the right temporal brain (sensors labeled MRT: N 5 21) and right frontal brain (sensors MRF: N 5 16) for left-hand SEFs and all sensors over the left temporal brain (MLT: N 5 21) and left frontal brain (MLF: N 5 16) for right-hand SEFs. The time point of maximum magnetic outflux was then determined from this sensor, and, subsequently, the sensor with maximum magnetic influx (red trace) at this time was selected from all remaining sensors. In Time Window 2, the characteristic field maps for left- and right-hand SEFs are similar and correspond to bilateral SII activation. For this window, for both left- and righthand SEFs, the sensor with maximum magnetic outflux (blue trace in Figure 1) was selected from all sensors over the left temporal brain (MLT) and the sensor with maximum magnetic influx (red trace) from all sensors over the right temporal brain (MRT). Left temporal outflux and right temporal influx are associated with left and right SII activation, respectively; return flux for both sources partially overlap and cancel out over central brain regions. Somatosensory-evoked responses were compared for waveform amplitude as well as wave shape similarity. Within-twinpair amplitude similarity was quantified by computing the squared Euclidean distance between corresponding SEF template time series (vectors T1 and T2) for the twins of every pair: (T1 " T2) # (T1 " T2) 0 , where # denotes the inner product and the transpose. Within-twin-pair wave shape similarity was assessed by means of Pearson’s linear correlation coefficient between corresponding SEF templates:

1042

D. van ’t Ent et al. Construction of SEF template Time Window 1: pT

Sensor with max. outflux at MRT and MRF Sensor with max. influx at time of maximum outflux

0.1

0 Max. influx

Max. outflux

–0.1 –0.1

0

0.1

0.2

0.3

0.4

s

Time Window 2: pT

Sensor with max. outflux at MLT Sensor with max. influx at MRT

0.1

0 Max. outflux

Max. influx

–0.1 –0.1

0

0.1 SEF Template:

0.2 =[

0.3 ] + [–1 x

0.4

s

]

2 Figure 1. Construction of a SEF template (in green) for Time Window 1 (top panel) and Time Window 2 (bottom panel) illustrated for a left-hand SEF (grand average across all twins). In each panel, an overlay of the response at all MEG sensors is shown, with stimulation onset at 0 s.

with 1 T!1 ¼ N

ðT1 " T1 Þ $ ðT2 " T2 Þ0 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðT1 " T1 Þ $ ðT1 " T1 Þ0 ðT2 " T2 Þ $ ðT2 " T2 Þ0 X j

T1j and T!2 ¼

1X T2j ; and N5number of samples: N j

To additionally compare MZ and DZ groups with regard to characteristics of the sources in the brain underlying the SEFs, we also performed a dipole source analysis. Dipole modeling was performed on the peak of the P35m SEF component that occurs at about 35 ms after onset of median nerve stimulation and corresponds to contralateral SI activation. We focused on this component because it was generally the most prominent and could be identified in every subject. In addition, we generally obtained the most reliable single dipole fit solution for this component. Dipole analysis was performed using the DipoleFit software tool provided by the MEG system manufacturer. A single sphere was used as a head model (sphere center at x 5 0 cm, y 5 0 cm, z 5 5 cm in the nasion-ear coordinate frame [de Munck et al., 2001]; sphere radius 5 7.5 cm). Similarities of source characteristics were quantified by computed differences in dipole location (Euclidean distance), orientation (vector difference angle), and strength.

Statistical Analysis SEF amplitude differences and wave shape correlations and differences in P35m source characteristics were computed for all MZ twin pairs and DZ twin pairs. In addition, but for qualitative comparison only, we computed amplitude differences and wave shape correlations for all possible couplings of twins from different pairs that could be constructed from our total sample of 68 MZ and DZ twins (N 5 2240 unique pairs of unrelated twins, without consideration of MZ or DZ status). Amplitude differences and wave shape correlations for these pairs were, however, not included in the statistical analyses, because they are not completely independent from the values computed for MZ and DZ pairs (as the pairs are constructed from twins of our MZ and DZ samples). Prior to statistical analysis, wave shape correlations were converted to Fisher Z scores to ensure a normal sampling distribution. Pearson correlation coefficients, rather than Z scores, are reported, however, to facilitate interpretation. SEF amplitude and wave shape similarities were evaluated using a general linear model with Stimulated Hand (within-twin-pair similarity for left-hand SEFs versus within-twin-pair similarity for right-hand SEFs) and Time Window (within-twin-pair SEF similarity for Time Window 1 vs, within-twin-pair SEF similarity for Time Window 2) as within-twin-pair factors and Twin Pair Type (MZ vs. DZ) as the between-twin-pair factor. To check for an influence of differences in absolute somatosensory response

Genetic influence on MEG-SEFs

1043

Table 1. Means (and Standard Deviations) of SEF Amplitude Differences and Wave Shape Correlations within Twin Pairs

Measure Amplitude differences Wave shape correlations

Stimulated hand Left Right Left Right

Time window 1

Time window 2

MZ

DZ

Unrelated

MZ vs. DZ

MZ

DZ

Unrelated

MZ vs. DZ

.39 (.17) .35 (.15) .64 (.41) .78 (.18)

.55 (.23) .64 (.24) .49 (.33) .38 (.34)

.66 (.26) .72 (.28) .30 (.39) .27 (.39)

.025 .001 .102 .000

0.29 (0.10) 0.26 (0.13) 0.43 (0.50) 0.61 (0.37)

0.34 (0.20) 0.43 (0.15) 0.32 (0.47) 0.37 (0.56)

0.39 (0.18) 0.45 (0.20) 0.11 (0.54) 0.18 (0.54)

.349 .001 .572 .216

Note: Means and standard deviations of amplitude differences (top rows, and in pT2) and wave shape correlations (bottom rows) for left- and hand righthand SEFs (column Stimulated hand) in Time Window 1 and Time Window 2; within MZ twin pairs (column MZ), DZ twin pairs (DZ) and within pairs of randomly coupled, unrelated, twins (Unrelated). See also Figure 2 for a graphical display of the data. Columns MZ vs. DZ show results of statistical comparisons (p values) between SEF similarities within MZ and DZ twin pairs.

SEF Amplitude Differences and Wave Shape Correlations Means and standard deviations of computed amplitude differences and wave shape correlations are shown in Table 1 and Figure 2. For SEF amplitude differences within MZ and DZ pairs, there was no significant main effect for variable Stimulated Hand, F(1,32) 5 1.15, p 5 .292, or a Stimulated Hand ! Time Window interaction, F(1,32) 5 0.06, p 5 .804. However, we did find an interaction between Stimulated Hand and Twin Pair Type, F(1,32) 5 6.29, p 5 .017, which was explained by the fact that evoked response amplitude differences tended to be smaller for right- compared to left-hand SEFs within MZ twin pairs, but smaller for left- compared to right-hand SEFs within DZ twin pairs. There was also a significant main effect for variable Time Window, F(1,32) 5 48.61, po.001, which indicated that withinpair differences in SEF amplitude were relatively reduced in Time Window 2 compared to Time Window 1, in particular for DZ twin pairs: Time Window ! Twin Pair Type interaction, F(1,32) 5 7.59, p 5 .010. This finding disappeared, however, when statistical evaluation was repeated after separately normalizing the somatosensory response amplitudes in both time windows, indicating that it reflected a systematic difference of SEF amplitudes within the first and second time windows. For wave shape correlations within MZ and DZ twin pairs, there were no differences for SEFs after left- versus right-hand stimulation or for SEFs in Time Window 1 versus Time Window 2: Stimulated Hand, F(1,32) 5 0.94, p 5 .340; Time Window, F(1,32) 5 2.45, p 5 .127; Stimulated Hand ! Time window, F(1,32) 5 1.98, p 5 .169; Stimulated Hand ! Twin Pair Type,

1.0

Amplitude difference (pT2)

Results

F(1,32) 5 1.39, p 5 .247; Time Window ! Twin Pair Type, F(1,32) 5 1.56, p 5 .220. Significant main effects of variable Twin Pair Type for both the analysis on amplitude differences, F(1,32) 5 17.31, po.001, and wave shape correlations, F(1,32) 5 11.96, p 5 .002, indicated

Time Window 1 Left hand SEFs

Time Window 2

0.5

1.0

Right hand SEFs

0.5

MZ 1.0

DZ

Unr.

MZ

DZ

Unr.

Unr.

MZ

DZ

Unr.

Left hand SEFs

0.5 Wave shape correlation

amplitudes on computed Euclidean amplitude differences, we also repeated the statistical evaluation on amplitude similarities after first separately normalizing the somatosensory response amplitudes in both time windows (to a maximum of 1 for the twin with the largest SEF amplitude). Finally, a bivariate Pearson correlation-based analysis was performed, again limited to SEF similarity values within MZ and DZ twin pairs, to test for a possible relation between SEF amplitude similarities and SEF wave shape similarities across left- and right-hand SEFs and both time windows (N 5 34 twin pairs ! 2 stimulation sides ! 2 time windows 5 136). Similarities of P35m dipole characteristics were evaluated using a general linear model with variable Stimulated Hand as within-twin-pair factors and variable Twin Pair Type as thebetween-twin-pair factor. There were no indications for sex differences; therefore all statistics were computed for male and female twins, combined.

1.0

Right hand SEFs

0.5

MZ

DZ

Figure 2. Mean amplitude differences (top) and wave shape correlations (bottom) for left- and right-hand SEFs in Time Window 1 (left bars) and Time Window 2 (right bars) within MZ twin pairs (MZ), DZ twin pairs (DZ), and pairs of unrelated twins (Unr). Error bars indicate standard deviations.

1044

D. van ’t Ent et al. Right hand SEF templates in time window 1 (sorted on within twin-pair wave shape correlation) MZ twin pairs

0.97

0.96

0.94

0.94

0.94

0.92

0.91

0.90

0.89

0.88

0.81

0.80

0.76

0.73

0.60

0.57

0.52

0.50

0.50

0.48

0.60

0.54

0.46

0.38

0.28

0.10

DZ twin pairs

0.84

0.79

0.73

0.70

0.06

0.01

– 0.05

–0.17

corr.

Figure 3. Templates for the initial part of the SEF, after right-hand stimulation, in Time Window 1. SEF data for the twin and co-twin (black vs. red colored traces) of MZ twin pairs (boxes displayed at the top) and DZ twin pairs (bottom boxes) are displayed, from left to right, according to a sort in descending order of within-twin-pair SEF wave shape correlations (indicated by the number in each box). Sorting was performed separately for MZ and DZ pairs.

that somatosensory-evoked brain responses showed higher resemblance (smaller amplitude differences and higher wave shape correlations) within MZ twin pairs as compared to DZ twin pairs. Qualitative comparison of amplitude difference and wave shape correlation values in Table 1 and Figure 2 suggests only a marginal tendency for higher SEF resemblance within DZ twin pairs compared to pairs of unrelated twins. Post hoc evaluation of every individual SEF comparison within MZ and DZ twin pairs (see Table 1, columns MZ vs. DZ) indicated that amplitude differences were significantly smaller within MZ twin pairs compared to DZ twin pairs for left- and right-hand SEFs in Time Window 1 and right-hand SEFs in Time Window 2 and that wave shape correlations were significantly higher within MZ compared to DZ twin pairs for right-hand SEFs in Time Window 1 (see also Figure 3). Finally, we found a significant negative correlation between SEF amplitude differences and SEF wave shape correlations within twin pairs (r 5 ! .28, p 5 .001), which demonstrates that SEF amplitudes tended to be more similar between twins (smaller amplitude differences) if SEF wave shapes were more similar (higher wave shape correlations).

Differences in Dipole Source Characteristics Table 2 shows means and standard deviations of within-twin-pair differences in position, orientation, and strength of equivalent current dipole (ECD) sources fitted to the P35m SEF component. Statistical analysis indicated a significant main effect for variable Stimulated Hand only for dipole position: position, F(1,32) 5 5.07, p 5 .031; orientation, F(1,32) 5 0.09, p 5 .767; strength, F(1,32) 5 0.88, p 5 .356, which was explained by the fact that within twin-pair differences in dipole location were smaller for left- compared to right-hand SEFs. There were no significant interactions between variables Stimulated Hand and Twin Pair Type for any of the dipole characteristics: position, F(1,32) 5 1.01, p 5 .323; orientation, F(1,32) 5 0.79, p 5 .381; strength, F(1,32) 5 0.80, p 5 .378. Significant main effects of variable Twin Pair Type, indicating higher similarity within MZ relative to DZ twin pairs, were found for dipole position, F(1,32) 5 9.44, p 5 .004, and strength, F(1,32) 5 4.68, p 5 .038, but not for dipole orientation, F(1,32) 5 0.31, p 5 .583. Post hoc evaluation of every individual comparison within MZ and DZ twin pairs (Table 2, column MZ vs. DZ) indicated that smaller differences in MZ twin pairs were evident in particular for po-

Table 2. Means (and Standard Deviations) of Within-Twin-Pair Differences in Characteristics of the Dipole Fitted to the P35m SEF Component Measure Position difference Orientation difference Strength difference

Stimulated hand

MZ

DZ

MZ vs. DZ

Left Right Left Right Left Right

1.24 (0.67) 1.44 (0.42) 19.31 (22.68) 23.20 (29.02) 8.38 (6.28) 8.47 (7.76)

1.65 (0.73) 2.18 (1.00) 28.88 (38.89) 21.04 (9.34) 11.52 (11.62) 15.38 (10.88)

.104 .006 .372 .791 .316 .038

Note: Means and standard deviations of differences in position (in centimeters), orientation (1), and strength (nanoAmpere meters) of the dipoles fitted to the P35m component of the SEF after left- and hand-right hand stimulation (column Stimulated hand), within MZ twin pairs (column MZ) and DZ twin pairs (DZ). Column MZ vs. DZ shows results of statistical comparisons (p values) between dipole parameter similarities within MZ and DZ twin pairs.

Genetic influence on MEG-SEFs

1045

0.4

MZ Twin pair A

MZ Twin pair B

twin

twin

co-twin

co-twin

0.2 0 –0.2 –0.4 0.4

pT

0.2 0 –0.2 –0.4

0

0.1

s 0.2

0

0.1

0.2

Figure 4. Somatosensory-evoked fields of two selected MZ twin pairs (overlay of all MEG sensors); the SEFs exhibit high within-twin-pair (columns) but lower between-twin-pair (rows) resemblance.

sition and strength of dipoles fitted to the P35m of SEFs after right-hand stimulation.

Discussion This is, to our knowledge, the first study to test for a genetic influence on sensory-evoked brain activity recorded with MEG. In line with previous EEG findings (Lewis et al., 1972), the results demonstrate that somatosensory-evoked responses show a high degree of correspondence between genetically identical MZ twins (e.g., Figure 4). We show this for the complete evoked response, as opposed to focusing on selected characteristics of representative brain potentials, which has been common practice in most previous EEG studies (for an overview, see van Beijsterveldt & Van Baal, 2002). Our findings thus indicate a substantial genetic influence on the complete time course of evoked brain activity. In combination with evidence for more similar dipole source characteristics of the SEF in MZ compared to DZ twins, our results are in line with previously reported evidence for strong genetic control of brain anatomy, including the sensory cortex (Lenroot et al., 2009; Peper, Brouwer, Boomsma, Kahn, & Hulshoff Pol, 2007; Peper et al., 2009). Although we tested a large number of individuals for a MEG study (N 5 68 in total), the sample size is still relatively limited for a MZ versus DZ twin-pairs comparison of resemblance. Nevertheless, the present finding of a genetic influence on amplitudes of evoked brain activity measured with MEG is in agreement with our recent findings for power of ongoing brain activity during rest (van ’t Ent et al., 2009) and with previous EEG studies on larger twin samples (van Beijsterveldt & Van Baal, 2002). The present results therefore substantiate that our previous conclusion that higher MZ resemblance for amplitudes of ongoing brain activity in EEG traces is not just reflecting greater

MZ concordance in intervening biological tissues (van ’t Ent et al., 2009) can also be extended to amplitudes of evoked brain responses. We also found that SEF amplitudes tended to be more similar between twins if SEF wave shapes were more similar. SEF wave shapes obviously are more alike if the constituent SEF components have more similar amplitudes. However, wave shape and amplitude are not necessarily strictly coupled. SEFs with highly similar wave shapes might, for example, differ in amplitude across the entire response or an extended section of the response. This can occur because of systematic differences in the physiology of the brain, such as individual differences in depth location, orientation, or strength of SEF sources. In particular for later SEF components, top-down processes may also play a role, such as differences in the amount of attention paid to the electrical stimulation (Eimer & Forster, 2003). The finding of a significant correlation between SEF morphology and amplitude therefore provides an indication that such systematic differences did not play a significant role in this study. The present data inform us that genetic factors influence SEF variation but do not allow us to draw definite conclusions on the exact mechanisms that underlie this influence. There might be a direct genetic effect, but it is also conceivable that genetic effects are indirect (e.g., through personality) through an influence on the environments that people expose themselves to. In either case, MEG appears to be useful as an endophenotype for individual differences in brain structure and function. Endophenotypes represent biological markers intermediate in the pathway between genetic variation and final individual differences of behavior and are a key construct of imaging genetics (de Geus, Goldberg, Boomsma, & Posthuma, 2008; Green et al., 2008). Measurements closer to the level of neural circuits underlying specific behaviors or behavior disorders are likely more tightly

1046

D. van ’t Ent et al.

associated with the effect of a single gene or a limited set of genes, which increases the power of genetic association testing. A number of genomic loci have already been linked to individual differences in cortical oscillations and event-related potentials

measured with EEG (Begleiter & Porjesz, 2006; Bodenmann et al., 2009; Espeseth, Rootwelt, & Reinvang, 2009; Liu et al., 2009). The present evidence suggest that MEG endophenotypes may prove similarly useful in genetic research.

REFERENCES Begleiter, H., & Porjesz, B. (2006). Genetics of human brain oscillations. International Journal of Psychophysiology, 60, 162–171. Bodenmann, S., Rusterholz, T., Durr, R., Stoll, C., Bachmann, V., Geissler, E., et al. (2009). The functional Val158Met polymorphism of COMT predicts interindividual differences in brain alpha oscillations in young men. Journal of Neuroscience, 29, 10855–10862. Boomsma, D. I., De Geus, E. J., Vink, J. M., Stubbe, J. H., Distel, M. A., Hottenga, J. J., et al. (2006). Netherlands Twin Register: From twins to twin families. Twin Research and Human Genetics, 9, 849–857. De Geus, E., Goldberg, T., Boomsma, D. I., & Posthuma, D. (2008). Imaging the genetics of brain structure and function. Biological Psychology, 79, 1–8. De Munck, J. C., Verbunt, J. P., van ’t Ent, D., & van Dijk, B. W. (2001). The use of an MEG device as 3D digitizer and motion monitoring system. Physics in Medicine and Biology, 46, 2041–2052. Eimer, M., & Forster, B. (2003). Modulations of early somatosensory ERP components by transient and sustained spatial attention. Experimental Brain Research, 151, 24–31. Espeseth, T., Rootwelt, H., & Reinvang, I. (2009). Apolipoprotein E modulates auditory event-related potentials in healthy aging. Neuroscience Letters, 459, 91–95. Green, A. E., Munafo, M. R., DeYoung, C. G., Fossella, J. A., Fan, J., & Gray, J. R. (2008). Using genetic data in cognitive neuroscience: From growing pains to genuine insights. Nature Reviews Neuroscience, 9, 710–720. Kohn, L. A. P. (1991). The role of genetics in craniofacial morphology and growth. Annual Review of Anthropology, 20, 261–278. Lenroot, R. K., Schmitt, J. E., Ordaz, S. J., Wallace, G. L., Neale, M. C., Lerch, J. P., et al. (2009). Differences in genetic and environmental influences on the human cerebral cortex associated with development during childhood and adolescence. Human Brain Mapping, 30, 163– 174. Lewis, E. G., Dustman, R. E., & Beck, E. C. (1972). Evoked response similarity in monozygotic, dizygotic and unrelated individuals: A comparative study. Electroencephalography and Clinical Neurophysiology, 32, 309–316. Liu, J., Kiehl, K. A., Pearlson, G., Perrone-Bizzozero, N. I., Eichele, T., & Calhoun, V. D. (2009). Genetic determinants of target and noveltyrelated event-related potentials in the auditory oddball response. NeuroImage, 46, 809–816. Okada, Y. C., Lahteenmaki, A., & Xu, C. (1999). Experimental analysis of distortion of magnetoencephalography signals by the skull. Clinical Neurophysiology, 110, 230–238. Peper, J. S., Brouwer, R. M., Boomsma, D. I., Kahn, R. S., & Hulshoff Pol, H. E. (2007). Genetic influences on human brain structure: A

review of brain imaging studies in twins. Human Brain Mapping, 28, 464–473. Peper, J. S., Schnack, H. G., Brouwer, R. M., Van Baal, G. C., Pjetri, E., Szekely, E., et al. (2009). Heritability of regional and global brain structure at the onset of puberty: A magnetic resonance imaging study in 9-year-old twin pairs. Human Brain Mapping, 30, 2184–2196. Smit, C. M., Wright, M. J., Hansell, N. K., Geffen, G. M., & Martin, N. G. (2006). Genetic variation of individual alpha frequency (IAF) and alpha power in a large adolescent twin sample. International Journal of Psychophysiology, 61, 235–243. Smit, D. J., Posthuma, D., Boomsma, D. I., & De Geus, E. J. (2005). Heritability of background EEG across the power spectrum. Psychophysiology, 42, 691–697. Van Baal, G. C., De Geus, E. J., & Boomsma, D. I. (1996). Genetic architecture of EEG power spectra in early life. Electroencephalography and Clinical Neurophysiology, 98, 502–514. Van Beijsterveldt, C. E., Molenaar, P. C., De Geus, E. J., & Boomsma, D. I. (1996). Heritability of human brain functioning as assessed by electroencephalography. American Journal of Human Genetics, 58, 562–573. Van Beijsterveldt, C. E., Molenaar, P. C., De Geus, E. J., & Boomsma, D. I. (1998). Individual differences in P300 amplitude: A genetic study in adolescent twins. Biological Psychology, 47, 97–120. Van Beijsterveldt, C. E., & Van Baal, G. C. (2002). Twin and family studies of the human electroencephalogram: A review and a metaanalysis. Biological Psychology, 61, 111–138. Van ’t Ent, D., van Soelen, I. L., Stam, C. J., De Geus, E. J., & Boomsma, D. I. (2009). Strong resemblance in the amplitude of oscillatory brain activity in monozygotic twins is not caused by ‘‘trivial’’ similarities in the composition of the skull. Human Brain Mapping, 30, 2142–2145. Wolters, C. H., Anwander, A., Tricoche, X., Weinstein, D., Koch, M. A., & MacLeod, R. S. (2006). Influence of tissue conductivity anisotropy on EEG/MEG field and return current computation in a realistic head model: A simulation and visualization study using high-resolution finite element modeling. NeuroImage, 30, 813–826. Wright, M. J., Hansell, N. K., Geffen, G. M., Geffen, L. B., Smith, G. A., & Martin, N. G. (2001). Genetic influence on the variance in P3 amplitude and latency. Behavior Genetics, 31, 555–565. Zietsch, B. P., Hansen, J. L., Hansell, N. K., Geffen, G. M., Martin, N. G., & Wright, M. J. (2007). Common and specific genetic influences on EEG power bands delta, theta, alpha, and beta. Biological Psychology, 75, 154–164. (Received July 13, 2009; Accepted November 2, 2009)

Psychophysiology, 47 (2010), 1047–1056. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01016.x

Schizotypal traits and N400 in healthy subjects

MARIE PRE´VOST,a,b MITCHELL RODIER,c LOUIS RENOULT,a,b YVONNE KWANN,a EMMANUELLE DIONNE-DOSTIE,a ISABELLE CHAPLEAU,a MATHIEU BRODEUR,a,c CLAIRE LIONNET,a and J. BRUNO DEBRUILLEa,b,c a

Douglas Mental Health University Institute, Montre´al, Que´bec, Canada Department of Neurology and Neurosurgery, McGill University, Montre´al, Que´bec, Canada c Department of Psychiatry, McGill University, Montre´al, Que´bec, Canada b

Abstract We examined whether correlations previously found between symptoms of schizophrenia patients and the amplitude of an event-related potential (ERP), the N400, could be also found between schizotypal experiences of healthy subjects and the N400. We chose a semantic categorization task previously used with patients. Schizotypal experiences were measured with the schizotypal personality questionnaire (SPQ). The effects of the other factors were controlled for when assessing the correlations between each SPQ factor and N400s. These correlations were assessed at each electrode site to see whether their distribution on the scalp follows that of the N400 effect. Disorganization and interpersonal scores were found to correlate with ERPs in the N400 time window, as previously reported for the comparable symptoms of patients. However, the scalp distribution of these correlations differed from that of the N400 effect. Descriptors: Delusional ideation, Disorganization, Interpersonal factor, Event-related brain potentials, N400

well, Kamath, & Compton, 2009; Johnston, Rossell, & Gleeson, 2008; Wilson, Christensen, King, Li, & Zelazo, 2008). Semantic processes are among the cognitive processes that have been widely studied in schizophrenia. The N400, an eventrelated potential (ERP) evoked by any potentially meaningful stimulus and thought to reflect semantic processing, has been related to reality distortion, disorganization, and negative symptoms of schizophrenia patients in various studies (e.g., Andrews, Shelley, Ward, Fox, Catts, & McConaghy, 1993; Debruille, Kumar, Saheb, Chintoh, Gharghi, et al., 2007; Kostova, Passerieux, Laurent, & Hardy-Bayle´, 2005, respectively). In the next paragraphs, these three relations will be reviewed. The first relation, that between reality distortion and the N400, has been studied because of the ostensible tendency of patients to maintain their delusions in the face of disconfirmatory evidence. It has been hypothesized that this ‘tendency’ is related to a deficit in integrating the information that challenges their beliefs (Freeman, 2007; Moritz & Woodward, 2006). The N400 was thus used as an index of this integration since the amplitude of this ERP appears to be proportional to the amount of effort that is spontaneously deployed to integrate unexpected semantic information into cognitive representations (e.g., Holcomb, 1993). Stimuli that arguably trigger more integration effort, such as stimuli that do not match the context, elicit N400s of greater amplitudes than stimuli that are primed because they match preceding stimuli. Consistent with the N400 integration hypothesis and with their deficit in integrating disconfirmatory evidence, schizophrenia patients who were more delusional were found to have smaller N400s than patients who were less delusional (Debruille et al., 2007). The link between N400 and delusion severity in schizophrenia patients was confirmed by the

Research into the symptoms of schizophrenia focuses primarily on the three dimensions that are frequently used to describe schizophrenia, which are reality distortion, disorganization, and negative symptoms (Lenzenweger, 1999). Interestingly, experiences which resemble these symptoms can be observed in the general population (Raine, 1991; Reynolds, Raine, Mellingen, Venables, & Mednick, 2000; Verdoux & van Os, 2002). This finding has raised the idea that, rather than a dichotomy, there could be a continuum between schizophrenia and normality (Claridge, 1997; Johns & van Os, 2001; Shevlin, Murphy, Dorahy, & Adamson, 2007; Strauss, 1969). Although the dichotomous view of psychopathology remains central to mental health care and research, there is a growing debate between these two approaches. One way to go forward is to test whether symptoms of schizophrenia patients and schizotypal personality traits of normals are associated with similar cognitive anomalies, since, as shown by Johns and van Os (2001), such similarities would reinforce the notion of a continuum. An increasing number of studies have looked at the relationship between schizotypal traits in healthy participants and cognitive functions. They observed in these participants similar cognitive anomalies as those found in schizophrenia patients (e.g., BedWe thank the subjects, whose participation was essential in the completion of this study. Funding was provided by Fonds de la Recherche en Sante´ Que´bec (scholarship 10084 to J. B. D.; fellowship 13542 to L. R.) and National Alliance for Research on Schizophrenia and Depression (NARSAD Young Investigator Award to J. B. D.). Address reprint requests to: Dr. J. Bruno Debruille, Human Neurocognitive Science Laboratory, F. B. C. Pavilion, Douglas Mental Health University Institute, 6875 Blvd. LaSalle, Montre´al, (QC) H4H 1R3, Canada. E-mail: [email protected] 1047

1048 results of Kiang, Kutas, Light, and Braff (2007). These authors reported that the severity of psychotic symptoms (i.e., delusions and hallucinations) correlates with a reduction of the N400 effect, that is, with a reduction of the difference between the amplitude of the N400 to mismatching stimuli and that to matching stimuli. Since the variations of one or of both of these amplitudes could be responsible for smaller N400 effects, one of the aims of the present study was to carefully differentiate the three N400 measures. In the rest of this paper, ‘N400 amplitude’ will be used only to designate the raw amplitude of the N400 deflections, whereas the phrase ‘N400 effect’ will be exclusively devoted to N400 differences. The second relation, that between disorganization and N400, has been studied partly because of the deficit found in schizophrenia patients in processing context efficiently and in keeping it in working memory during long stimulus onset asynchronies (SOAs) (Cohen & Servan-Schreiber, 1992; Hardy-Bayle´, Sarfati, & Passerieux, 2003; Hemsley, 2005). Accordingly, a word that is related to a following target word induces less priming in patients than in normal controls. One should thus observe a correlation in the direction opposite to that predicted for delusions. N400s elicited by primed target words should be larger in more- than in less-disorganized patients (see Discussion section for another hypothesis leading to the same prediction). The results of previous works support that prediction. Correlations were found between thought disorder and larger N400 amplitudes for congruent targets in Kuperberg, Sitnikova, Goff, and Holcomb (2006) and in Kostova et al. (2005, personal communication), which possibly explain the smaller N400 effects observed in Ditman and Kuperberg (2007). Interestingly, one study described greater N400s for the incongruent targets for higher thought disorder scores (Salisbury, O’Donnell, McCarley, Nestor, & Shenton, 2000), while another mentioned greater amplitudes for the mean of congruent targets and incongruent targets (Andrews et al., 1993). It has to be noted that in all the above mentioned studies, the word or the sentence used to prime the target stimulus was shown just once in the experiment. This can only increase the impact of the problematic maintenance of this context information in working memory and further jeopardize the facilitation this context induces and account for the larger N400 amplitudes obtained for the targets. However, correlations between N400s and the two symptoms mentioned above were not always observed (Ditman & Kuperberg, 2007; Kostova et al., 2005; Salisbury et al., 2000). For delusion, one study reported an opposite relationship, that is, greater N400 amplitudes with higher delusion scores (Kiang, Kutas, Light, & Braff, 2008). For disorganization, the correlations with N400s sometimes failed to be found (Kiang et al., 2007, 2008), and one study described a relationship going in the reverse direction, that is, reduced N400 amplitudes for stimuli that do not match the context (non-exemplar category) for highly disorganized patients (Debruille et al., 2007). Nevertheless, these discrepancies may not be so surprising. Because delusions and disorganization can be positively correlated to each other (Lenzenweger & Dworkin, 1996; for comparable experiences in the general population, see Raine, Reynolds, Lencz, Scerbo, Triphon, & Kim, 1994; Reynolds et al., 2000) their opposite influences on N400 amplitudes could cancel each other, or one could even reverse the influence of the other. Therefore, in the present study, which aimed at evaluating the relationship between schizotypal traits in healthy participants and the N400, it was decided to use partial correlations in order to assess the effect of

M. Pre´vost et al. one schizotypal trait on N400s while controlling for the effects of the two other traits. Similarly, the third relation, that between the negative symptoms of schizophrenia and the N400, was not found in some studies (Ditman & Kuperberg, 2007; Kiang et al., 2008; Kuperberg et al., 2006; Salisbury et al., 2000). But two studies with schizophrenia patients found a tendency for a reduced N400 effect with more negative symptoms (Kiang et al., 2007; Kostova et al., 2005). Moreover, when looking at similar schizotypal traits in healthy subjects, one study reported significant results. Kiang and Kutas (2005) observed a correlation between diminished N400 effects and higher scores for the interpersonal items of the Schizoptypal Personality Questionnaire, the SPQ (Raine, 1991), which reflect experiences that resemble negative symptoms in patients. As this interpersonal factor also includes paranoid ideation and social anxiety, the authors proposed that a mistrust bias might be responsible for this finding. They based their proposition on a study where mistrust was induced in university students (Schul, Mayo, & Burnstein, 2004). In this study, behavioral responses showed both an increased semantic priming effect for words having a meaning opposite to that of the prime (e.g., hollow-full) and a decreased effect for words that were synonyms (e.g., hollow-empty). The authors of this behavioral study proposed that ‘‘mistrust may be associated with increased activation of message-incongruent associations and decreased activation of message-congruent associations, by encouraging a person to focus more on the possibility that a message is invalid.’’ Kiang and Kutas (2005) thus related their smaller N400 effects to this decreased activation of message-congruent associations. In addition to the interpersonal score, the SPQ can also be used to measure the disorganization trait and delusional-like ideation in healthy subjects. The idea that these experiences of healthy subjects resemble schizophrenia symptoms has been supported in several studies (Raine, 1991; Reynolds et al., 2000; Verdoux & van Os, 2002; for experiences which resemble delusions, see Peters, Joseph, & Garety, 1999; and van Os, Hanssen, Bijl, & Ravelli 2000; for traits which resemble disorganization, see Coleman, Levy, Lenzenweger, & Holzman, 1996; and Gooding, Tallent, & Hegyi, 2001; for traits that resemble negative symptoms, see Chapman, Edell, & Chapman, 1980; and Cadenhead, Kumar, & Braff, 1996). Nevertheless, to stress the fact that there may be qualitative and quantitative differences between the symptoms of patients and schizotypal traits, a different terminology is used for healthy subjects. The phrase ‘disorganization trait’ is used to refer to experiences that resemble disorganization and the phrase ‘delusional-like ideation’ is used to refer to experiences evaluated by the SPQ that are observed in healthy subjects and that resemble delusions in patients. Recent studies using schizotypal personality measures have demonstrated that ERPs may be used to study neurocognitive processes related to schizotypal traits in healthy individuals (Gassab, Mechri, Dogui, Gaha, d’Amato, et al., 2006; Kiang & Kutas, 2005; Wan, Crawford, & Boutros, 2006; Wang, Miyazato, Hokama, Hiramatsu, & Kondo, 2004). Although to a lesser extent than in the case of schizotypal personality disorders, the existence of measurable schizotypal traits in healthy populations may be seen as an opportunity for approaching the mechanisms of schizophrenia symptoms in a less confounded context. Indeed, whereas some patients’ anomalies might also be the consequences of the long-term disability induced by symptoms, this is less likely to be the case in healthy populations. Moreover, as with schizotypal personality disorders, studying

Schizotypal traits and N400 in healthy subjects healthy subjects allows interpretations free from the problems of patients’ medication. The aim of the present study was thus to explore the neurocognitive mechanisms underlying schizotypal personality traits in healthy individuals. More specifically, our goal was to see whether we could find in these subjects the relationships observed in patients between N400 and the three schizophrenia symptoms, namely, the smaller N400s with higher delusions scores, the greater N400s with more disorganization, and the smaller N400 effects in case of more severe negative symptoms. Although acknowledging that schizotypal personality traits of healthy subjects may be quantitatively and qualitatively different from schizophrenia symptoms, we aimed at testing whether these traits are accompanied by cognitive processes similar to those of patients, which would bring further support to the idea of a continuum between schizophrenia and normality (Peters et al., 1999; van Os et al., 2000). In addition to the above-mentioned use of partial correlations to control for the effect of the other schizotypal factors when focusing on the relation of one of them with the N400, three features characterized the present study. First, we used the protocol of our previous study in patients (Debruille et al., 2007). This was done both to compare results obtained in patients to those in normals and to circumvent the following interpretation problem. In studies where each target word was preceded by a particular context word, the anomalous N400s to targets observed were seen as due to an abnormal priming of targets. This abnormality was thought to be related to a deficient processing of the context or to a problematic maintenance of the prime in working memory. However, if the processing of context words is impaired, the processing of targets also may be impaired and nothing can help to determine which of the two possibilities accounts for the anomalous target N400s. In the semantic categorization task of Debruille et al. (2007), the same category word was used as a context (i.e., first word of the trial) for all the trials for which participants had to make a decision. This was aimed at facilitating the processing of the context so that it produced the same effects on the processing of the target word (i.e., on the second word of the trial) in patients as in normals. A previous study using this particular protocol reported an absence of significant differences between patients and normal controls in the ERPs evoked by such invariant context words (Debruille, Kumar, Saheb, Chintoh, Gharghi, et al., 2010). Thus, if the disorganization trait is found to have an effect on the N400s to the targets of our study, it will be more likely to be due to a deficit in processing the targets themselves, which would, nevertheless, have to be confirmed by an absence of effects of schizotypal traits on each of the components of the ERPs elicited by the context word. The second feature that characterizes the present study is to compute the correlations for each electrode site. For written words, the N400 effect reaches a maximum over midline centroparietal electrodes and is slightly larger over right than left parietal sites (Holcomb, 1988; Kutas, van Petten, & Besson, 1988). Correlations involving the N400 effect can thus be expected to reach a maximum at these sites. The third feature of this study derives from the fact that variations in the size of the N400 effects may have several causes. A smaller N400 effect can be due to larger N400 deflections for matching stimuli, to smaller N400 deflections for mismatching stimuli, or to both. As mentioned above, when discussing the larger N400 deflections found with disorganization and with the interpersonal factor in the matching conditions, such results have

1049 different functional significances. Therefore, correlations for each of the three schizotypal factors were measured three times, once with the N400s elicited by matching stimuli, once with the N400s for mismatching stimuli, and once for the N400 effects.

Methods Participants Forty-nine right-handed participants (35 women, 14 men) aged between 18 and 51 years were recruited by advertisements in an English and in a French newspaper. Participants who answered these ads were asked what their mother tongue was. Only English and French were accepted. The rest of the procedure was carried out in the language of the participants. They had to have normal or corrected-to-normal vision and were screened by telephone and systematically excluded for any history of DSM-IV Axis I psychiatric illnesses, except for depressive episodes that resolved at least 2 years ago. The use of these a priori criteria led to include 2 participants who had suffered from depression. Both had recovered and were off medication for more than 6 years. Participants with a history of head injury with loss of consciousness longer than 10 min were excluded, as well as participants with neurological or medical conditions known to compromise brain function and participants abusing drugs. The procedure to recruit the participants included first a short questionnaire made up of the 16 items of the ‘odd belief and magical thinking’ and ‘ideas of reference’ subscales of the SPQ (Dumas, Bouafia, Gutknecht, Saoud, Dalery, & d’Amato, 2000; Dumas, Rosenfeld, Saoud, Dalery, & d’Amato, 1999; Raine, 1991). These were used to preassess delusional-like ideation over the phone. In the first phase of the recruitment, only those who scored 5 or more out of 16 (the maximal score) were asked to participate in the study (n 5 25) in order to have enough participants with high delusional-like ideation scores. No or low delusional-like ideation participants (n 5 24) were recruited in a second phase among people with delusional-like ideation scores smaller than 5 and whose demographic characteristics (age, gender, and number of years of education) matched those of the high delusional-like ideation score participants. At their arrival in the lab, participants gave written informed consent after the procedures were described according to the criteria of the Research Ethics Board of the Douglas Mental Health University Institute. Before having their ERPs recorded, participants had to complete the SPQ. This questionnaire is based on the DSM-III-R criteria for schizotypal personality disorder and includes nine subscales that can be grouped into three clusters or factors: (1) interpersonal, (2) disorganization, and (3) cognitive-perceptual (Raine et al., 1994). The validity of the whole SPQ has been demonstrated (Raine, 1991). Its clusters have been defined by a factor analysis (Raine et al., 1994) and used in many previous studies (e.g., Dickey et al., 2005; Kiang & Kutas, 2005; Sommer, Daalman, Rietkerk, Diederen, Bakker, et al., 2008). The interpersonal score is computed by adding the scores for the ‘social anxiety,’ the ‘no close friends,’ the ‘constricted affect,’ and the ‘paranoid ideation’ subscale. The disorganization trait score is computed by adding the score for the ‘odd speech’ subscale to the score for the ‘odd or eccentric behavior’ subscale. Both subscales load on the same factor (disorganization) in the general population and in clinical population (Reynolds et al., 2000). The ‘cognitive and perceptual’ cluster could not be used to evaluate

1050

M. Pre´vost et al.

Table 1. Characteristics of the Participants (1st column) and Correlations (Pearson’s r Coefficients) between these Characteristics (following columns) Mean (Stdev) Age Years of study SPQ Interpersonal Disorganization Delusion

30.3 (10.4) 14.3 (1.7) 18.9 (14.1) 7.2 (6.2) 4.5 (3.7) 5.1 (4.2)

Age

Years of study

SPQ

Interpersonal

Disorganization

0.00 " 0.31n " 0.30n " 0.34n " 0.22

" 0.13 " 0.15 " 0.17 " 0.04

0.90nnn 0.85nnn 0.85nnn

0.68nnn 0.62nnn

0.62nnn

Note: SPQ: Total score for the schizotypal personality questionnaire. n po0.05; nnnpo0.001.

delusional-like ideation since it also includes hallucinations. It is thus the scores for the two subscales ‘idea of references’ and ‘odd beliefs and magical thinking’ that were used here, as previously done in a study performed to assess the neurocognitive mechanisms underlying delusional-like ideation in normals (e.g., Woodward, Buchy, Mortiz, & Liotti, 2007). Here, these two scores were simply added. Note that paranoid ideation was not integrated into the delusional-like ideation score because this subscale does not include delusional-like items per se, but rather suspiciousness, which is already included in the interpersonal factor. Pearson correlations between subjects’ characteristics are displayed in Table 1. Each SPQ factor correlated significantly with the two others. Additionally, age negatively correlated with the overall SPQ scores and with the interpersonal and disorganization factors but not with delusional-like ideation. Task Subjects were seated comfortably in a dimly lit room in front of a computer screen placed 1 m from their eyes. Black stimuli were presented on a white background at the center of this screen. Trials were made of two serially presented words. In two-thirds of the trials, the first word was the question word ‘ANIMAL?’ and, in one-third of the trials, the first word was the instruction ‘INACTION.’ These words were then followed by the target word. The stimulus onset asynchrony was 2 sec and each word was displayed for 1 sec. The target word was either an exemplar of the animal category (e.g., lion) or a non-exemplar of this category (e.g., hammer). Subjects had to decide whether or not the target word belongs to the animal category as rapidly and as accurately as possible by pressing one of two keys with their right index finger. The ‘INACTION’ word, introduced so that subjects have to pay attention to the first word, signaled to the participants that they should not respond to the target stimulus, which was also either an exemplar or a non-exemplar of the animal category. The non-exemplar stimuli comprised names of tools, pieces of furniture, office and kitchen objects, buildings and parts of buildings, and transportations means. The target word was followed 2 to 2.5 sec later by the word ‘Blink,’ giving the subjects the opportunity to blink without disrupting the electroencephalogram (EEG) signal of the trial. All target words used were selected from among familiar words using the Content, Mousty, and Radeau’s database (1990) for the French words and the Kucera and Francis (1967) counts for the English words. In the French version, in both the exemplar and the non-exemplar categories, there were forty-seven words with a frequency comprised between 0 and 1000 out of 100 million occurrences, nine

words with a frequency comprised between 1000 and 3000, and four words with a frequency higher than 3000. In the English version, in both categories, there were forty-seven words with a frequency comprised between 0 and 9 out of one million; ten words with a frequency comprised between 10 and 17, and three words with a frequency comprised between 35 and 117. The average length of words in each category did not differ from one another. In the French version, the mean number of letters was 6.5 ! 1.7 in the exemplar category and 6.7 ! 1.5 in the non-exemplar category. In the English version, the mean number of letters was 5.5 ! 1.8 in the exemplar category and 5.75 ! 1.7 in the non-exemplar category. In the action condition, there were 60 trials with exemplar target words and 60 trials with non-exemplar target words. For the ‘INACTION’ condition, there were 30 trials for each of the two stimulus categories. So, for both conditions, the probability of occurrence of an animal target word was 50%. Each target word was presented only once, thus the probability of occurrence of a particular target word was 1 out of 180. Data Acquisition Behavioral responses were recorded for each trial. The EEG was recorded with tin electrodes from the ECI cap (Electro-Cap International, Eaton, OH), which were placed according to the modified expanded 10–20 system (American EEG Society, 1991) with a right ear lobe initial reference. Twenty-eight electrodes were used. Sites were grouped into three subsets. The sagittal subset included Fz, FCz, Cz, and Pz; the parasagittal subset, FP1/2, F3/4, FC3/4, C3/4, CP3/4, P3/4, O1, and O2; and the lateral subset, F7/8, FT7/8, T3/4, TP7/8, T5, and T6. Eye movements and blinks were monitored with F7 and F8 for horizontal movements and with FP2, and an additional electrode placed on the right cheekbone for vertical movements. The impedance was kept below 5KO. The gain of the Contact Precision Instruments amplifiers (Boston, MA) used was set at 20,000. The half amplitude cut-offs of high and low pass frequency filters were set at 0.01 and 100 Hz, respectively. In addition, a 60-Hz electronic notch filter was used. EEG signals were digitized at a 256-Hz sampling frequency1 and stored along with the stimulus and response codes. The EEG was re-referenced off-line to the mean of the left and right earlobes signals. Measures EEG epochs corresponding to trials with incorrect responses or with reaction times shorter than 200 ms or longer than 2000 ms 1 The 256-Hz sampling frequency respects Nyquist’s law, but does not eliminate the possibility of signal distortion in some cases.

Schizotypal traits and N400 in healthy subjects were rejected. We also rejected trials with excessive eye movements, amplifier saturations, or analog-to-digital clippings lasting more than 100 ms. Our baseline was set between ! 200 and 0 ms. Averages were calculated over the time period between target word onset and 1000 ms later. The minimum numbers of artifactfree trials averaged by subject was 28 and 30 for the exemplar and the non-exemplar categories, respectively. The mean number of trials averaged was 49 (SD 8) in both the exemplar category and the non-exemplar category. The N400 amplitude was measured by computing the mean voltage of the ERPs in a 350–550 ms time window, relatively to the baseline. This time window was centered on the negative peak observed on the grand average of the non-exemplar–exemplar ERP subtractions. For each subject, a mean voltage was computed at each electrode for the N400 in the exemplar category, for the N400 in the non-exemplar category, and for the N400 difference (non-exemplar minus exemplar amplitudes). Analyses Mean reaction times (RTs) and accuracies (As) were obtained in the action condition, that is, for all trials except those with INACTION as the prime word. Two subjects were excluded from all analyses because their error rate was greater than 10%. RTs and As were analyzed with separate one-way repeated-measure analysis of variance (ANOVAs) (i.e., t-tests) with category (exemplar versus non-exemplar) as the within-subject factor. Assuming a linear decomposition, partial correlations were run to control for the effect of the two other schizotypal factors. These correlations were run between each of the three factors and RTs on the one hand, and each of the three factors and As on the other hand. This was done in an exploratory way to generate a priori hypotheses for future studies. For mean voltages of ERPs in the N400 time window, a twoways repeated measure omnibus ANOVA was run with category (exemplar vs. non-exemplar) and electrode site as within-subjects factors. The three subsets of electrodes were analyzed separately when an interaction between the category factor and the electrode was observed in the omnibus ANOVA. For the parasagittal and lateral subsets of electrodes, the hemi-scalp (right versus left) was added as a third within-subject factor. We used the Greenhouse and Geisser (1959) procedure to compensate for heterogeneous variances for the factor having more than two levels, that is, the electrode factor. In each case, the original degrees of freedom are reported together with the Epsilon (E) correction factor and the corrected probability level. The mean voltage values of the N400s in the exemplar category, the N400s in the non-exemplar category, and the N400 effect (non-exemplar minus exemplar differences) were correlated (two-tailed) to the delusional-like ideation, the interpersonal and the disorganization factor scores at each electrode, to assess the scalp distribution of the effects. Partial correlations between ERP measures at each electrode and each SPQ subscale scores were then computed. This allowed controlling for the two remaining SPQ factors scores, which was necessary since significant correlations were found between these factors (see Table 1) and since their comparable symptoms have been found to have opposite effects on ERPs in the N400 time window in studies with schizophrenia patients (e.g., Andrews et al., 1993; Kuperberg et al., 2006). Partial correlations were possible since correlations could be assumed to be linear. For the correlation between disorganization trait and interpersonal, the Pearson’s r was 0.68 (see Table 1). The use of a 6 degrees polynomial function

1051 to find the best fit only improves the r to 0.71. The initial straight line was thus an adequate solution. For the correlation between the disorganization trait and delusional-like ideation, the Pearson’s r was 0.62. The use of the same type of complex function only improves the r to 0.64. Finally, for the correlation between delusional-like ideation and the interpersonal trait, for which the Pearson’s r was 0.62, the use of a complex polynomial function only improves the r to 0.64. Pearson, and thus linear correlation, was therefore a good model to represent the association between factors. Results Behavioral Data Mean RTs were shorter in the exemplar (843 ms " 122) than in the non-exemplar (923 ms " 160) category (F(1,45) 5 55.4, po.001). Accuracies (i.e., error rates) did not differ significantly between the exemplar (2.8% " 3.5) and the non-exemplar (2.3% " 3.1) categories. No correlation with delusional-like ideation, disorganization trait, or interpersonal scores was found, either with RTs or with error rates, even when running partial correlations and thus when studying the effect of one factor while controlling for the two others. ERP Data ERPs in the N400 time window were more negative in the nonexemplar than in the exemplar category as showed by the omnibus ANOVA (F(1,45) 5 53.6, po.001). Since the category factor interacted with electrode (F(27,1215) 5 6.5, E 5 0.169, po.001), we conducted ANOVAs for each subset of electrodes and observed that ERP amplitudes in the non-exemplar category were greater than in the exemplar category for the three subsets (sagittal: F(1,46) 5 45.2, po.001, parasagittal: F(1,45) 5 51.5, po.001, and lateral: F(1,46) 5 46.5, po.001), as observed in Figure 1. ERPs were more negative over the left than over the right hemi-scalp at the parasagittal (F(1,45) 5 25.2; po.001) and lateral (F(1,46) 5 13.1, po.001) subsets (Figure 2). In contrast, a category # hemi-scalp interaction at the parasagittal subset (F(1,45) 5 4.2, p 5 .045) revealed that the difference between the exemplar and the non-exemplar categories was larger over the right than over the left side of the scalp (Figures 3 & 4). The post hoc analyses performed showed that this difference was significant for each hemi-scalp at the parasagittal subset (left: F(1,45) 5 44.9, po.001; right: F(1,45) 5 53.8, po.001). The Pearson correlations between each SPQ factor and ERP amplitudes in the N400 time window were assessed for each of the 28 electrodes. Partial correlations were then systematically computed to control for the two other SPQ factors. As displayed in Figure 5, no significant bivariate correlation between delusional-like ideation and ERP amplitudes was found over centroparietal electrodes. Some bivariate correlations were observed at frontal electrode sites that were no longer significant when we controlled for the interpersonal and the disorganization scores. The disorganization scores negatively correlated with ERP voltages in the N400 time window over all the frontal and central electrodes in the exemplar and non-exemplar categories. The higher the disorganization scores, the more negative the ERPs were in the N400 time window at these sites in both categories. These correlations remained significant after adjustment for delusional-like ideation and interpersonal scores, as shown by Figure 5. However, the N400 effect did not correlate with the disorganization factor, which suggests that the increase in

1052

M. Pre´vost et al.

Figure 1. Grand average ERPs for the exemplar and the non-exemplar categories (n 5 47).

negativity induced by the disorganization trait was similar in the exemplar and the non-exemplar category. Regarding the interpersonal factor, no correlation was found with the ERP amplitude in the exemplar and non-exemplar categories after adjustment for delusional-like ideation and disorganization scores (Figure 5). In contrast, there were positive correlations between the N400 effect and interpersonal scores, which remained significant over the left central electrodes when delusional-like ideation and disorganization scores were controlled for. Smaller N400 differences were associated with higher interpersonal scores. As age correlated with disorganization and interpersonal factors, each partial correlation was also run controlling for age in addition to the other factors. Correlations were also run without the two participants with a history of depression. This did not change the correlations found or their scalp distribution.

Figure 2. Spline interpolated voltage maps of the N400 amplitude for the exemplar and the non-exemplar categories in the 350–550 ms time window.

Discussion We investigated in healthy subjects the relationship between delusional-like ideation, disorganization trait, and interpersonal factors, and the N400 potential elicited during a semantic categorization task. As sometimes observed in patients (Debruille et al., 2007; Lenzenweger & Dworkin, 1996), and often in healthy populations (Raine et al., 1994; Reynolds et al., 2000), the scores for each factor were significantly correlated with the scores for the two other factors. Individuals who had higher delusion-like ideation scores also tended to have higher disorganization scores and greater scores for the interpersonal factor (see Table 1). The effect of each schizotypal factor was thus studied while controlling for the effects of the two remaining factors. The N400 protocol used was designed to circumvent the effect of the deficit of

Figure 3. Grand average ERPs of the N400 effect (non-exemplar minus exemplar ERPs) (n 5 47).

Schizotypal traits and N400 in healthy subjects

Figure 4. Spline interpolated voltage maps of the N400 effect (nonexemplar minus exemplar ERPs) in the 350–550 ms time window.

the processing of prior context (Debruille et al., 2010). The source of the variations of N400 effects was assessed by measuring correlations in both the exemplar and the non-exemplar categories. Finally, the correlations were computed at each electrode site, to test whether the relationship under investigation follows the scalp distribution of the N400 effect obtained in the experiment between exemplar and non-exemplar targets. The results obtained demonstrated the relevance of this method. Correlations between these ERPs and disorganization and interpersonal factors were observed, but their scalp topography did not follow that of the N400 effect. The few correlations found between ERPs in the N400 time window and delusion-like ideation became insignificant when controlling for the two other factors (Figure 5). The N400 protocol used herein led to a negative deflection in the N400 time window for both categories. This deflection was widespread over the scalp. The N400 in the non-exemplar category was significantly greater than the N400 in the exemplar category, and this ERP difference had a centro-parietal maximum greater over the right than over the left hemiscalp (Figure 4). These results suggest that our protocol, where a single category is shown as a first word, elicits the classical N400 effect found for written words (Holcomb, 1988; Kutas & Federmeier, 2000). Significant correlations between N400 amplitudes and disorganization scores were found over centro-parietal electrode sites. This replicates the results of previous studies done with patients (Andrews et al., 1993; Kostova et al., 2005; Kuperberg et al., 2006; Salisbury et al., 2000). The more disorganized the subjects, the larger their N400 deflections. In the present study, this was significant not only in the exemplar but also in the non-exemplar category. The correlations remained significant after controlling for the two other factors. Interestingly, no correlation with the N400 effect was observed, as in Kiang and Kutas’ study (2005) with healthy subjects. In both studies, the disorganization trait may have enhanced N400 amplitudes similarly in the exemplar and non-exemplar categories, leaving the N400 effect unchanged. The fact that this was obtained here in a task created to circumvent the effect of the deficit in the processing of the prior context suggests that larger target N400s could be caused by an anomalous semantic processing of target stimuli themselves. Such anomalous processing could have also existed in previous studies, in addition to the deficiency in the processing of the context that was suggested. Most interestingly, while part of the

1053 correlations were observed at N400 sites, that is, at centro-parietal sites, the strongest correlations were found at right frontocentral sites. These results should be considered with caution since, in contrast with centro-parietal locations, they do not correspond to a priori hypotheses. Nevertheless, they suggest that the amplitude of an ERP component other than the N400 could also depend on the severity of the disorganization trait. Future research should be devoted to assess the functional significance of this component. In any case, the N400 effect is known to be generated by multiple brain regions (Halgren, Dhond, Christensen, Van Petten, Marinkovic, et al., 2002), having multiple contributing electrophysiological sub-components (Franklin, Dien, Neely, Huber, & Waterson, 2007). Disorganization may thus affect a sub-component of the N400 that is not maximal over centro-parietal electrodes. As for the correlations observed at centro-parietal sites, given that a semantic categorization task was used, they are most likely due to a modulation of the classical N400 by the disorganization factor. This association of greater N400s with higher disorganization scores may have different functional significances depending on the hypothesis of reference. Regarding the hypothesis that N400 amplitudes index the automatic spreading activation within the lexicon (Deacon, Hewitt, Yang, & Nagata, 2000; Kutas & Hillyard, 1984), our results are consistent with studies using schizophrenia patients (Andrews et al., 1993; Salisbury et al., 2000), which suggest that schizophrenia, and more specifically thought disorders, may be related to an overspread activation in the network of word representations (Maher, Manschreck, Hoover, & Weisstein, 1987; Manschreck, Maher, Milavetz, Ames, Weisstein, & Schneyer, 1988; Spitzer, Braun, Hermle, & Maier, 1993). This idea that words induce a broader activation in morethan in less-disorganized individuals implies that targets also induce broader activation. Therefore, according to the hypothesis that N400 is generated by the activation, all targets should elicit larger N400 deflections, and N400 effects may not change, as observed here. A similar interpretation could be made at the level of the network coding for meanings, following the hypothesis that N400 indexes activation in semantic memory (see Kutas & Federmeier, 2000; Lau, Phillips, & Poeppel, 2008). Alternatively, according to the hypothesis that the amplitude of the N400 is proportional to the efforts deployed by the brain to integrate the meaning of the stimulus in the context (Holcomb, 1993), our results indicate that healthy people with more disorganization scores deploy greater efforts to try to integrate target stimuli than people without disorganization. This may be a consequence of an initial overspread of activation since it arguably increases difficulty. However, the absence of effects of disorganization scores on error rates and reaction times does not provide further support to these lexical and semantic access views of N400s, because these processes are likely to be involved in the categorization task. Alternatively, according to a fourth hypothesis, the N400s index a late inhibition of inappropriate representations (Debruille, 1998; Debruille, Ramirez, Wolf, Schaefer, Nguyen, et al., 2008; for a review, see Debruille, 2007). In the case of the present experiment, these inappropriate representations could be those related to animals that were activated by the question word ‘ANIMAL?’ and that are inappropriate for non-exemplars targets. Due to their overspread of activation, subjects with higher disorganization scores would have activated more of these representations and thus would have more inhibition to perform, hence the greater N400 amplitudes. This hypothesis may also be used to interpret the greater N400s obtained in the exemplar

1054

M. Pre´vost et al.

Figure 5. Spline interpolated maps of the correlation coefficients between mean ERP voltages in the 350–550 ms time window and schizotypal features. Bivariate correlations for each factor are followed by partial correlations that controlled for the two other factors’ scores. The blue colors used for negative correlation coefficients mean that greater N400 amplitudes correlate to higher factor scores, as greater N400s (for exemplars, non-exemplars, and effect) are associated with more negative values. Significant correlations at one electrode site are tagged by a small ring for po.05, by an intermediate ring for po.01, and by a large ring for po.001. No correction of the alpha level was necessary here despite the large number of statistical tests done because of the a priori hypothesis (i.e., that strongest correlation would be observed at centro-parietal sites, that is, at the sites where the N400 effect with written words is usually maximal). A Bonferroni correction taking into account all the tests done (3 traits ! 28 electrodes 5 84) would set the alpha level at 0.0005. With a less conservative correction, considering familywise errors, the alpha would be set at 0.0017 (for 28 electrodes).

condition. The first word ‘ANIMAL’ may activate knowledge corresponding to animals other than the target. This inappropriate activation may be more important in cases of more marked disorganization. Another finding of the present study is that, in the N400 time window, the subtractions of the voltages of the exemplar category from those of the non-exemplar category positively correlated with higher scores for the interpersonal factor of the SPQ. The higher the score, the smaller the N400 effect, since this effect has a negative voltage. These results are consistent with those of a previous study (Kiang & Kutas, 2005), which reported a correlation between ERP amplitude in the N400 time window and interpersonal scores in normals. Our study replicates this finding but also shows that it persists when controlling for the two re-

maining schizotypal factors. Moreover, it suggests that these correlations are maximal over left centro-parietal electrodes (see Figure 5) whereas the typical sites where the N400 effect is maximal for written word stimuli are usually at the right centroparietal part of the scalp (Holcomb, 1988; Kutas et al., 1988). However, these results should be considered with caution since they do not correspond to a priori hypotheses. Nevertheless, they suggest that the correlations may be due to the modulations of an ERP other than the N400, the functional significance of which would have to be investigated. Alternatively, the interpersonal factor might preferentially affect a sub-component of the N400 that is not maximal over centro-parietal electrodes. This left centro-parietal preferential location is difficult to compare with findings in the literature. The correlations reported in other

Schizotypal traits and N400 in healthy subjects

1055

studies pertained to only one scalp location. In Kiang and Kutas (2005), it was the midline parietal electrode (Pz). In their more recent study in schizophrenia patients, Kiang et al. (2007) found that smaller N400 effects at the central electrode, Cz, tend to correlate with the severity of negative symptoms, and thus with symptoms that resemble those measured by the SPQ interpersonal factor. Meanwhile, Kostova et al. (2005) found the same relationship between negative symptoms of patients and the N400 effect at Pz, but this was only a trend (p 5 .08). However, no measures of the correlations between the different clinical factors were made. It thus remains unclear whether it would have been necessary to control for other factors in their subjects’ samples too, and whether the correlation with ERPs would persist with these controls. The maps of the bivariate correlations obtained for the disorganization trait and those obtained for the interpersonal score are interesting to compare to the maps of partial correlations (Figure 5). This comparison reveals that controlling for the two other schizotypal factors may have dramatic effects. At some sites, this control induced a switch in the direction of the correlations. By emphasizing the importance of the control, these findings may help to reconcile some of the discrepancies present in the N400 studies of schizophrenia. These discrepancies could be due to varying patterns of symptom correlations across the population samples studied. Regarding delusion-like ideation, the atypically localized correlations found between the severity of this factor and the amplitude of the ERPs in the N400 time window (Figure 5) disappeared when we controlled for the effect of the two other factors, unlike the correlations found in patients (Debruille et al.,

2007), which resisted this control. Therefore, it seems that delusion-like ideation may not be correlated to N400 in normals and that healthy subjects with delusion-like ideation do not have abnormal semantic processes. The present study thus suggests that delusion-like ideation observed in healthy subjects might not be similar to delusional beliefs of patients as they are not underlain by the same neurocognitive mechanisms. These results are at odds with the findings of Debruille et al. (2007) and Kiang et al. (2007) in schizophrenia patients and could be used to support a distinction between normal population and schizophrenia patients in contrast with the idea of a continuum between the two (Garety, Bebbington, Fowler, Freeman, & Kuipers, 2007; Shevlin et al., 2007; Verdoux & van Os, 2002). However, the results of Pre´vost, Rodier, Lionnet, Brodeur, King, and Debruille (forthcoming) suggest that smaller N400s can be found in healthy subjects with high scores on delusion-like ideation items of the SPQ, when boosting paranoid feelings, which probably diminishes some of the differences between schizophrenia patients and healthy subjects. In summary, the results of the present study suggest that the semantic processes accompanying a particular schizotypal personality trait should be studied while controlling for the other traits. As with other brain imaging techniques, various localizations have to be examined to test whether the correlations actually pertain to the neurocognitive component under focus. Moreover, our study supports the view of a continuum between schizophrenia and normality at least for disorganization and interpersonal factors to the extent that they were found, in healthy subjects, to be accompanied by semantic processes biases similar to those observed in schizophrenia patients.

REFERENCES American EEG Society. (1991). Electrode Nomenclature Committee. Modified expanded 10–20 system. Journal of Clinical Neurophysiology, 35, 200–202. Andrews, S., Shelley, A. M., Ward, P. B., Fox, A., Catts, S. V., & McConaghy, N. (1993). Event-related potential indices of semantic processing in schizophrenia. Biological Psychiatry, 34, 443–458. Bedwell, J. S., Kamath, V., & Compton, M. T. (2009). The relationship between interview-based schizotypal personality dimension scores and the continuous performance test. Schizophrenia Research, 108, 158–162. Cadenhead, K., Kumar, C., & Braff, D. (1996). Clinical and experimental characteristics of ‘‘hypothetically psychosis prone’’ college students. Journal of Psychiatric Research, 30, 331–340. Chapman, L. J., Edell, W., & Chapman, J. P. (1980). Physical anhedonia, perceptual aberration, and psychosis proneness. Schizophrenia Bulletin, 6, 639–653. Claridge, G. (1997). Final remarks and future directions. In G. Claridge (Ed.), Schizotypy. Implications for illness and health (pp. 301–317). Oxford: Oxford University Press. Cohen, J. D., & Servan-Schreiber, D. (1992). Context, cortex, and dopamine: A connectionist approach to behavior and biology in schizophrenia. Psychological Review, 99, 45–77. Coleman, M. J., Levy, D. L., Lenzenweger, M. F., & Holzman, P. S. (1996). Thought disorder, perceptual aberrations, and schizotypy. Journal of Abnormal Psychology, 105, 469–473. Content, A., Mousty, P., & Radeau, M. (1990). Une base de donne´es lexicales informatise´e pour le franc¸ais e´crit et parle´. L’Anne´e Psychologique, 90, 551–566. Deacon, D., Hewitt, S., Yang, C., & Nagata, M. (2000). Event-related potential indices of semantic priming using masked and unmasked words: Evidence that the N400 does not reflect a post-lexical process. Cognitive Brain Research, 9, 137–146. Debruille, J. B. (1998). Knowledge inhibition and N400: A study with words that look like common words. Brain and Language, 62, 202– 220.

Debruille, J. B. (2007). The N400 potential could index a semantic inhibition. Brain Research Review, 56, 472–427. Debruille, J. B., Kumar, N., Saheb, D., Chintoh, A., Gharghi, D., Lionnet, C., et al. (2007). Delusions and processing of discrepant information: An event-related brain potential study. Schizophrenia Research, 89, 261–277. Debruille, J. B., Kumar, N., Saheb, D., Chintoh, A., Gharghi, D., Lionnet, C., et al. (2010). Circumventing the deficit of context processing in schizophrenia: An event-related brain potential study. International Journal of Psychophysiology, 75, 167–176. Debruille, J. B., Ramirez, D., Wolf, Y., Schaefer, A., Nguyen, T. V., Bacon, B. A., et al. (2008). Knowledge inhibition and N400: A within- and a between-subjects study with distractor words. Brain Research, 1187, 167–183. Dickey, C. C., McCarley, R. W., Niznikiewicz, M. A., Voglmaier, M. M., Seidman, L. J., Kim, S., et al. (2005). Clinical, cognitive, and social characteristics of a sample of neuroleptic-naive persons with schizotypal personality disorder. Schizophrenia Research, 78, 297–308. Ditman, T., & Kuperberg, G. R. (2007). The time course of building discourse coherence in schizophrenia: An ERP investigation. Psychophysiology, 44, 991–1001. Dumas, P., Bouafia, S., Gutknecht, C., Saoud, M., Dalery, J., & d’Amato, T. (2000). [Validation of the French version of the Raine Schizotypal Personality Disorder QuestionnaireFCategorial and dimensional approach to schizotypal personality traits in a normal student population]. Encephale, 26, 23–29. Dumas, P., Rosenfeld, F., Saoud, M., Dalery, J., & d’Amato, T. (1999). [Translation and French adaptation of the Raine Schizotypal Personality Questionnaire]. Encephale, 25, 315–322. Franklin, M. S., Dien, J., Neely, J. H., Huber, E., & Waterson, L. D. (2007). Semantic priming modulates the N400, N300, and N400RP. Clinical Neurophysiology, 118, 1053–1068. Freeman, D. (2007). Suspicious minds: The psychology of persecutory delusions. Clinical Psychology Review, 27, 425–457.

1056 Garety, P. A., Bebbington, P., Fowler, D., Freeman, D., & Kuipers, E. (2007). Implications for neurobiological research of cognitive models of psychosis: A theoretical paper. Psychological Medicine, 37, 1377– 1391. Gassab, L., Mechri, A., Dogui, M., Gaha, L., d’Amato, T., Dalery, J., et al. (2006). Abnormalities of auditory event-related potentials in students with high scores on the Schizotypal Personality Questionnaire. Psychiatry Research, 144, 117–122. Greenhouse, G. W., & Geisser, S. (1959). On methods of analysis of profile data. Psychometrika, 24, 1582–1589. Gooding, D. C., Tallent, K. A., & Hegyi, J. V. (2001). Cognitive slippage in schizotypic individuals. The Journal of Nervous and Mental Disease, 189, 750–756. Halgren, E., Dhond, R. P., Christensen, N., Van Petten, C., Marinkovic, K., Lewine, J. D., et al. (2002). N400-like magnetoencephalography responses modulated by semantic context, word frequency, and lexical class in sentences. NeuroImage, 17, 1101–1116. Hardy-Bayle´, M. C., Sarfati, Y., & Passerieux, C. (2003). The cognitive basis of disorganization symptomatology in schizophrenia and its clinical correlates: Toward a pathogenetic approach to disorganization. Schizophrenia Bulletin, 29, 459–471. Hemsley, D. R. (2005). The schizophrenic experience: Taken out of context? Schizophrenia Bulletin, 31, 43–53. Holcomb, P. J. (1988). Automatic and attentional processing: An eventrelated brain potential analysis of semantic priming. Brain and Language, 35, 66–85. Holcomb, P. J. (1993). Semantic priming and stimulus degradation: Implications for the role of the N400 in language processing. Psychophysiology, 30, 47–61. Johns, L. C., & van Os, J. (2001). The continuity of psychotic experiences in the general population. Clinical Psychology Review, 21, 1125–1141. Johnston, A. E., Rossell, S. L., & Gleeson, J. F. (2008). Evidence of semantic processing abnormalities in schizotypy using an indirect semantic priming task. The Journal of Nervous and Mental Disease, 196, 694–701. Kiang, M., & Kutas, M. (2005). Association of schizotypy with semantic processing differences: An event-related brain potential study. Schizophrenia Research, 77, 329–342. Kiang, M., Kutas, M., Light, G. A., & Braff, D. L. (2007). Electrophysiological insights into conceptual disorganization in schizophrenia. Schizophrenia Research, 92, 225–236. Kiang, M., Kutas, M., Light, G. A., & Braff, D. L. (2008). An eventrelated brain potential study of direct and indirect semantic priming in schizophrenia. American Journal of Psychiatry, 165, 74–81. Kostova, M., Passerieux, C., Laurent, J. P., & Hardy-Bayle´, M. C. (2005). N400 anomalies in schizophrenia are correlated with the severity of formal thought disorder. Schizophrenia Research, 78, 285– 291. Kucera, H., & Francis, W. N. (1967). Computational analysis of presentday American English. Providence, RI: Brown University Press. Kuperberg, G. R., Sitnikova, T., Goff, D., & Holcomb, P. J. (2006). Making sense of sentences in schizophrenia: Electrophysiological evidence for abnormal interactions between semantic and syntactic processing. Journal of Abnormal Psychology, 115, 251–265. Kutas, M., & Federmeier, K. D. (2000). Electrophysiology reveals semantic memory use in language comprehension. Trends in Cognitive Sciences, 4, 463–470. Kutas, M., & Hillyard, S. A. (1984). Brain potentials during reading reflect word expectancy and semantic association. Nature, 307, 161– 163. Kutas, M., van Petten, C., & Besson, M. (1988). Event-related potential asymmetries during the reading of sentences. Electroencephalography and Clinical Neurophysiology, 69, 218–233. Lau, E. F., Phillips, C., & Poeppel, D. (2008). A cortical network for semantics: (De)constructing the N400. Nature Review Neuroscience, 9, 920–933. Lenzenweger, M. F. (1999). Schizophrenia: Refining the phenotype, resolving endophenotypes. Behaviour Research and Therapy, 37, 281– 295. Lenzenweger, M. F., & Dworkin, R. H. (1996). The dimensions of schizophrenia phenomenology. Not one or two, at least three, perhaps four. British Journal of Psychiatry, 168, 432–440. Maher, B. A., Manschreck, T. C., Hoover, T. M., & Weisstein, C. C. (1987). Thought disorder and measured features of language pro-

M. Pre´vost et al. duction in schizophrenia. In P. Harvey & E. Walker (Eds.), Positive and negative symptoms in psychosis: Description, research and future directions (pp. 195–215). Hillsdale, NJ: Erlbaum. Manschreck, T. C., Maher, B. A., Milavetz, J. J., Ames, D., Weisstein, C. C., & Schneyer, M. L. (1988). Semantic priming in thought disordered schizophrenic patients. Schizophrenia Research, 1, 61–66. Moritz, S., & Woodward, T. S. (2006). A generalized bias against disconfirmatory evidence in schizophrenia. Psychiatry Research, 142, 157–165. Peters, E. R., Joseph, S. A., & Garety, P. A. (1999). Measurement of delusional ideation in the normal population: Introducing the PDI (Peters et al. Delusions Inventory). Schizophrenia Bulletin, 25, 553– 576. Pre´vost, M., Rodier, M., Lionnet, C., Brodeur, M., King, S., & Debruille, J. B. Paranoid induction reduces N400s of healthy subjects with delusional-like ideation. Psychophysiology, forthcoming. Raine, A. (1991). The SPQ: A scale for the assessment of schizotypal personality based on DSM-III-R criteria. Schizophrenia Bulletin, 17, 555–564. Raine, A., Reynolds, C., Lencz, T., Scerbo, A., Triphon, N., & Kim, D. (1994). Cognitive-perceptual, interpersonal, and disorganized features of schizotypal personality. Schizophrenia Bulletin, 20, 191–201. Reynolds, C. A., Raine, A., Mellingen, K., Venables, P. H., & Mednick, S. A. (2000). Three-factor model of schizotypal personality: Invariance across culture, gender, religious affiliation, family adversity, and psychopathology. Schizophrenia Bulletin, 26, 603–618. Salisbury, D. F., O’Donnell, B. F., McCarley, R. W., Nestor, P. G., & Shenton, M. E. (2000). Event-related potentials elicited during a context-free homograph task in normal versus schizophrenic subjects. Psychophysiology, 37, 456–463. Schul, Y., Mayo, R., & Burnstein, E. (2004). Encoding under trust and distrust: The spontaneous activation of incongruent cognitions. Journal of Personality and Social Psychology, 86, 668–679. Shevlin, M., Murphy, J., Dorahy, M. J., & Adamson, G. (2007). The distribution of positive psychosis-like symptoms in the population: A latent class analysis of the National Comorbidity Survey. Schizophrenia Research, 89, 101–109. Sommer, I. E., Daalman, K., Rietkerk, T., Diederen, K. M., Bakker, S., Wijkstra, J., et al. (2008). Healthy individuals with auditory verbal hallucinations; Who are they? Psychiatric assessments of a selected sample of 103 subjects. Schizophrenia Bulletin, in press. Spitzer, M., Braun, U., Hermle, L., & Maier, S. (1993). Associative semantic network dysfunction in thought-disordered schizophrenic patients: Direct evidence from indirect semantic priming. Biological Psychiatry, 34, 864–877. Strauss, J. S. (1969). Hallucinations and delusions as points on continua function. Rating scale evidence. Archives of General Psychiatry, 21, 581–586. van Os, J., Hanssen, M., Bijl, R. V., & Ravelli, A. (2000). Strauss (1969) revisited: A psychosis continuum in the general population? Schizophrenia Research, 45, 11–20. Verdoux, H., & van Os, J. (2002). Psychotic symptoms in non-clinical populations and the continuum of pychosis. Schizophrenia Research, 54, 59–65. Wan, L., Crawford, H. J., & Boutros, N. (2006). P50 sensory gating: Impact of high vs. low schizotypal personality and smoking status. International Journal of Psychophysiology, 60, 1–9. Wang, J., Miyazato, H., Hokama, H., Hiramatsu, K., & Kondo, T. (2004). Correlation between P50 suppression and psychometric schizotypy among non-clinical Japanese subjects. International Journal of Psychophysiology, 52, 147–157. Wilson, C. M., Christensen, B. K., King, J. P., Li, Q., & Zelazo, P. D. (2008). Decomposing perseverative errors among undergraduates scoring high on the Schizotypal Personality Questionnaire. Schizophrenia Research, 106, 3–12. Woodward, T. S., Buchy, L., Moritz, S., & Liotti, M. (2007). A bias against disconfirmatory evidence is associated with delusion proneness in a nonclinical sample. Schizophrenia Bulletin, 33, 1023–1028.

(Received January 31, 2009; Accepted November 21, 2009)

Psychophysiology, 47 (2010), 1057–1065. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01025.x

BRIEF REPORT

The role of temporal predictability in the anticipatory biasing of sensory cortex during visuospatial shifts of attention

JESSICA J. GREEN and JOHN J. McDONALD Department of Psychology, Simon Fraser University, Burnaby, British Columbia, Canada

Abstract The presentation of an attention-directing cue elicits a lateralized ERP deflection called the late directing attention positivity (LDAP) and lateralized changes in alpha-band elelctroencephalogram oscillations. Both of these electrophysiological responses have been independently linked to biasing of visual cortex in anticipation of an impending target. However, the LDAP is not always observed, and the link between the ERP and alpha-band modulations remains unclear. Here, we examined the effect of advance knowledge of the time of target onset on the ERP and alphaband responses to cues. The LDAP was present only when the attention-directing cues accurately indicated the time of target appearance, whereas two sequential attention-related alpha-band modulations were observed regardless of the temporal information provided by the cues. Thus, alpha-band activity may be a more reliable index of pretarget biasing of visual cortical activity than lateralized ERP effects. Descriptors: Cognition, EEG/ERP, Normal volunteers

attention-directing cues that provide advance information about the upcoming target’s likely whereabouts (e.g., Harter, Miller, Price, LaLonde, & Keyes, 1989; Hopf & Mangun, 2000; Mangun, 1994). By comparing the activity that occurs over the hemispheres ipsilateral and contralateral to the to-be-attended direction indicated by the cue, lateralized activity that varies with the direction of attention can be identified. Researchers have identified possible electrophysiological indices of anticipatory priming of visual cortex using two methods for extracting eventrelated changes in the electroencephalogram (EEG). First, examination of the ERPs typically shows a component called the late directing attention positivity (LDAP; Harter et al., 1989), which appears as a relative positivity over posterior electrode sites contralateral to the to-be-attended location beginning 400– 500 ms after the onset of an attention-directing cue. Second, examination of changes in EEG oscillations typically show relative increases in alpha-band (8–14 Hz) activity over posterior electrode sites ipsilateral to the to-be-attended location in the same time period as the LDAP (Kelly, Lalor, Reilly, & Foxe, 2006; Rihs, Michel, & Thut, 2007; Thut, Nietzel Brandt, & Pascual-Leone, 2006; Worden, Foxe, Wang, & Simpson, 2000; Yamagishi, Goda, Callan, Anderson, & Kawato, 2005). Although the LDAP and alpha effects have been linked independently to the anticipatory priming of visual cortex in preparation for an upcoming visual target, a number of questions remain regarding the functions of these electrophysiological responses and the relationship between them. There is some evidence that the presentation of a stimulus induces the phase

It is well established that voluntarily orienting attention to a spatial location in preparation for an impending visual stimulus will enhance processing of that stimulus once it occurs at that location (e.g., Kastner & Ungerleider, 2000). Detection and identification of stimuli appearing at the attended location are faster and more accurate than to stimuli appearing at unattended locations (Luck, Hillyard, Mouloua, & Hawkins, 1996; Prinzmetal, McCool, & Park, 2005). Typically, this improvement in performance for attended stimuli is accompanied by the enhancement of event-related potential (ERP) components that are thought to reflect early visual processing (e.g., the P1 and N1; Hillyard & Anllo-Vento, 1998; Luck et al., 1994; Mangun, 1994; Mangun & Hillyard, 1991). These attentional modulations of stimulus processing are assumed to be the result of anticipatory activity in visual cortex occurring prior to the onset of the target as a result of top-down attentional control (Kastner & Ungerleider, 2000), effectively priming the neural areas that will respond to the target once it appears. To examine the pretarget biasing of visual cortex, researchers have focused on the electrophysiological responses elicited by

This study was supported by grants from the Natural Sciences and Engineering Research Council of Canada, Canadian Foundation for Innovation, Canada Research Chairs program, and British Columbia Knowledge Development Fund. Address correspondence to: Jessica J. Green, Center for Cognitive Neuroscience, Duke University, Box 90999, Durham, NC 27708, USA. E-mail: [email protected] 1057

1058 resetting of ongoing oscillations to be time-locked to the stimulus and that this plays a role in generating ERPs (Makeig et al., 2002; Mazaheri & Picton, 2005). It has also been speculated that the EEG oscillations and ERPs share common neural generators (Engel, Fries, & Singer, 2001; LaBerge, 2005; Mazaheri & Picton, 2005). Thus, modulations of the ERPs and alpha-band EEG could reflect the same anticipatory attention processes. However, the ERP and alpha effects could also reflect independent, albeit complementary, processes. The LDAP has been interpreted as reflecting primarily the enhancement of activity contralateral to the to-be-attended location (e.g., Harter et al., 1989), whereas the lateralized alpha effects have been interpreted as reflecting either the enhancement of the to-be-attended location (Thut et al., 2006; Yamagishi et al., 2005) or the active suppression of to-beignored locations (e.g., Kelly et al., 2006; Rihs et al., 2007; Worden et al., 2000). Despite the frequent interpretation of the LDAP as reflecting anticipatory biasing of visual cortex, there is enough variability in the ERP effects across attention-cuing studies to question the reliability of these components in indexing anticipatory processes. Some studies report a relative negativity in the lateralized ERP, rather than a positivity, that is sustained until target onset (Grent-’t-Jong & Woldorff, 2007; Van der Stigchel, Heslenfeld, & Theeuwes, 2006; Worden et al., 2000). In addition, the LDAP is not always sustained until the onset of the target, as would be expected if it reflected pretarget biasing processes, and seems to occur at the same latency regardless of the time interval between the cue and the target (van Velzen, Forster, & Eimer, 2002). Finally, and most importantly for the current study, in a few studies no LDAPs were observed, despite the presence of other cue-elicted ERP activities and attention effects on target processing (Mangun, 1994; Nobre, Sebestyen, & Miniussi, 2000; Talsma, Slagter, Nieuwenhuis, Hage, & Kok, 2005; Yamaguchi, Tsuchiya, & Kobayashi, 1994). Thus, the determinants of the LDAP and the link between the LDAP and the increased alphaband activity are not yet clear. One commonality between the few studies that have not observed an LDAP is the use of multiple, randomly intermixed stimulus-onset asynchronies (SOAs) between the attention-directing cue and the target stimulus. Most ERP studies of the trialby-trial cuing of attention have utilized a single cue–target SOA to maximize the signal-to-noise ratios of target ERPs. As a consequence, the appearance of the target stimulus is temporally predictable as well as spatially predictable in these single-SOA experiments. Because attention can be directed not only to where a target is likely to appear but also to when a target is likely to appear (e.g., Coull & Nobre, 1998; Miniussi, Wilding, Coull, & Nobre, 1999), it is possible that the LDAP reflects a process that occurs only when attention is oriented in both space and time. For example, orienting attention to the cued location might lead to changes in the ongoing EEG rhythms within visual cortical regions in anticipation of the expected target, but the changes in EEG rhythms might be precisely phase-locked across trials only when the onset of the target is temporally predictable. Consistent with this possibility, the temporal predictability of a stimulus event is known to lead to entrainment of neural oscillations (Lakatos, Karmos, Mehta, Ulbert, & Schroeder, 2008; Large & Jones, 1999). Such entrainment would reduce variability in the phases of event-related EEG oscillations across trials, thereby making it more likely for temporally predictable changes in EEG oscillations to be visible in the ERPs (i.e., to remain after epochs of EEG are averaged; e.g., Pfurtscheller & Lopes da Silva, 1999).

J.J. Green and J.J. McDonald Upon first consideration, the hypothesis that the LDAP and alpha-band modulations reflect the same underlying attentionalbiasing process appears unlikely because the LDAP appears as a modulation of slow-wave activity that lasts for several hundred milliseconds, whereas one alpha cycle lasts for only 70–125 ms. However, recent studies have demonstrated that modulations of the higher frequency alpha-band oscillations can produce lower frequency changes in the ERPs due to asymmetric amplitude modulations of the oscillations (Mazaheri & Jensen, 2008) or alpha-associated baseline shifts that remain in the ERPs after averaging (Nikulin et al., 2007). Moreover, the event-related changes in alpha-band oscillations also show up as slow, longduration changes using standard methods to average the induced activities (e.g., Worden et al., 2000). Thus, the LDAP could reflect an alpha-band oscillatory process that always occurs during voluntary spatial orienting but is only visible in the ERP when the time of target onset is known in advance and the oscillatory activity becomes phase-locked. On the basis of this hypothesis, when the onset of the targets is unpredictable, as is the case when multiple SOAs are utilized, no LDAP should be observed. The goals of the present study were (a) to test the hypothesis that the LDAP would only be observed when the onset of the target was temporally predictable and (b) to examine the LDAP and alpha modulations in the same task to determine if they are modulated in a similar manner by manipulations of temporal predictability. To this end, we performed two variants of a visual cueing experiment that included three randomly intermixed cue– target SOAs (300 ms, 900 ms, and 1500 ms). In one variant of the task the cues provided advance information about the target’s location but provided no information about when the target would appear. In the other variant of the task, the cues provided information about both the target’s location and time of appearance. We predicted that the LDAP would be observed only when the onset of the target was predictable, but that alpha-band EEG modulations that index anticipatory biasing of visual cortex would be observed regardless of the temporal predictability of the target onset.

Methods The Simon Fraser University research ethics board approved all experimental procedures. Participants Twenty-nine neurologically typical university students participated in the experiment after providing informed written consent. Fifteen people participated in Experiment 1a and 14 people participated in Experiment 1b. Data from 3 participants from Experiment 1a and 2 participants from Experiment 1b were discarded because of excessive eye movement or blink artifacts. The remaining subjects (Experiment 1a: 12 participants, 6 female, 11 right-handed, mean age 5 18.9 years; Experiment 1b: 12 participants, 6 female, all right-handed, mean age 5 20.3 years) all reported normal or corrected-to-normal visual acuity and normal color vision. Stimuli and Apparatus The experiment was conducted in a sound attenuated, electrically shielded chamber containing a 19-in. CRT monitor for stimulus presentation (see McDonald & Green, 2008, for a detailed de-

Temporal predictability and anticipatory biasing

1059

scription of our stimulus presentation and EEG recording apparatus). All visual stimuli were presented on a black monitor with a background luminance of 0.02 cd/m2 and a viewing distance of 65 cm. Throughout the experiment, three unfilled gray squares (2.51 ! 2.51, RGB 5 201, 201, 201; 32.2 cd/m2) were presented on the screen as landmarks indicating the potential target locations, along with a small gray fixation box presented at the center of the screen. Two of the landmarks were presented to the left and right of fixation, 1.51 above the horizontal meridian. The third landmark was presented directly above fixation. All landmarks were presented 81 (center to center) from the fixation point. Cue stimuli consisted of a black circle overlaid with diagonal lines that were colored cyan (RBG 5 0, 255, 255; 50.2 cd/m2), yellow (RGB 5 255, 255, 0; 55.4 cd/m2), or magenta (RGB 5 255, 0, 255; 29.3 cd/m2). These colored lines were arranged to form either a leftward, rightward, or upward pointing arrow-like shape (see Figure 1) 1.51 in diameter that was presented for 100 ms directly above fixation. Target stimuli consisted of five parallel horizontally or vertically oriented bars presented within one of the three target landmarks for 50 ms. The target lines were initially a medium gray (RGB 5 100, 100, 100), and luminance was varied across trials depending on the participants’ performance (see Procedure for details). The mask was a checkerboard pattern that remained visible until a response was made. Procedure Participants were instructed to maintain fixation on the fixation box throughout the experiment. Each trial began with the presentation of a leftward-, rightward-, or upward-pointing cue.

After a SOA of 300, 900, or 1500 ms, a target was presented at one of the three possible target locations. On two thirds of the trials the target appeared at the location indicated by the cue stimulus. On the remaining one-third of trials the target appeared in one of the two uncued locations. Following the target a checkerboard mask was presented and remained on screen until the participant made a response. The task was to discriminate the orientation of the target (vertical or horizontal) and respond with a button press. To motivate participants to attend to the cued location, the task was made difficult by adapting the luminance of the target bars to keep target discrimination accuracy across all trials at 80%. This was done using parameter estimation by sequential testing (PEST; Taylor & Creelman, 1967). The PEST algorithm reduced or increased target luminance on the next trial when overall accuracy for target discrimination fell above or below 80%. This ensured that accuracy would be below ceiling and also enabled us to look at different levels of accuracy across different cue-validity conditions. Each participant performed a single 1-h session comprised of 864 trials, with a short rest break given after every 36 trials. In total, there were 96 trials for each type of cue (left, right, and up) at each cue–target SOA (300, 900, and 1500 ms). In Experiment 1a, the color of the cue stimulus indicated, with 100% validity, the time of target onset. For example, for one participant the cyan cue indicated that the target would appear 300 ms following the cue, the yellow cue indicated that the target would appear 900 ms following the cue, and the magenta cue indicated that the target would appear 1500 ms following the cue. The color–SOA associations were counterbalanced across par-

Mask until response

Target 50 ms

ISI 300, 900, or 1500 ms

Cue 100 ms

Figure 1. Sequence of events on a single trial.

1060 ticipants. In Experiment 1b, the color of the cue was not predictive of the cue–target SOA, and the target was equally likely to occur 300, 900, or 1500 ms following the cue. Electrophysiological Recording Electrophysiological recordings were obtained from 63 tin electrodes, using a modified 10–10 montage that included five nonstandard sites inferior to the standard occipital locations (Green, Conder, & McDonald, 2008). All EEG signals were referenced to the right mastoid. The horizontal electrooculogram (HEOG) was recorded bipolarly using two electrodes positioned lateral to the external canthi. Electrode impedances were kept below 15 kO. All signals were recorded with a bandpass of 0.1–100 Hz (! 12 dB/octave, 3 dB attenuation) and digitized at a rate of 500 Hz. Artifact rejection was performed using a semi-automated procedure to remove epochs that contained horizontal eye movements, blinks, and amplifier blocking (Green et al., 2008). Eye movements were detected on the HEOG channel and blinks were detected at FP1, located over the left eye. Artifact-free data were then used to create ERP waveforms. The averaged waveforms were digitally low-pass filtered (! 3 dB point at 25 Hz) and digitally rereferenced to the average of the left and right mastoids. Data Analysis Behavioral performance. The effects of spatial cueing and temporal predictability on behavioral performance were assessed by statistically evaluating target discrimination RTs. Because target discrimination accuracy could vary between experimental conditions (i.e., only overall task accuracy was controlled via our adaptive procedure), we also performed an analogous statistical test on accuracy to ensure that RT modulations were not the result of speed–accuracy trade-offs. Each analysis of variance (ANOVA) had within-subject factors for cue type (valid, invalid) and SOA (300 ms, 900 ms, 1500 ms) and a between-subjects factor of temporal predictability (predictable vs. unpredictable cues). Greenhouse–Geisser adjusted p values are reported where appropriate. Paired-samples t tests were then performed to examine spatial cueing effects for each SOA and temporal predictability separately by comparing performance for validly and invalidly cued targets. Planned pairwise comparisons were twotailed and Bonferroni adjusted to maintain familywise error rates at .05. Cue-elicted activity. To examine ERPs elicited by the cues, EEG epochs were averaged separately for leftward- and rightward-directing cue stimuli for each of the 900 ms and 1500 ms SOAs. Analyses were not performed for the 300 ms SOA as the LDAP does not typically begin until approximately 400 ms following the cue, which would overlap with the target ERPs in the 300-ms SOA condition. ERPs elicited by shift-left and shift-right cues were collapsed to examine waveforms recorded contralaterally and ipsilaterally to the cued location. To examine alpha-band activity elicited by the cues, the EEG epochs were first transformed into the time-frequency domain using temporal-spectral evolution (TSE). The TSE procedure involves first band-pass filtering the EEG in the frequency of interest (8–14 Hz) and then full-wave rectifying the signal before averaging the epochs separately for leftward- and rightward-directing cues (Salmelin & Hari, 1994). The resulting waveforms were then plotted as a percentage increase or decrease relative to a 200-ms prestimulus baseline interval and collapsed in the same

J.J. Green and J.J. McDonald manner as the ERP waveforms to examine alpha-band activity ipsilateral and contralateral to the cued location. For both the ERP and alpha-band EEG activities, mean amplitudes of the contralateral and ipsilateral waveforms were measured for each subject at occipital (PO7/PO8) electrode sites. Amplitudes were measured relative to a 200-ms prestimulus baseline in two time intervals, one during the typical onset of the LDAP (400–600 ms) and one at the end of the cue–target interval immediately preceding the target (700–900 ms in the 900-ms SOA condition and 1300–1500 ms in the 1500-ms SOA condition). Paired-samples t tests between amplitudes ipsilateral and contralateral to the cued location were then performed to test for the presence or absence of lateralized differences in activity for each SOA (900 ms and 1500 ms) and cue type (temporally predictive or nonpredictive) separately. Topographical mapping. To characterize the scalp topography of the LDAP and alpha-band modulations, we used a variant of the antisymmetric mapping method (Praamstra, Stegman, Horstink, & Cools, 1996). The contralateral–ipsilateral voltage differences measured at homologous left and right electrode sites (e.g., PO7/PO8) are projected to one side of the scalp and copied to the other side of the scalp with their polarities inverted. The resulting maps are perfectly symmetric but have foci of opposite polarity in the two hemispheres. Voltages at midline electrode sites are set to zero (Green et al., 2008).

Results Behavioral Measures As can be seen in Table 1, participants performed well on the target discrimination task and were generally faster and more accurate to respond to targets appearing at validly cued locations than to targets appearing at invalidly cued locations. For RTs, the between-groups comparison revealed a main effect of cue validity, F(1,22) 5 53.65, po.0001, but no interaction between validity and temporal predictability, F(1,22) 5 .18, p 5 .68, indicating that the effect of spatial cueing on RTs was the same for both groups. The main effect of SOA, the validity–SOA interaction, and their interactions with temporal predictability were also nonsignificant (all Fso1.5, all ps4.23). Pairwise comparisons to assess RTcueing effects separately at each SOA revealed significant cueing effects for all SOAs when the cues were temporally predictive and for the 300-ms and 900-ms SOAs when the cues were temporally nonpredictive (see Table 1). A cueing effect was apparent for the nonpredictive 1500-ms SOA, but the effect was not significant after correction for multiple comparisons. Although the experiment was designed to keep overall accuracy near 80%, target discrimination accuracy was generally higher for validly cued targets than for invalidly cued targets, ruling out a speed–accuracy trade-off. In the between-groups comparison, a significant main effect of cue validity was observed, F(1,22) 5 12.62, p 5 .002, but there was no significant interaction between cue validity and temporal predictiveness, F(2,22) 5 .02, p 5 .88. The main effect of SOA, the validity–SOA interaction, and their interaction with temporal predictability were also nonsignificant (all Fso2.53, all ps4.09). Planned pairwise comparisons of the cueing effects on discrimination accuracy revealed that significant accuracy effects were only observed for the 300- and 900-ms SOAs for the temporally predictive cues. For the temporally nonpredictive cues, the cue-

Temporal predictability and anticipatory biasing

1061

Table 1. Summary of Behavioral Performance 300-ms SOA Temporally predictive RT (mean [SEM]) Valid 655 (30.5) Invalid 682 (32.5) Cueing effect " 26.8 (8.9)nn ACC (% corr [SEM]) Valid 80.2 (1.8) Invalid 75.2 (1.7) Cueing effect 5.0 (0.9)nn Temporally nonpredictive RT (mean [SEM]) Valid 656 (34.6) Invalid 693 (36.6) Cueing effect " 36 (10.1)nn ACC (% corr [SEM]) Valid 80.4 (2.0) Invalid 78.3 (2.0) Cueing effect 2.1 (2.1)

900-ms SOA

1500-ms SOA

649 (24.7) 685 (27.7) " 36 (9.5)nn

642 (27.9) 687 (32.9) " 45 (9.5)nn

79.7 (2.0) 75.2 (2.0) 4.5 (1.8)nn

78.3 (1.8) 79.6 (2.6) " 1.3 (2.0)

656 (32.5) 693 (38.3) " 36 (11.7)nn

656 (35.8) 680 (34.0) " 24 (9.2)n

79.4 (1.1) 75.7 (1.6) 3.8 (1.7)n

80.1 (1.0) 78.4 (2.2) 1.7 (2.1)

n po.05; nnpo.016 (significant after Bonferroni correction for three pairwise tests).

ing effect for the 900-ms SOA was largest (! 4%) but was not significant after corrections for multiple comparisons. Lateralized Cue-Elicited Activity As can be seen in Figure 2, the LDAP was present in the ERP waveforms following temporally predictive cues but not temporally nonpredictive cues. For temporally predictive cues, the LDAP appeared in the same general time rangeFapproximately 400–700 ms after cue onsetFregardless of the actual SOA (900 or 1500 ms). No reliable LDAP was observed in the early time Temporally Predictive

interval (400–600 ms) following temporally nonpredictive cues or in the 200 ms immediately preceding the target for either temporally predictive or nonpredictive cues. In contrast, as can be seen in Figure 3, alpha-band modulations began approximately 300 ms postcue and were sustained until the onset of the target regardless of SOA or temporal predictability of the cues. Mean amplitudes of the contralateral-minus-ipsilateral differences and the statistical results for both the LDAP and alpha-band EEG modulations can be seen in Table 2. Although both the LDAP, when present, and the alpha-band modulations showed maxima over posterior occipital scalp sites, there were variations in scalp topography that suggested the occurrence of multiple processes. The LDAP was maximal at occipital electrodes PO7/PO8 but extended laterally toward the temporal lobes. In contrast, during the same time interval, the alpha activity was maximal at more inferior electrodes (I3/I4) and spread dorsally to more parietal scalp sites (e.g., P3/P4). In addition, the early and late periods of the alpha activity also showed some variability in topography, with the activity immediately preceding the target focused primarily around lateral occipital scalp sites PO7/PO8. This change in scalp topography over time suggests that the sustained alpha-band modulation observed until the end of the cue–target interval may reflect multiple, sequential processes rather than simply sustained pretarget biasing in visual cortex. To explore the spatiotemporal pattern of alpha-band modulation, we isolated the lateralized activity elicited by leftward- and rightward-directing cues by subtracting out the activity elicited by the upward-directing cues. All cues were expected to elicit similar sequences of attentional control and pretarget anticipatory preparation. Unlike leftward- and rightward-directing cues, however, the upward-directing cues should not have elicited any lateralized activity because the to-be-attended location was on Temporally Non-predictive

_ 1µV

1500 ms SOA

900 ms SOA

+

Figure 2. Lateralized ERP waveforms elicited by leftward- and rightward-directing cues. Waveforms for leftward- and rightward-directing cues were collapsed to display waveforms contralateral and ipsilateral to the to-be-attended direction. For each condition the early and late time intervals used for statistical analysis are demarcated with boxes. Shaded boxes indicate time intervals where a significant difference between contralateral and ipsilateral waveforms was observed. Unshaded boxes indicate that no difference between contralateral and ipsilateral waveforms was detected. The scalp topography of the contralateral–ipsilateral difference for each time interval tested is displayed below the corresponding ERP waveforms.

1062

J.J. Green and J.J. McDonald Temporally Predictive

Temporally Non-predictive

900 ms SOA

5%

1500 ms SOA

Contralateral to cued location Ipsilateral to cued location Cued to midline location

Figure 3. Alpha-band TSE waveforms elicited by leftward-, rightward-, and upward-directing cues. Waveforms for leftward- and rightward-directing cues were collapsed to display waveforms contralateral and ipsilateral to the to-be-attended direction. For more accurate comparison to the ipsilateral and contralateral waveforms, activity elicited by upward-directing cues was averaged across left- and right-hemisphere electrodes to remove hemispheric main effects. For each condition the early and late time intervals used for statistical analysis are demarcated with boxes. Shaded boxes indicate time intervals where a significant difference between contralateral and ipsilateral waveforms was observed. Unshaded boxes indicate that no difference between contralateral and ipsilateral waveforms was detected. The scalp topography of the contralateral–ipsilateral difference for each time interval tested is displayed below the corresponding TSE waveforms.

the vertical meridian (i.e., was nonlateralized). The resulting difference waves (e.g., left cue minus up cue) would help to pinpoint lateralized changes in alpha-band activity related to attention control (for additional discussion, see McDonald & Green, 2008). As can be seen in Figure 3, during the early time interval (400–600 ms) the activity following upward-directing cues is similar to the activity observed over ipsilateral scalp following leftward- and rightward-directing cues, suggesting that during this time interval the modulations of activity are primarily occurring in contralateral cortex. This pattern begins to reverse around 700 ms postcue, and by the end of the cue–target interval the activity following upward-directing cues is similar to the activity observed over contralateral scalp following leftward- and

rightward-directing cues. This suggests that during the time interval immediately preceding the target the alpha-band modulations are primarily occurring in ipsilateral cortex (i.e., contralateral to the to-be-ignored location). The transition from an early contralateral decrease in alphaband activity to a late ipsilateral increase in alpha-band activity was apparent for both temporally predictive and nonpredictive cues at both the 900-ms and 1500-ms SOA and showed similar patterns for both leftward- and rightward-directing cues, relative to upward-directing cues. Figure 4 shows the scalp topography for the isolated alpha-band modulations for the 1500-ms SOA (activity for the 900-ms SOA was nearly identical, so only the 1500-ms SOA is shown here for simplicity). During the early time

Table 2. Summary of Lateralized ERP and Alpha-Band EEG Activity Observed at Occipital Electrode Pair PO7/PO8

Temporally predictive 900-ms SOA Mean (! SEM) t value 1500-ms SOA Mean (! SEM) t value Temporally nonpredictive 900-ms SOA Mean (! SEM) t value 1500-ms SOA Mean (! SEM) t value po.05; nnpo.016.

n

LDAPFearly

LDAPFlate

AlphaFearly

AlphaFlate

.52 (.12) 4.37nn

" .13 (.21) " .61

" 4.02 (1.27) " 3.17nn

" 4.89 (1.52) " 3.21nn

.64 (.21) 2.96nn

" .03 (.21) " .12

" 3.01 (.89) " 3.36nn

" 9.48 (2.73) " 3.40nn

.07 (.12) .59

" .21 (.11) .08

" 3.45 (1.49) " 2.30n

" 5.24 (1.96) " 2.68n

.16 (.26) .54

" .08 (.31) " .25

" 4.20 (1.19) " 3.54nn

" 5.37 (1.77) " 3.03nn

Temporal predictability and anticipatory biasing

1063

Temporally Predictive Left - Up

Right - Up

Temporally Non-predictive Left - Up

Right - Up

Early (400–600 ms)

Late (1300–1500 ms)

Figure 4. Scalp topographies of lateralized alpha-band activity isolated using the upward-directing cue in both early and late time intervals. As the pattern of activity was the same for both the 900-ms and 1500-ms SOAs, only the maps for the 1500-ms SOA are displayed.

period the isolated alpha activity was apparent primarily over midline scalp sites but was slightly lateralized in the contralateral direction. During the late time period the isolated alpha activity was apparent at more lateral scalp sites ipsilateral to the to-beattended location.

Discussion The LDAP component has been considered as one of the more reliable ERP indices of attentional preparation in spatial cueing tasks. Indeed, there is increasing evidence that two other cueelicited ERP components, the EDAN (e.g., van Velzen & Eimer, 2003) and the ADAN (e.g., Green et al., 2008), that were initially thought to reflect deployment of attention to the cued location reflect processes other than those necessary for the voluntary control of attention. Despite the general robustness of the LDAP (e.g., it occurs even when attention is cued in other sensory modalities; Green & McDonald, 2006; Green, Teder-Sa¨leja¨rvi, & McDonald, 2005), several puzzling questions have remained. The present study dealt with two of these questions. First, why is the LDAP sometimes absent when it otherwise appears that observers deployed attention to the cued location? Second, are the LDAP and alpha-band modulations, both considered indices of attentional preparation, similarly modulated by temporal predictability? Regarding the first question, we hypothesized that the temporal predictability of target onset was an important factor that influenced the presence of the LDAP, as the handful of studies that did not report any pretarget biasing effects used unpredictable intervals between the cue and target (Nobre et al., 2000; Talsma et al., 2005; Yamaguchi et al., 1994). Our results confirmed this hypothesis by showing that the LDAP was present following temporally predictive cues but was absent following temporally nonpredictive cues. Thus, the presence of the LDAP in prior studies can be linked to the cue being predictive of both where and when the target was likely to appear. Interestingly, two visual cueing studies with variable cue–target SOAs did report biasing-related activity in their ERPs (Grent-’t-Jong & Woldorff, 2007; Jongen, Smulders, & van der Heiden, 2007). At first glance, these observations are inconsistent with our hypothesis that the presence of the LDAP is dependent on the temporal predictability of target onset. However, both studies used only two randomly intermixed SOAs, whereas the

current study and previous studies that did not observe an LDAP (Mangun, 1994; Nobre et al., 2000; Talsma et al., 2005; Yamaguchi et al., 1994) all used three or more SOAs. It is possible that when only two SOAs are used, the uncertainty about when the target will appear is minimal, and the LDAP is still visible. For example, participants could prepare for the shortest SOA on every trial, and if the target did not occur at the expected time it was known that it must occur at the later SOA, allowing participants to effectively prepare for the target. When three or more SOAs are randomly intermixed, it becomes increasingly difficult to use temporal information to prepare for the target, and the LDAP is no longer observed. Regarding the second question, we compared the LDAP to lateralized modulations of the alpha-band EEG, another seemingly reliable electrophysiological index of attentional preparation, to determine if they were modulated in a similar manner by temporal predictability. Our results showed different patterns for the LDAP and alpha activities. The LDAP was only present following temporally predictive cues, whereas alpha-band EEG was modulated in a similar manner regardless of the temporal predictability of the target onset. One possible explanation for the discrepancy between the LDAP and alpha activities is that they reflect the same underlying process, but that induced activities that are visible in the alpha-band EEG are only visible in the ERPs when the temporal interval between the cue and target is predictable and the phase of the alpha oscillations can be entrained to the target onset. This may account for the absence of the LDAP following temporally nonpredictive cues, but it cannot fully explain the pattern of ERP and alpha effects observed. Although the LDAP and alpha modulations begin at approximately the same time (around 400 ms postcue), the LDAP appears to fizzle out by 700 ms whereas the alpha modulations continue until the onset of the target. In addition, alpha-band activities isolated using a cue to shift attention to a nonlateralized location showed two distinct phasesFan early contralateral decrease in alpha that occurs in the same time interval as the LDAP and a later ipsilateral increase in alpha. It is possible that the LDAP and the early phase of the alphaband modulations reflect the same neurocognitive process, but that the alpha oscillations are not phase-locked following temporally nonpredictive cues and therefore are not visible in the ERP. However, in a previous visual cueing experiment we utilized an upwarddirecting cue to isolate the lateralized LDAP in the same manner as the alpha-band activity was isolated in the present study. In that

1064

J.J. Green and J.J. McDonald

study the isolated LDAP appeared as a negativity ipsilateral to the tobe-attended location rather than as a positivity contralateral to the tobe-attended location that this ERP component has been presumed to reflect (McDonald & Green, 2008). In that study the isolated LDAP also showed a much more lateral scalp distribution than the predominantly medial distribution of early isolated alpha-band activity observed in the current study. Thus, it seems unlikely that the LDAP and early alpha activities measure the same underlying process. Additionally, in the present study, an ipsilateral increase in alpha activity was observed in the interval immediately preceding the target, regardless of the temporal predictability of the cue stimuli. This ipsilateral activity was unique to the alpha-band EEG modulations and highly consistent with the observation of ipsilateral increases in theta-band activity in the inferior occipital gyrus in the few hundred milliseconds before target onset in our previous study (Green & McDonald, 2008). In that paper the amplitude of the increase in theta-activity in occipital regions was highly correlated with subsequent attentional benefits on target discrimination accuracy, suggesting that it is an essential component of the voluntary control and deployment of attention. Given that theta (4–7 Hz) and alpha (8–14 Hz) are neighboring frequencies and some degree of frequency smearing will occur when the EEG data are transformed into the time–frequency domain, it is possible that the ipsilateral alpha modulations observed here are indexing the same neurocognitive process indexed by ipsilateral theta modulations in our previous study (Green & McDonald, 2008). The question remains as to why modulations of activity early on occur contralateral to the to-be-attended location, whereas modulations of activity immediately before the onset of a taskrelevant target occur ipsilateral to the to-be-attended location. It is possible that the decrease in alpha over contralateral scalp in the present study reflects activity of a neural generator in medial regions of ipsilateral visual cortex that is oriented such that it appears as a slightly contralateral effect at the scalp. If this is the case, the two phases of the alpha-band modulations could simply reflect the spread of activity from lower visual areas (e.g., V1/V2) to higher visual areas (e.g., V4, IT) over time in the ipsilateral cortex. However, it is also possible that the early activity is occurring in contralateral visual cortex but that pretarget enhancement of neuronal activity in the cortical regions that will process the target do not need to continue until the onset of the target. It has previously been hypothesized that preparatory changes in neural excitability continue only until they reach a preset maximum level and then remain stable at that level until the target occurs (Hopf & Mangun, 2000). It may be that enhancement of processing in contralateral visual cortex reaches its peak around 700 ms postcue, around the time that the LDAP dissipates and the focus of the alpha activity moves to ipsilateral cortex. Regardless of temporal predictability of the cues or the cue– target SOA, we consistently observed increased alpha-band activity

over scalp sites ipsilateral to the to-be-attended location in the few hundred milliseconds before the target appeared. This ipsilateral effect likely reflects a mechanism for active suppression of the tobe-ignored locations (Kelly et al., 2006; Rihs et al., 2007; Worden et al., 2000). Because the target was most likely to occur at the cued location and the discrimination task was perceptually demanding, the suppression of the uncued locations would be useful for funnelling processing resources to the cortical regions that were most likely to be stimulated by the target when it appeared. A number of studies have now provided evidence that topdown attentional control processes lead to modulation of activity in visual cortex prior to the appearance of an expected target. However, the precise nature of this pretarget biasing (i.e., enhancement of the to-be-attended location vs. suppression of the to-be-ignored location) and whether the ERP and oscillatory activities that have been independently linked to pretarget biasing show similar modulations have remained unclear. Our results suggest that anticipatory biasing of visual cortex in preparation for the target likely involves multiple overlapping processes, one that may enhance activity in cortical areas contralateral to the tobe-attended location and another that suppresses activity in cortex contralateral to the to-be-ignored location (see also McDonald & Green, 2008). Contralateral decreases (Thut et al., 2006; Yamagishi et al., 2005) and ipsilateral increases (Kelly et al., 2006; Rihs et al., 2007; Worden et al., 2000) in pretarget alpha-band activity have been reported separately in previous studies and have been interpreted as reflecting either facilitative or suppressive processes, respectively. Our results suggest that voluntarily attending to a spatial location in preparation for an upcoming target can involve both the enhancement of attended locations and suppression of unattended locations. These two processes likely complement each other, enabling the efficient processing of the target when it appears at the attended location. In addition, in the present study the same patterns of alpha activity were observed regardless of the temporal predictability of the attention-directing cues whereas the ERP modulations were only apparent early in the cue–target interval when the cues provided advance knowledge of when the target would appear. Examining alpha-band EEG modulations, or oscillatory activity in other frequency bands, enables the examination of neural processes that may not be precisely phase-locked in the EEG. Care must be taken to isolate attention-related changes in alphaband activity from stimulus-driven changes, however. In the present study, we isolated lateralized attention-related changes in alpha-band activity using a simple subtraction technique. Other, more elaborate methods have also been employed (Kelly et al., 2006). The results of the present study suggest that alpha-band EEG modulations may be a more reliable index of spatially specific anticipatory biasing in sensory cortex than the lateralized ERP responses.

REFERENCES Coull, J. T., & Nobre, A. C. (1998). Where and when to pay attention: The neural systems for directing attention to spatial locations and to time intervals as revealed by both PET and fMRI. Journal of Neuroscience, 18, 7426–7435. Engel, A. K., Fries, P., & Singer, W. (2001). Dynamic predictions: Oscillations and synchrony in top-down processing. Nature Reviews Neuroscience, 2, 704–716. Green, J. J., Conder, J. A., & McDonald, J. J. (2008). Lateralized frontal activity elicited by attention-directing visual and auditory cues. Psychophysiology, 45, 579–587.

Green, J. J., & McDonald, J. J. (2006). An event-related potential study of supramodal attentional control and crossmodal attention effects. Psychophysiology, 43, 161–171. Green, J. J., & McDonald, J. J. (2008). Electrical neuroimaging reveals timing of attentional control activity in human brain. PLoS Biology, 6, e81. Green, J. J., Teder-Sa¨leja¨rvi, W. A., & McDonald, J. J. (2005). Control mechanisms mediating shifts of attention in auditory and visual space: A spatio-temporal ERP analysis. Experimental Brain Research, 166, 358–369.

Temporal predictability and anticipatory biasing Grent-’t-Jong, T., & Woldorff, M. G. (2007). Timing and sequence of brain activity in top-down control of visual-spatial attention. PLoS Biology, 5, e12. Harter, M. R., Miller, S. L., Price, N. J., LaLonde, M. E., & Keyes, A. L. (1989). Neural processes involved in directing attention. Journal of Cognitive Neuroscience, 1, 223–237. Hillyard, S. A., & Anllo-Vento, L. (1998). Event-related brain potentials in the study of visual selective attention. Procedings of the National Academy of Sciences, USA, 95, 781–787. Hopf, J. M., & Mangun, G. R. (2000). Shifting visual attention in space: An electrophysiological analysis using high spatial resolution mapping. Clinical Neurophysiology, 111, 1241–1257. Jongen, E. M., Smulders, F. T., & Van der Heiden, J. S. (2007). Lateralized ERP components related to spatial orienting: Discriminating the direction of attention from processing sensory aspects of the cue. Psychophysiology, 44, 968–986. Kastner, S., & Ungerleider, L. G. (2000). Mechanisms of visual attention in the human cortex. Annual Review of Neuroscience, 23, 315–341. Kelly, S. P., Lalor, E. C., Reilly, R. B., & Foxe, J. J. (2006). Increases in alpha oscillatory power reflect an active retinotopic mechanism for distracter suppression during sustained visuospatial attention. Journal of Neurophysiology, 95, 3844–3851. LaBerge, D. (2005). Attentional processing. Cambridge, MA: Harvard University Press. Lakatos, P., Karmos, G., Mehta, A. D., Ulbert, I., & Schroeder, C. E. (2008). Entrainment of neuronal oscillations as a mechanism of attentional selection. Science, 320, 110–113. Large, E. W., & Jones, M. R. (1999). The dymanics of attending: How people track time-varying events. Psychological Review, 106, 119– 159. Luck, S. J., Hillyard, S. A., Mouloua, M., & Hawkins, H. L. (1996). Mechanisms of visual-spatial attention: Resource allocation or uncertainty reduction? Journal of Experimental Psychology: Human Perception and Performance, 22, 725–737. Luck, S. J., Hillyard, S. A., Mouloua, M., Woldorff, M. G., Clark, V. P., & Hawkins, H. L. (1994). Effects of spatial cueing on luminance detectability: Psychophysical and electrophysiological evidence for early selection. Journal of Experimental Psychology: Human Perception and Performance, 20, 887–904. Makeig, S., Westerfield, M., Jung, T.-P., Enghoff, S., Townsend, J., Courchesne, E., et al. (2002). Dynamic brain sources of visual evoked responses. Science, 295, 690–694. Mangun, G. R. (1994). Orienting attention in the visual fields: An electrophysiological analysis. In H.-J. Heinze, T. F. Munte, & G. R. Mangun (Eds.), Cognitive Electrophysiology (pp. 81–101). Boston: Birkhauser. Mangun, G. R., & Hillyard, S. A. (1991). Modulations of sensoryevoked brain potentials indicate changes in perceptual processing during visual-spatial priming. Journal of Experimental Psychology: Human Perception and Performance, 17, 1057–1074. Mazaheri, A., & Jensen, O. (2008). Asymmetric amplitude modulations of brain oscillations generate slow evoked responses. Journal of Neuroscience, 28, 7781–7787. Mazaheri, A., & Picton, T. W. (2005). EEG spectral dynamics during discrimination of auditory and visual targets. Cognitive Brain Research, 24, 81–96. McDonald, J. J., & Green, J. J. (2008). Isolating event-related potential components associated with voluntary control of visuo-spatial attention. Brain Research, 1227, 96–109.

1065 Miniussi, C., Wilding, E. L., Coull, J. T., & Nobre, A. C. (1999). Orienting attention in the time domain: Modulation of potentials. Brain, 122, 1507–1518. Nikulin, V. V., Linkenkaer-Hansen, K., Nolte, G., Lemm, S., Mu¨ller, K. R., Ilmoniemi, R. J., et al. (2007). Novel mechanism for evoked responses in the human brain. European Journal of Neuroscience, 25, 3146–3154. Nobre, A. C., Sebestyen, G. N., & Miniussi, C. (2000). The dynamics of shifting visuospatial attention revealed by event-related potentials. Neuropsychologia, 38, 964–974. Pfurtscheller, G., & Lopes da Silva, F. H. (1999). Event-related EEG/ MEG synchronization and desynchronization: Basic principles. Clinical Neurophysiology, 110, 1842–1857. Praamstra, P., Stegeman, D. F., Horstink, M. W. I. M., & Cools, A. R. (1996). Dipole source analysis suggests selective modulation of the supplementary motor area contribution to the readiness potential. Electroencephalography and Clinical Neurophysiology, 98, 468–477. Prinzmetal, W., McCool, C., & Park, S. (2005). Attention: Reaction time and accuracy reveal different mechanisms. Journal of Experimental Psychology General, 134, 73–92. Rihs, T. A., Michel, C. M., & Thut, G. (2007). Mechanisms of selective inhibition in visual spatial attention are indexed by a-band EEG synchronization. European Journal of Neuroscience, 25, 603–610. Salmelin, R., & Hari, R. (1994). Spatiotemporal characteristics of sensorimotor neuromagnetic rhythms related to thumb movement. Neuroscience, 60, 537–550. Talsma, D., Slagter, H. A., Nieuwenhuis, S., Hage, J., & Kok, A. (2005). The orienting of visuospatial attention: An event-related brain potential study. Cognitive Brain Research, 25, 117–129. Taylor, M. M., & Creelman, C. D. (1967). PEST: Efficient estimates on probability functions. Journal of the Acoustical Society of America, 41, 782–787. Thut, G., Nietzel, A., Brandt, S. A., & Pascual-Leone, A. (2006). Alphaband electroencaphalographic activity over occipital cortex indexes visuospatial attention bias and predicts visual target detection. Journal of Neuroscience, 26, 9494–9502. Van der Stigchel, S., Heslenfeld, D. J., & Theeuwes, J. (2006). An ERP study of preparatory and inhibitory mechanisms in a cued saccade task. Brain Research, 1105, 32–45. Van Velzen, J., & Eimer, M. (2003). Early posterior ERP components do not reflect the control of attentional shifts toward expected peripheral events. Psychophysiology, 40, 827–831. Van Velzen, J., Forster, B., & Eimer, M. (2002). Temporal dynamics of lateralized ERP components elicited during endogenous attentional shifts to relevant tactile events. Psychophysiology, 39, 874–878. Worden, M. S., Foxe, J. J., Wang, N., & Simpson, G. V. (2000). Anticipatory biasing of visuospatial attention indexed by retinotopically specific alpha-band electroencephalography increases over occipital cortex. Journal of Neuroscience, 20, RC63. Yamagishi, N., Goda, N., Callan, D. E., Anderson, S. J., & Kawato, M. (2005). Attentional shifts towards an expected visual target alter the level of alpha-band oscillatory activity in the human calcarine cortex. Cognitive Brain Research, 25, 799–809. Yamaguchi, S., Tsuchiya, H., & Kobayashi, S. (1994). Electroencephalographic activity associated with shifts of visuospatial attention. Brain, 117, 553–562. (Received June 19, 2009; Accepted December 15, 2009)

Psychophysiology, 47 (2010), 1066–1074. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01014.x

Effects of cycling exercise on vigor, fatigue, and electroencephalographic activity among young adults who report persistent fatigue

ROD K. DISHMAN,a NATHANIEL J. THOM,a,b TIMOTHY W. PUETZ,a PATRICK J. O’CONNOR,a and BRETT A. CLEMENTZb,c a

Department of Kinesiology, The University of Georgia, Athens, Georgia, USA BioImaging Research Center, Biomedical and Health Sciences Institute, The University of Georgia, Athens, Georgia, USA c Departments of Psychology and Neuroscience, The University of Georgia, Athens, Georgia, USA b

Abstract We previously reported that 6 weeks of exercise training had positive effects on feelings of vigor and fatigue among college students who reported persistent fatigue. Here we examined whether transient mood changes after single sessions of exercise would mimic those chronic effects and whether they would be related to changes in brain activity measured by electroencephalography (EEG). Feelings of vigor were higher after both low- and moderate-intensity exercise during Weeks 1, 3, and 6 compared to a control condition. Feelings of fatigue were lower after low-intensity exercise during Weeks 3 and 6. Posterior theta activity accounted for about half the changes in vigor. Studies that manipulate mood, EEG activity, or both during exercise are needed to determine whether EEG changes after exercise are causally linked with mood. Descriptors: Aerobic exercise, Mood, Theta activity

Those results suggested that chronic effects of exercise training on feelings of persistent fatigue or low energy might result from a gradual accumulation over several weeks of repeated, acute responses to each exercise session. To our knowledge, this possibility has not been investigated. We report here ancillary analyses of the aforementioned trial to examine the pattern of change in transient feelings of vigor and fatigue in response to single, repeated sessions of low-intensity or moderate-intensity exercise. We further examined whether the expected changes in fatigue and vigor after acute exercise would be related to changes in brain activity measured by electroencephalography (EEG). Although the chronic results of the trial were positive, they did not clarify whether the effects of physical activity on feelings of vigor and fatigue were plausibly explainable by neurobiological mechanisms or whether they might have been partly artifact, influenced by experimental demand (i.e., responses according to expectations of a trial’s goals) that bias participants’ responses to selfreport measures having transparent content (Morgan, 1997). Putative health benefits of physical activity are commonly known, and it is not feasible to blind participants to their participation in an exercise program. Hence, demonstration that participants’ transient responses to acute exercise are consistent across time and are related to biological markers plausibly linked to feelings of vigor or fatigue would bolster the evidence to support real psychological benefits of exercise for people who report persistent fatigue.

Feelings of fatigue or low energy are a burden on public health. Persistent fatigue has an estimated point prevalence of 20% among adults in the United States and worldwide (Wessely, Hotopf, & Sharpe, 1998). It is a common reason for doctor visits and is often treated inadequately by health care providers (Lange, Cook, & Natelson, 2005). The cumulative evidence suggests a role for physical activity as an adjuvant in the primary and secondary prevention of persistent fatigue. Population-based studies show that physical activity is associated with lower odds that people will report feelings of fatigue and low energy (Puetz, 2006). Clinical trials of chronic exercise among sedentary adults with fatigue-related medical conditions showed improvement in feelings of energy and fatigue that compares favorably with effects of cognitive–behavioral interventions or treatment with drugs (Puetz, Beasman, & O’Connor, 2006; Puetz, O’Connor, & Dishman, 2006). We recently reported positive effects on feelings of vigor and fatigue during 6 weeks of exercise training among inactive college students who complained of persistent fatigue (Puetz, Flowers, & O’Connor, 2008). Vigor was increased similarly regardless of exercise intensity, but fatigue was decreased only after low-intensity training. Results were most pronounced during the last week of training and were independent of changes in fitness. Address reprint requests to: Rod K. Dishman, Department of Kinesiology, The University of Georgia, Athens, GA 30602-6554, USA. E-mail: [email protected] 1066

Exercise, EEG, and vigor, and fatigue A substantial literature, including studies involving moderate submaximal and maximal exercise intensities (e.g., Hall, Ekkekakis, & Petruzzello, 2007; Petruzzello, Hall, & Ekkekakis, 2001; Petruzzello & Tate, 1997), is consistent with the theoretical view that hemispheric asymmetry in frontal alpha activity is both a neural mediator and a moderator of affective experience (Coan & Allen, 2004). The importance of EEG oscillatory activity in other frequencies and at other scalp locations for understanding feelings of energy or fatigue is uncertain. In a recent study of female college students without complaints of energy or fatigue problems, Woo, Kim, Kim, Petruzzello, and Hatfield (2009) reported that feelings of vigor were elevated above resting levels after 30 min of moderately intense treadmill running. That effect was attenuated after controlling for EEG activity in delta, theta, and alpha frequency bands measured at a right (F4), but not left (F3), anterior sensor location. The authors’ interpretation was that the increase in vigor was explainable by a reduction in behavioral withdrawal processes, consistent with the theoretical view that asymmetrical activation of right anterior brain regions is a predominant feature of the neural circuitry of unpleasant affective experience (Davidson, 2001). However, although activity in each frequency band was increased at the right anterior site and decreased at the left anterior site, only the reduction in theta activity at the left anterior site (F3) was statistically different after exercise compared to rest. An alternative theoretical position postulates that variations in posterior brain regions can be influential in the arousal dimension of affect (Heller, 1993). Although this theory has not yet been tested with people suffering from fatigue, predictions about increased left-hemisphere activity among people classified with high anxious apprehension relative to those classified with high anxious arousal have been supported (Heller, Nitschke, Etienne, & Miller, 1997). The potential relevance to the present work is that fatigue often accompanies anxious apprehension (Nitschke, Heller, Palmieri, & Miller, 1999), and other theorists have linked persistent fatigue to a dysregulation of arousal in the central nervous system (Pfaff, 2006). We previously reported on the cumulative evidence indicating that acute exercise typically increases EEG activity in delta, theta, alpha, and beta frequencies regardless of recording site (Crabbe & Dishman, 2004), and we hypothesized that the increased metabolic arousal of exercise might elicit diffuse increases in cortical EEG activity via thalamic-cortical or thalamic-reticular-cortical circuitry (Steriade, Gloor, Llina´s, Lopes de Silva, & Mesulam, 1990). In the study reported here, we were particularly interested in alpha and theta responses measured at both anterior and posterior sites because of three additional types of evidence. First, neurofeedback that elevates theta activity relative to alpha activity in the midline parietal region has been shown to increase feelings of energy (Raymond, Varney, Parkinson, & Gruzelier, 2005). Second, research on attention and executive cognitive functions such as working memory has revealed the importance of understanding dissociations between tonic and phasic changes in alpha and theta frequencies, especially in anterior–posterior neural networks (Klimesch, Freunberger, Sauseng, & Gruber, 2008). It was important to evaluate whether similar findings would be obtained for feelings of energy or fatigue after physical exertion. Third, viewing affective pictures that are rated as moderately or highly arousing has resulted in increased left frontal theta activity and larger bilateral increases at posterior parietal and occipital sites, with mixed and variable influences on alpha activity (Aftanas, Varlamov, Pavlov, Makhnev, & Reva, 2002).

1067 The purpose of this report, therefore, was to examine whether acute sessions of exercise repeated during 6 weeks of training would result in transient changes in feelings of vigor and fatigue (i.e., ‘‘right now’’ feelings assessed immediately before and after each exercise session) that would be similar to the chronic effects (i.e., more enduring feelings indexed using a ‘‘past week, including today’’ time frame) that we reported earlier (Puetz et al., 2008). Further, we examined whether the expected changes in vigor and fatigue after acute exercise would be related to concurrent changes in brain activity measured by EEG. Our goal was not to test neural correlates of emotional responding (e.g., Crabbe, Smith, & Dishman, 2007) or neural predictors of mood responses after exercise (e.g., Petruzzello et al., 2001) but, rather, to test whether acute exercise would modify feelings of energy and fatigue and whether those effects would be mediated by altered alpha and theta activity at anterior or posterior parietal and occipital sites. We also examined whether the effects would differ either according to exercise intensity or across the weeks of training.

Materials and Methods Participants Healthy male and female college students between the ages of 18 and 35 were recruited from the University of Georgia via electronic mail sent to several campus groups. Recruitment procedures and participant characteristics have been described in detail elsewhere (Puetz et al., 2008). Participants were paid $60 as an incentive to participate in one of the conditions for 6 weeks. Controls also were offered participation in a 6-week supervised exercise program that started after their participation in the control condition was completed. An a priori statistical power analysis (D’Amico, Neilands, & Zambarano, 2001) showed that a sample of 36 participants would provide a statistical power of ! 0.80 for the main and interaction effects of the study design assuming a two-tailed a value of .05, a high correlation across repeated measures (r 5 .80), and an expected Cohen’s d effect size of 0.60 for both EEG (Crabbe & Dishman, 2004) and feelings of fatigue and energy (Puetz et al., 2008). Exclusion criteria were as follows: (a) the absence of persistent fatigue (! 30 days), defined as a raw score of 17 or higher on the vitality scale of the SF-36 Health Survey; a raw score of 16 is approximately one-half standard deviation below U.S. population norms (Ware, 1993). Patients with chronic fatigue syndrome would have a longer chronicity of symptoms (! 6 months vs. ! 30 days) and SF-36 vitality scores likely to be 1–2 standard deviations lower than the participants in the present study (Hardt et al., 2001). (b) The presence of contraindications to maximal exercise based on professional guidelines (American College of Sports Medicine, 2006). (c) A physically active lifestyle, defined as a weekly energy expenditure greater than one-half standard deviation below college-aged norms as measured by the 7-Day Physical Activity Recall questionnaire (Blair et al., 1985; Dishman & Steinhardt, 1988). (d) The self-reported use of any antidepressant medication within the last month. (e) Scores at or above a cutoff suggesting any of the following DSM-IV psychiatric disorders measured by the Psychiatric Diagnostic Screening Questionnaire (Zimmerman & Mattia, 2001a, 2001b): generalized anxiety disorder, panic disorder, social anxiety disorder, major depressive disorder, and substance abuse.

1068 Design and Procedures Thirty-six participants were randomly assigned to one of three conditions: moderate intensity aerobic exercise, low intensity aerobic exercise, or no-treatment control. Blocked randomization, which assured 12 participants in each condition, was performed using Research Randomizer (www.randomizer.org). Following randomization, the week before the intervention began, mood, physical activity, and medication/nutritional supplement data were obtained. Week-to-week changes in the use of medications and nutritional supplements were monitored with a questionnaire created for use in this investigation. Participants in each condition subsequently visited the laboratory on 18 occasions over a 6-week period. Laboratory visits occurred three days per week. Each laboratory visit took place at approximately the same time each day. In addition to the 18 testing sessions, aerobic fitness testing occurred the week before the start and the week after the end of the 6-week study period. Aerobic Fitness Testing During aerobic fitness testing, all participants performed an incremental exercise test on an electronically braked, computerdriven cycle ergometer (Lode BV, Groningen, The Netherlands) in order to measure peak oxygen consumption (VO2peak). The purpose of the VO2peak test was to examine changes in fitness over the course of the study. Test results also were used to ensure that subsequent submaximal exercise training bouts were completed at an exercise intensity that was equal for each participant relative to his or her peak oxygen consumption and that the two exercise conditions differed so that participants were below ventilatory threshold during the low-intensity condition and above ventilatory threshold during the moderate-intensity condition. Participants were fitted to the cycle ergometer and provided with standardized, taperecorded instructions for providing overall ratings of perceived exertion using a 6–20 scale (Borg, 1998). After inserting a mouthpiece for collecting expired gases, the participants performed a 5-min warm up at 25 W. The initial work rate for the exercise test was 50 W and the work rate continuously increased at a rate of 24 W per minute until the participant reached volitional exhaustion (Storer, Davis, & Caiozzo, 1990). Pedal rate was maintained at 50–70 rpm throughout the test and was verified by an automated revolution counter. An integrated metabolic measurement system (Parvo Medics TrueOne 2400) was used to measure ventilation, oxygen consumption, and carbon dioxide production and calculate respiratory exchange ratio measures every 15 s. Heart rate was measured continuously using a Polar S120 heart rate monitor (Polar Electro Oy, Kempele, Finland). Heart rate, perceived exertion, and work rate were recorded during the last 10 s of every minute during the test of peak oxygen consumption. Peak oxygen consumption was defined as the highest VO2 value when two of three following criteria were satisfied: (a) respiratory exchange ratio ! 1.10, (b) heart rate within 10 beats per minute of age-predicted maximum (i.e., 220–age), or (c) rating of perceived exertion ! 18. All participants satisfied at least two of the three criteria. Ventilatory threshold (Beaver, Wasserman, & Whipp, 1986) was 60.8% (95% confidence interval [CI] 5 57.6%– 64.0%) of peak VO2 and did not differ between groups, F(2,26) 5 0.807, p 5 .457, Z2 5 .058, across time, F(1,26) 5 3.0, p 5 .095, Z2 5 .103, or between groups across time, F(2,26) 5 0.206, p 5 .815, Z2 5 .016. Differences between and within groups ranged from 5% to 8%, which did not exceed the standard error of the test.

R.K. Dishman et al. Intervention Conditions Moderate-intensity exercise condition. During the 6-week training intervention, the participants in the moderate-intensity aerobic exercise condition performed 20 min of exercise at 75% VO2peak on a cycle ergometer. Participants warmed up by cycling at 25 W for 5 min. Then participants cycled against a resistance that produced an intensity of 75% VO2peak for 20 min. During the 4th and 14th minute of exercise, VO2, heart rate, and perceived exertion data were collected. Based on the oxygen consumption measurements, an investigator made necessary adjustments to the work rate in order to maintain an intensity of 75% VO2peak throughout the remainder of the exercise bout. The typical adjustment was a reduction in pedal resistance, resulting in a reduction in associated power output to account for the slow component rise in VO2 that occurs during exercise at intensities exceeding ventilatory threshold. Intensity exceeded or equaled (1 participant) ventilatory threshold for all participants. After the 20-min exercise bout, participants cooled down by cycling at 25 W for 5 min. During the entire 30-min period, pedal rate was maintained at approximately 60 rpm. Participants completed all exercise sessions in a small room with the door closed in the presence of a single investigator who supervised the exercise. Social interactions were purposefully kept to a minimum in order to minimize possible effects of social interactions on mood and standardize the exercise setting. Low-intensity exercise condition. The participants assigned to the low-intensity exercise condition performed 20 min of exercise at 40% VO2peak on a cycle ergometer. Procedures and measures were identical to those previously described for the moderate intensity aerobic exercise condition except participants cycled at 40% VO2peak for 20 min. Intensity was below ventilatory threshold for all participants. No-treatment control condition. The participants assigned to the no-treatment control condition sat quietly on a cycle ergometer. Procedures and measures were identical to those previously described for the aerobic exercise conditions except participants sat on the cycle ergometer for 30 min instead of performing a 5-min warm up, 20-minute training session, and 5min cool down. This group also was asked to maintain their current physical activity level for the duration of the study. Adherence to exercise conditions. The overall adherence rate to the assigned condition for the three groups combined was 98.3% and did not differ between the groups, F(2,33) 5 2.03, p 5 .15, Z2 5 .11. The moderate intensity exercise, low intensity exercise, and no-treatment control groups completed an average of 97.2%, 98.3%, and 99.4% of the 18 prescribed sessions, respectively. Outcome Measures Mood States. Vigor and fatigue mood states were measured immediately before and 10 min after the acute exercise or no-treatment control conditions during the 1st (Week 1), 9th (Week 3) and 18th (Week 6) laboratory visit using the 30-item Profile of Mood States–Short Form (POMS-SF; McNair, Lorr, & Droppleman, 1992). Participants rated the intensity of 30 mood items on a 5-point Likert-type scale ranging from not at all to extremely. Participants were asked to respond as to how they felt ‘‘right now.’’

Exercise, EEG, and vigor, and fatigue EEG measurement. Four minutes of eyes-closed continuous EEG data were recorded prior to and after conditions of exercise or quiet rest according to standard procedures (Pivik et al., 1993; Pizzagalli, 2007). The timing of postcondition EEG assessments started ! 6 min after exercise stopped. EEG data were recorded vertex referenced using a 256-sensor Geodesic Sensor Net and NetAmps 200 amplifiers (Electrical Geodesics; EGI, Eugene, OR). The sensor net was adjusted until all pedestals were properly seated on the scalp (i.e., not sitting on thick mats of hair that could result in bridging between sensors; e.g., Greischar et al., 2004). Individual sensor impedances were adjusted until they were below 50 kO (Ferree, Luu, Russell, & Tucker, 2001). In addition, an electrolyte bridge test was conducted between all pairs of sensors prior to recording (Tenke & Kayser, 2001), and, if there was evidence of bridging, sensors were adjusted until bridging was no longer evident (this was rarely required). Data were sampled at 500 Hz with an analog filter bandpass of 0.1–200 Hz. Data Reduction EEG data screening. Data were preprocessed following recommendations (with minimal modification) made by Junghofer, Elbert, Tucker, and Rockstroh (2000). Raw data were visually inspected off-line for bad sensor recordings. Bad sensors were interpolated (no more than 5% of sensors for any subject) using a spherical spline interpolation method as implemented in BESA 6.0 (MEGIS Software, Gra¨felfing, Germany). Data were transformed to an average reference and digitally filtered from 0.5 to 58 Hz (12 db/octave rolloff, zero-phase) and notch-filtered at 60 Hz (4 Hz width). Eyeblink and cardiac artifact correction was achieved by using the Independent Component Analysis (ICA) toolbox in EEGLAB 4.515 (Delorme & Makeig, 2004) under Matlab (Version 7.0, MathWorks, Natick, MA). ICA allows artifact removal without spatially distorting the data by using higher-order statistics to produce temporally independent signals (Onton, Westerfield, Townsend, & Makeig, 2006). Independent components representing saccades, blinks, and heart rate artifact were removed according to published guidelines (Jung et al., 2000). Data were then transformed to an 81-sensor virtual surface Laplacian montage using BESA to improve spatial resolution (Nunez & Srinivasan, 2006). The surface Laplacian is the second spatial derivative of the scalp potential recorded at the sensors, which accentuates neural activities associated with radially oriented superficial cortical sources, the signals that are best measured with EEG (Nunez & Srinivasan, 2006). EEG data reduction. After artifact removal and before transformation to the frequency domain, the mean and linear trends were removed using Matlab. The data were then fast-Fourier transformed to absolute power (mV2/Hz) using a moving Hanning window (50% overlap) on 10-s segments of the data and averaged over each minute. Intraclass correlation coefficients (ICC-2) were high (mean ICC-2 5 .96, range: .722–.997) for each minute of data obtained prior to each of the three acute exercise or control conditions; therefore, each minute was averaged across the 4 min. Because of nonnormality, a natural log transform was performed, with FFT power expressed as logn(mV2/Hz) averaged across four frequency bands: theta (4–7 Hz), alpha (8– 13 Hz), low beta (13–20 Hz), and high beta (20–30 Hz). To evaluate which scalp locations were most active, grand average top-meridian plots were constructed in BESA for each frequency band. These plots were visually inspected for regions of highly

1069 active sensor groupings, and then mean power was calculated within these regions. Two sensor groupings (anterior and posterior) consistently emerged from the visual inspection of the top meridian plots and were therefore used for subsequent statistical analyses (see Figure 1). Channels included in each frequency band are as follows: anterior theta, low beta, and high beta 5 11 sites centered on Fpz and AFz; posterior theta and high beta 5 15 sites centered on Oz; posterior alpha 5 10 sites centered on P3 and P4; posterior low beta 5 26 sites centered on POz and Oz. ICC values within these regions were high (ICC-2 4.90). Data Analysis The primary dependent measures were mood ratings of vigor and fatigue and EEG power. Results did not differ according to hemisphere, so, as for mood ratings, EEG effects were tested for anterior and posterior regions using a series of 3 (group: moderate- and low-intensity exercise, quiet rest control) " 2 (time: pre- vs. post-condition) " 3 (week: Week 1, Week 3, Week 6) mixed model analyses of variance (RM-ANOVA) with time and week repeated (SPSS Windows Version 15, Chicago, IL). Sphericity adjustments were made using Huynh-Feldt e. After determining directional effects of change in each group, hypothesis testing was conducted by RM-ANCOVA using the precondition score as a time-varying covariate. Bonferroni corrected simple contrasts of adjusted means were used to decompose significant effects (two-tailed). Effect sizes associated with the F statistics are expressed as Z2. To examine whether changes in brain activity mediated changes in mood, EEG power was added as a time-varying covariate in each RM-ANCOVA for mood (Coan & Allen, 2004). This resulted in a 3 (group: moderate- and low-intensity exercise, quiet rest control) " 3 (week: Week 1, Week 3, Week 6) RMANCOVA controlling for precondition differences on the mood scores and for change scores (post–pre) of EEG activity. For this model, statistical findings were of interest if the addition of EEG as a covariate changed the mood model results from a statistically significant F value to an insignificant F, which would indicate that the EEG response accounted for some of the changes in mood (i.e., partially mediated the effect of experimental condition; MacKinnon, Fairchild, & Fritz, 2007). Results Mood For vigor, there was a Group " Time effect, F(2,30) 5 9.238, p 5 .001, Z2 5 .38, but not a Group " Time " Week effect, F(4,60) 5 1.97, p 5 .11, Z2 5 .12, indicating differences between the groups in change after each session (see Figure 2). Vigor was increased after low-intensity, F(1,9) 5 9.32, p 5 .014, Z2 5 .51, and moderate-intensity, F(1,11) 5 7.26, p 5 .021, Z2 5 .40, exercise and was reduced after the control condition, F(1,10) 5 8.88, p 5 .014, Z2 5 .47. The Group " Time effect was not changed after controlling for precondition scores, F(2,29) 5 8.77, p 5 .001, Z2 5 .38. Contrasts indicated that feelings of vigor were higher after both low-intensity exercise (p 5 .002) and moderate-intensity exercise (p 5 .032) compared to the control condition and that vigor was similar after moderate- and low-intensity exercise (p 5 .629; see Figure 2). For fatigue, there was a Group " Time " Week effect, F(4,60) 5 2.84, p 5 .032, Z2 5 .16, that was quadratic (p 5 .015; see Figure 3). Fatigue was reduced after low-intensity exercise during the last two sessions (Time " Week quadratic effect),

1070

R.K. Dishman et al.

Figure 1 (A) Top-down maps depicting the channels that were included in the analysis for each frequency band. (B) Top meridian plots of grand average FFT power (averaged across group, time, and week) for each frequency band across the 4-min data collection period. These plots indicate that there might be anterior/posterior mirroring of alpha power, although we used the surface Laplacian transformation, which should serve to minimize such effects (Nunez & Srinivasan, 2006).

F(1,9) 5 5.20, p 5 .048, Z2 5 .37, but was unchanged after moderate-intensity exercise (p values 4.145) and the control condition (p4.440). The Week ! Group effect was attenuated slightly after controlling for precondition scores, F(4,59) 5 2.40, p 5 .060, Z2 5 .14. Contrasts indicated that fatigue was lower after low-intensity exercise compared to both the moderateintensity and control conditions during Sessions 2 (p 5 .016) and 3 (p 5 .05) but not Session 1 (p 5 .395). EEG Power There were no significant main or interaction effects for EEG activity in the anterior channels (all p values " .51, all Z2s # .03), indicating that EEG activity was not different between groups or across time over anterior sites. However, sig-

8 week 1 week 3

6

week 6

Vigor

4

2

0

–2

low

moderate

control

Figure 2. POMS-SF vigor change scores (postminus precondition) $ SEM for moderate- and low-intensity cycling exercise and the control condition.

nificant Group ! Time effects were detected for EEG activity in posterior regions for all frequency bands (all F ratios " 6.00, all p values # .006, all Z2s " .27). There were no Group ! Week or Time ! Week effects (all Z2s # .05), indicating activity in all frequency bands was different between groups after each session, but not across weeks. Alpha activity was increased after low-intensity exercise, F(1,11) 5 6.23, p 5 .03, Z2 5 .36, but activity in other frequencies was unchanged after low-intensity exercise (all F ratios # 1.74, all p values " .214, all Z2s # .14). Conversely, activity in all frequencies was decreased after moderate-intensity exercise and the control condition (all F ratios " 9.44, all p values # .01, all Z2s " .36) with the exception that alpha activity was unchanged after the control condition, F(1,12) 5 2.35, p 5 .15, Z2 5 .18. Results were not different for group effects (all F ratios " 5.56, all p values # .008, all Z2s " .26) and Group ! Week effects (all F ratios # .89, all p values " .474, all Z2s # .052) after controlling for precondition values. Contrasts indicated that activity in all frequencies was higher after low-intensity exercise than after the control condition (all p values # .033). Alpha activity after low-intensity exercise was also higher than after moderate-intensity exercise (p 5 .047). Otherwise, activity did not differ between exercise conditions or between the moderate-intensity and control conditions (all p values " .314). Relationship between Changes in EEG Power and Mood The higher vigor observed after both low- and moderate-intensity exercise repeated during Weeks 1, 3, and 6, F(2,29) 5 8.77, p 5 .001, Z2 5 .38, was partially mediated by change in theta activity (4–7 Hz) in the posterior region (see Figure 4): adjusted model, F(2,29) 5 2.78, p 5 .08, Z2 5 .17. EEG activity in all other regions and frequency bands did not account for the change in feelings of vigor (all p values # .05 for all adjusted models). The lower fatigue observed after low-intensity exercise during Sessions 2 and 3, F(4,59) 5 2.40, p 5 .060, Z2 5 .14, was

Exercise, EEG, and vigor, and fatigue low

0

moderate

1071 control

Fatigue

–1

–2 week 1 week 3 –3

week 6

–4 Figure 3. POMS-SF fatigue change scores (post- minus precondition) $ SEM for low- and moderate-intensity cycling exercise and the control condition.

not explained by change in EEG activity (p values ! .065 and Z2 " .139 for all adjusted models). Discussion There were two important outcomes of the study. First, transient increases in feelings of vigor and decreases in feelings of fatigue experienced after each exercise session of low or moderate intensity paralleled results of our prior report that chronic exercise training led to improved feelings of vigor and fatigue experienced during the preceding week (Puetz et al., 2008). Reduced fatigue occurred only after low-intensity exercise and only during the second and third exercise sessions, suggesting that feelings of vigor are more sensitive to the effects of acute exercise than are feelings of fatigue among young people who report persistent fatigue. Second, and most importantly, half of the effect of exercise on feelings of vigor was mediated by changes in occipitalparietal brain activity in the theta frequency range, suggesting that exercise-induced modifications in mood are not simply a function of experimental demand characteristics. The significance of these findings for understanding the relationships between exercise and its effects on mood and specific cortical brain activity is discussed next.

week 1

0.4 Spectral power

week 3 0.2

week 6

0 –0.2 –0.4

low

moderate

control

Figure 4. Change scores (post- minus precondition) for posterior thetaband spectral density and low-intensity and moderate-intensity cycling exercise and the control condition.

People without clinical mood disorders rank exercise near the top among the behaviors they use to self-manage their moods (Thayer, Newman, & McClain, 1994). Even short bouts (i.e., 10 min) of walking for exercise in natural settings are accompanied by feelings of increased energy (Thayer, 1987). To our knowledge, however, this is the first experimental report indicating that acute exercise improves transient feelings of energy and fatigue in people who report persistent fatigue. Our participants’ scores on other scales of the POMS were in the normal range for college students and were not affected by exercise (data not shown). Hence, our results do not appear to represent an artifact of participant expectancies of general benefits of exercise for improving mood. Despite suggestions that the mood-enhancing effects of exercise may be slow to develop among persons less experienced with an exercise routine (Hoffman & Hoffman, 2008), the present findings offer further promise for exercise as a clinical tool to combat the deleterious public health costs of persistent fatigue and low energy (e.g., Lange et al., 2005; Puetz, 2006). Previous research used the POMS to assess the effects on mood of acute exercise in young women without persistent fatigue. During short bouts (# 8 min), higher-intensity exercise (cycling at 100 W) led to increases in subjective fatigue, whereas vigor was increased after low-intensity exercise (cycling at 25 W) among female medical students (Steptoe & Cox, 1988). In another sample of young women, fatigue was increased and vigor was decreased similarly after each of two maximal exercise tests separated by 1 month (Pronk, Crouse, & Rohack, 1995). Among middle-aged women and men without cardiopulmonary or metabolic diseases, 20 min of self-selected, moderate intensity treadmill exercise were followed by increased vigor and decreased fatigue among regular exercisers but not among those who had not exercised regularly during the preceding 6 months (Hoffman & Hoffman, 2008). Women and men who were receiving treatment for major depressive disorder reported reductions in fatigue as well as an increase in vigor after 30 min of moderately intense (60%–70% maximal heart rate) treadmill exercise (Bartholomew, Morrison, & Ciccolo, 2005). Only the changes in vigor were greater than those reported after a similar period of quiet rest. In another study of middle-aged women and men diagnosed with depression or anxiety disorders, 20 min of cycling exercise were followed by decreased feelings of fatigue (measured by a subjective exercise experiences scale) when the intensity was 50% of maximal heart rate reserve (# 58 W), but fatigue was unchanged after a self-selected intensity that was higher (# 66 W; Knapen et al., 2008). An interesting question for future research is whether transient feelings of vigor after early sessions of low- or moderate-intensity exercise might alter people’s awareness or interpretation of their fatigue after exertion and subsequently contribute to reductions in persistent feelings of fatigue or low energy. Most studies and literature reviews of the effect of acute exercise on brain electrocortical activity measured by EEG have emphasized, or have been restricted to, reports of changes in alpha frequency band activity (e.g., Kubitz & Pothakos, 1997). The common premise has been that increased alpha activity reflects a state of decreased cortical activation indicative of fatigue, relaxation, or decreased anxiety (Boutcher & Landers, 1988; Pineda & Adkisson, 1961). Alpha activity in response to exercise may be related to increased activity in somatosensory afferents (Youngstedt, Dishman, Cureton, & Peacock, 1993). This idea is consistent with evidence, including the data we report here, of increased alpha activity measured at central and temporal sites

1072 that overlay portions of the sensorimotor cortex that receive afferent feedback during exercise. Sensorimotor alpha has been called the mu rhythm (Kuhlman, 1978), and neuroimaging data indicate that it is directly associated with brain activation in the sensorimotor cortex (Ritter, Moosmann, & Villringer, 2009). Alpha activity in the present study was increased only after lowintensity exercise, and that change did not account for feelings of vigor or fatigue after exercise. Despite the historical focus on alpha activity, a systematic review of the exercise literature showed that acute exercise is accompanied by increases in EEG activity in all the measured frequency bands regardless of recording site (Crabbe & Dishman, 2004). A pattern of nonspecific changes in frequency bands associated with exercise would be consistent with research on the neuronal and neurophysiological architectures supporting oscillatory brain activity. For instance, alpha and theta frequencies traditionally are considered to recruit different neuronal pools when measured with scalp-level recordings (Pizzagalli, Oakes, & Davidson, 2003; Schu¨rmann & Basar, 1994), and to index unique cognitive processes or states of the nervous system (Pfurtscheller, 1992). However, recent work at the microscopic (cellular) level indicated a direct link between relative amounts of alpha/theta composition in the EEG and thalamocortical inputs in the same neuronal architecture (see Hughes & Crunelli, 2005). Thalamocortical neurons can generate alpha/theta oscillations after the activation of either metabotropic glutamate receptors (mGLUR) or muscarinic acetylcholine receptors (mAchR) (Hughes et al., 2008). In cat lateral geniculate nucleus preparations, pharmacological activation of mGLUR1a leads to increased alpha activity. Deactivation at the same synapses leads to a dose-response-related decrease in the idling frequency of these neurons and, therefore, a relative increase in amount of theta activity (Hughes et al., 2004). Activation of mGLURs is also associated with increased high threshold burst firing in the alpha/theta frequency range in the motor and somatosensory regions of the thalamus (Hughes et al., 2008). This work suggests that a decrease in thalamocortical input via mGLUR or mAchR activation may determine relative alpha/ theta concentrations in the spontaneous EEG. Also, activation of afferent inflow from carotid and cardiopulmonary baroreceptors into cardiovascular centers in the brain stem is alerting and increases indices of arousal, including hippocampal theta activity and cortical EEG activity in the insular cortex, a cardiovascular–limbic interface (Morgane, Galler, & Mokler, 2005; Oppenheimer, 1992; Spyer, 1989; Yu & Blessing, 2000). In humans, acute exercise increases cerebral blood flow in both the thalamus and insular cortex (Williamson, McColl, Mathews, Ginsburg, & Mitchell, 1999). Generalized effects of exercise on attention or arousal (Magnie et al., 2000) can plausibly influence brain electrocortical activity in several frequencies and loci, and it is plausible that brain cortical systems are altered generally in response to the increased metabolic arousal of physical exertion and regulation of physical fatigue by the brain (Nielsen, Hyldig, Bidstrup, GonzalezAlonso, & Christoffersen, 2001; Nybo & Nielsen, 2001; Nybo & Secher, 2004). Other research also implicates thalamocortical feedback in modification of oscillatory brain activities associated with exercise via sympathetically mediated changes in epinephrine and norepinephrine at higher intensities of exercise that elicit mechanoceptive and nociceptive signaling (Stock et al., 1996). The extent to which specifically cortico-thalamic input and associated neurocircuity are regulated or modified, or both, by varying intensities of exercise, therefore, would be a valuable avenue to pursue in subsequent research.

R.K. Dishman et al. Most prior studies that assessed nonspecific EEG activity associated with feelings of energy after acute exercise focused on frontal cortex activity (e.g., Petruzzello et al., 2001; Woo et al., 2009). Researchers recently have begun to examine more brain regions, reporting varying mood-related effects (e.g., increased calmness, reduced psychological strain) of different intensities of running on EEG alpha, beta, and gamma activity measured at anterior, temporal, parietal, and occipital sites among experienced runners (Schneider et al., 2009). Smit, Eling, Hopman, and Coenen (2005) reported that 40 min of moderately intense cycling exercise were followed by increased feelings of general activation and increased EEG power in low beta frequencies measured with eyes closed at the midline parietal region. There were no effects of exercise on theta or alpha frequencies at either frontal or parietal sites. Similarly, in the present report, there were no between-groups effects on EEG activity measured over frontal cortex. Rather, all effects were associated with activity over posterior brain regions. The present finding of associations between mood and posterior cortical activations is consistent with previous reports. For instance, Heller and colleagues have reported increases in self-ratings of arousal and cardiovascular and sympathetic nervous system activity related to electrocortical activation in parietal and temporal regions (Heller, 1990; Heller et al., 1997; Heller, Nitschke, & Miller, 1998). There are many possible reasons for the difference between previous studies and the present report on distributions of brain activations associated with exercise and changes of mood, one of which may be the study here of persons reporting persistent fatigue. Additionally, EEG theta power may vary in scalp topography as a function of caffeine sensitivity (Retey et al., 2006), which may be mediated by a polymorphism of the adenosine A2A receptor gene (Retey et al., 2007). Such findings are also of interest given numerous investigations reporting a link between theta activity and alertness or feelings of fatigue (e.g., Lal & Craig, 2001). Caffeine sensitivity and its relationship to exerciseinduced changes in EEG and mood have not been investigated. In our prior quantitative review of brain electrocortical activity during and after exercise, we analyzed a dozen potential moderators of the effects of acute exercise on EEG alpha activity. Several variables were found to significantly moderate the effect, including exercise duration, EEG recording latency from the end of exercise, time of day, whether the eyelids were opened or closed, the body position during EEG recording, whether the EEG reference scheme was reported, age, and whether the study did or did not include a control condition (Crabbe & Dishman, 2004). Differences between previous reports and the present study might also be attributed to EEG quantification schemes. Neural activity recorded at the scalp at any one sensor and at any single time point contains activity from multiple neural generators (Dien, Beal, & Berg, 2005). Locations of sources in the brain, therefore, cannot be directly inferred from the spatial distribution of EEG activity without first performing a source-space transformation. For the present report we used the surface Laplacian transformation, which has the desirable properties of requiring few underlying assumptions for its computation (Nunez & Srinivasan, 2006) but still being proportional to the dura potential (Srinivasan, Nunez, Tucker, Silberstein, & Cadusch, 1996). On the one hand, this transformation accentuates superficial neural activities that are oriented radially with respect to the sensors, which are the signals best measured with EEG. On the other hand, the surface Laplacian attenuates deep, nonfocal

Exercise, EEG, and vigor, and fatigue

1073

sources that can be either the cause of diffuse noise when examining cortical activity or be related to true brain activations (Nunez & Srinivasan, 2006). The quantification scheme used here, therefore, was conservative with respect to the signals measured, which may have resulted in the attenuation of real neural activations that are difficult to discern from nonspecific neural or

system noise, or both. The best way to resolve this issue in future investigations is by using multiple imaging modalities to collect brain activity data from the same participants (e.g., Dale & Halgren, 2001). Such multimodal neuroimaging work has the potential to more fully elucidate how brain activity mediates the relationship between changes in mood and exercise.

REFERENCES Aftanas, L. I., Varlamov, A. A., Pavlov, S. V., Makhnev, V. P., & Reva, N. V. (2002). Time-dependent cortical asymmetries induced by emotional arousal: EEG analysis of event-related synchronization and desynchronization in individually defined frequency bands. International Journal of Psychophysiology, 44, 67–82. American College of Sports Medicine. (2006). ACSM’s guidelines for exercise testing and prescription (7th ed). Philadelphia: Lippincot Williams & Wilkins. Bartholomew, J. B., Morrison, D., & Ciccolo, J. T. (2005). Effects of acute exercise on mood and well-being in patients with major depressive disorder. Medicine and Science in Sports and Exercise, 37, 2032– 3037. Beaver, W. L., Wasserman, K., & Whipp, B. J. (1986). A new method for detecting anaerobic threshold by gas exchange. Journal of Applied Physiology, 60, 2020–2027. Blair, S. N., Haskell, W. L., Ho, P., Paffenbarger, R. S. Jr., Vranizan, K. M., Farquhar, J. W., et al. (1985). Assessment of habitual physical activity by a seven-day recall in a community survey and controlled experiments. American Journal of Epidemiology, 122, 794–804. Borg, G. (1998). Borg’s perceived exertion and pain scales. Champaign: Human Kinetics. Boutcher, S. H., & Landers, D. M. (1988). The effects of vigorous exercise on anxiety, heart rate, and alpha activity of runners and nonrunners. Psychophysiology, 25, 696–702. Coan, J. A., & Allen, J. J. (2004). Frontal EEG asymmetry as a moderator and mediator of emotion. Biological Psychology, 67, 7–49. Crabbe, J. B., & Dishman, R. K. (2004). Brain electrocortical activity during and after exercise: A quantitative synthesis. Psychophysiology, 41, 563–574. Crabbe, J. B., Smith, J. C., & Dishman, R. K. (2007). Emotional & electroencephalographic responses during affective picture viewing after exercise. Physiology & Behavior, 90, 394–404. Dale, A. M., & Halgren, E. (2001). Spatiotemporal mapping of brain activity by integration of multiple imaging modalities. Current Opinion in Neurobiology, 11, 202–208. D’Amico, E. J., Neilands, T. B., & Zambarano, R. (2001). Power analysis for multivariate and repeated measures designs: A flexible approach using the SPSS MANOVA procedure. Behavior Research Methods, 33, 479–484. Davidson, R. J. (2001). Toward a biology of personality and emotion. Annals of the New York Academy of Sciences, 935, 191–207. Delorme, A., & Makeig, S. (2004). Eeglab: An open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. Journal of Neuroscience Methods, 134, 9–21. Dien, J., Beal, D. J., & Berg, P. (2005). Optimizing principal components analysis of event-related potentials: Matrix type, factor loading weighting, extraction, and rotations. Clinical Neurophysiology, 116, 1808–1825. Dishman, R., & Steinhardt, M. (1988). Reliability and concurrent validity for a 7-d re-call of physical activity in college students. Medicine and Science in Sports and Exercise, 20, 14–25. Ferree, T. C., Luu, P., Russell, G. S., & Tucker, D. M. (2001). Scalp electrode impedance, infection risk, and EEG data quality. Clinical Neurophysiology, 112, 536–544. Greischar, L. L., Burghy, C. A., van Reekum, C. M., Jackson, D. C., Pizzagalli, D. A., Mueller, C., et al. (2004). Effects of electrode density and electrolyte spreading in dense array electroencephalographic recording. Clinical Neurophysiology, 115, 710–720. Hall, E. E., Ekkekakis, P., & Petruzzello, S. J. (2007). Regional brain activity and strenuous exercise: Predicting affective responses using EEG asymmetry. Biological Psychology, 75, 194–200. Hardt, J., Buchwald, D., Wilks, D., Sharpe, M., Nix, W. A., & Egle, U. T. (2001). Health-related quality of life in patients with chronic

fatigue syndrome: An international study. Journal of Psychosomatic Research, 51, 431–434. Heller, W. (1990). The neuropsychology of emotion: Developmental patterns and implications for psychopathology (Vol. 1). Hillsdale, NJ: Lawrence Erlbaum Associates. Heller, W. (1993). Neuropsychological mechanisms of individual differences in emotion, personality, and arousal. Neuropsychology, 7, 476– 489. Heller, W., Nitschke, J. B., Etienne, M. A., & Miller, G. A. (1997). Patterns of regional brain activity differentiate types of anxiety. Journal of Abnormal Psychology, 106, 376–385. Heller, W., Nitschke, J. B., & Miller, G. A. (1998). Lateralization in emotion and emotional disorders. Current Directions in Psychological Science, 7, 26–32. Hoffman, M. D., & Hoffman, D. R. (2008). Exercisers achieve greater acute exercise-induced mood enhancement than nonexercisers. Archives of Physical Medicine and Rehabilitation, 89, 358–363. Hughes, S. W., & Crunelli, V. (2005). Thalamic mechanisms of EEG alpha rhythms and their pathological implications. The Neuroscientist, 11, 357–372. Hughes, S. W., Errington, A., Lorincz, M. L., Ke´kesi, K. A., Juha´sz, G., Orba´n, G., et al. (2008). Novel modes of rhythmic burst firing at cognitively-relevant frequencies in thalamocortical neurons. Brain Research, 1235, 12–20. Hughes, S. W., Lo¨rincz, M., Cope, D. W., Blethyn, K. L., Ke´kesi, K. A., Parri, H. R., et al. (2004). Synchronized oscillations at alpha and theta frequencies in the lateral geniculate nucleus. Neuron, 42, 253–268. Jung, T. P., Makeig, S., Humphries, C., Lee, T. W., McKeown, M. J., Iragui, V., et al. (2000). Removing electroencephalographic artifacts by blind source separation. Psychophysiology, 37, 163–178. Junghofer, M., Elbert, T., Tucker, D. M., & Rockstroh, B. (2000). Statistical control of artifacts in dense array EEG/MEG studies. Psychophysiology, 37, 523–532. Klimesch, W., Freunberger, R., Sauseng, P., & Gruber, W. (2008). A short review of slow phase synchronization and memory: Evidence for control processes in different memory systems? Brain Research, 1235, 31–44. Knapen, J., Sommerijns, E., Vancampfort, D., Sienaert, P., Pieters, G., Haake, P., et al. (2008). State anxiety and subjective well-being responses to acute bouts of aerobic exercise in patients with depressive and anxiety disorders. British Medical Journal, 43, 756–759. Kubitz, K. A., & Pothakos, K. (1997). Does aerobic exercise decrease brain activation? Journal of Sport and Exercise Psychology, 19, 291– 301. Kuhlman, W. N. (1978). Functional topography of the human mu rhythm. Electroencephalography and Clinical Neurophysiology, 44, 83–93. Lal, S. K. L., & Craig, A. (2001). A critical review of the psychophysiology of driver fatigue. Biological Psychology, 55, 173–194. Lange, G., Cook, D. B., & Natelson, B. H. (2005). Rehabilitation and treatment of fatigue. In J. DeLuca (Ed.), Fatigue as a window to the brain (pp. 301–316). Cambridge, MA: MIT Press. MacKinnon, D. P., Fairchild, A. J., & Fritz, M. S. (2007). Mediation analysis. Annual Review of Psychology, 58, 593–614. Magnie, M. N., Bermon, S., Martin, F., Madany-Lounis, M., Suisse, G., Muhammad, W., et al. (2000). P300, n400, aerobic fitness, and maximal aerobic exercise. Psychophysiology, 37, 369–377. McNair, D. M., Lorr, M., & Droppleman, L. F. (1992). Edits manual for the profile of mood states. San Diego, CA: Edits/Educational and Industrial Testing Service. Morgan, W. P. (1997). Methodological considerations. In W. P. Morgan (Ed.), Physical activity and mental health (pp. 3–32). Washington, DC: Taylor & Francis.

1074 Morgane, P. J., Galler, J. R., & Mokler, D. J. (2005). A review of systems and networks of the limbic forebrain/limbic midbrain. Progress in Neurobiology, 75, 143–160. Nielsen, B., Hyldig, T., Bidstrup, F., Gonzalez-Alonso, J., & Christoffersen, G. R. J. (2001). Brain activity and fatigue during prolonged exercise in the heat. Pflu¨gers Archives: European Journal of Physiology, 442, 41–48. Nitschke, J. B., Heller, W., Palmieri, P. A., & Miller, G. A. (1999). Contrasting patterns of brain activity in anxious apprehension and anxious arousal. Psychophysiology, 36, 628–637. Nunez, P. L., & Srinivasan, R. (2006). Electric fields of the brain: The neurophysics of EEG. New York: Oxford University Press. Nybo, L., & Nielsen, B. (2001). Perceived exertion is associated with an altered brain activity during exercise with progressive hyperthermia. Journal of Applied Physiology, 91, 2017–2023. Nybo, L., & Secher, N. H. (2004). Cerebral perturbations provoked by prolonged exercise. Progress in Neurobiology, 72, 223–261. Onton, J., Westerfield, M., Townsend, J., & Makeig, S. (2006). Imaging human EEG dynamics using independent component analysis. Neuroscience and Biobehavioral Reviews, 30, 808–822. Oppenheimer, S. M. (1992). Cardiovascular effects of human insular cortex stimulation. Neurology, 42, 1727–1732. Petruzzello, S. J., Hall, E. E., & Ekkekakis, P. (2001). Regional brain activation as a biological marker of affective responsivity to acute exercise: Influence of fitness. Psychophysiology, 38, 99–106. Petruzzello, S. J., & Tate, A. K. (1997). Brain activation, affect, and aerobic exercise: An examination of both state-independent and statedependent relationships. Psychophysiology, 34, 527–533. Pfaff, D. (2006). Brain arousal and information theory: Neural and genetic mechanisms. Cambridge, MA: Harvard University Press. Pfurtscheller, G. (1992). Event-related synchronization (ERS): An electrophysiological correlate of cortical areas at rest. Electroencephalography and Clinical Neurophysiology, 83, 62–69. Pineda, A., & Adkisson, M. A. (1961). Electroencephalographic studies in physical fatigue. Texas Reports on Biology and Medicine, 19, 332– 342. Pivik, R. T., Broughton, R. J., Coppola, R., Davidson, R. J., Fox, N., & Nuwer, M. R. (1993). Guidelines for the recording and quantitative analysis of electroencephalographic activity in research contexts. Psychophysiology, 30, 547–558. Pizzagalli, D. A. (2007). Electroencephalography and high-density electrophysiological source localization. In J. T. Cacioppo, L. G. Tassinary, & G. G. Berntson (Eds). Handbook of psychophysiology (3rd ed, pp. 56–84). Cambridge, UK: Cambridge University Press. Pizzagalli, D. A., Oakes, T. R., & Davidson, R. J. (2003). Coupling of theta activity and glucose metabolism in the human rostral anterior cingulate cortex: An EEG/PET study of normal and depressed subjects. Psychophysiology, 40, 939–949. Pronk, N. P., Crouse, S. F., & Rohack, J. J. (1995). Maximal exercise and acute mood response in women. Physiology and Behavior, 57, 1–4. Puetz, T. W. (2006). Physical activity and feelings of energy and fatigue: Epidemiological evidence. Sports Medicine, 36, 767–780. Puetz, T. W., Beasman, K. M., & O’Connor, P. J. (2006). The effect of cardiac rehabilitation exercise programs on feelings of energy and fatigue: A meta-analysis of research from 1945 to 2005. European Journal of Cardiovascular Prevention & Rehabilitation, 13, 886–893. Puetz, T. W., Flowers, S. S., & O’Connor, P. J. (2008). A randomized controlled trial of the effect of aerobic exercise training on feelings of energy and fatigue in sedentary young adults with persistent fatigue. Psychotherapy and Psychosomatics, 77, 167–174. Puetz, T. W., O’Connor, P. J., & Dishman, R. K. (2006). Effects of chronic exercise on feelings of energy and fatigue: A quantitative synthesis. Psychological Bulletin, 132, 866–876. Raymond, J., Varney, C., Parkinson, L. A., & Gruzelier, J. H. (2005). The effects of alpha/theta neurofeedback on personality and mood. Cognitive Brain Research, 23, 287–292. Retey, J. V., Adam, M., Gottselig, J. M., Khatami, R., Durr, R., Achermann, P., et al. (2006). Adenosinergic mechanisms contribute to individual differences in sleep deprivation-induced changes in neurobehavioral function and brain rhythmic activity. Journal of Neuroscience, 26, 10472–10479. Retey, J. V., Adam, M., Khatami, R., Luhmann, U. F. O., Jung, H. H., Berger, W., et al. (2007). A genetic variation in the adenosine a2a

R.K. Dishman et al. receptor gene (adora2a) contributes to individual sensitivity to caffeine effects on sleep. Clinical Pharmacology and Therapeutics, 81, 692–698. Ritter, P., Moosmann, M., & Villringer, A. (2009). Rolandic alpha and beta EEG rhythms’ strengths are inversely related to fmri-bold signal in primary somatosensory and motor cortex. Human Brain Mapping, 30, 1168–1187. Schneider, S., Askew, C. D., Diehl, J., Mierau, A., Kleinert, J., Abel, T., et al. (2009). EEG activity and mood in health orientated runners after different exercise intensities. Physiology & Behavior, 96, 709– 716. Schu¨rmann, M., & Basar, E. (1994). Topography of alpha and theta oscillatory responses upon auditory and visual stimuli in humans. Biological Cybernetics, 72, 161–174. Smit, A. S., Eling, P. A., Hopman, M. T., & Coenen, A. M. (2005). Mental and physical effort affect vigilance differently. International Journal of Psychophysiology, 57, 211–217. Spyer, K. M. (1989). Neural mechanisms involved in cardiovascular control during affective behaviour. Trends in Neuroscience, 12, 506– 513. Srinivasan, R., Nunez, P. L., Tucker, D. M., Silberstein, R. B., & Cadusch, P. J. (1996). Spatial sampling and filtering of EEG with spline Laplacians to estimate cortical potentials. Brain Topography, 8, 355– 366. Steptoe, A., & Cox, S. (1988). Acute effects of aerobic exercise on mood. Health Psychology, 7, 329–340. Steriade, M., Gloor, P., Llina´s, R. R., Lopes de Silva, F. H., & Mesulam, M. M. (1990). Report of IFCN Committee on Basic Mechanisms. Basic mechanisms of cerebral rhythmic activities. Electroencephalography and Clinical Neurophysiology, 76, 481–508. Stock, C., Baum, M., Rosskopf, P., Schober, F., Weiss, M., & Liesen, H. (1996). Electroencephalogram activity, catecholamines, and lymphocyte subpopulations after resistance exercise and during regeneration. European Journal of Applied Physiology, 72, 235–241. Storer, T. W., Davis, J. A., & Caiozzo, V. J. (1990). Accurate prediction of VO2max in cycle ergometry. Medicine and Science in Sports and Exercise, 22, 704–712. Tenke, C. E., & Kayser, J. (2001). A convenient method for detecting electrolyte bridges in multichannel electroencephalogram and eventrelated potential recordings. Clinical Neurophysiology, 112, 545–550. Thayer, R. E. (1987). Energy, tiredness, and tension effects of a sugar snack versus moderate exercise. Journal of Personality and Social Psychology, 52, 119–125. Thayer, R. E., Newman, J. R., & McClain, T. M. (1994). Self-regulation of mood: Strategies for changing a bad mood, raising energy, and reducing tension. Journal of Personality and Social Psychology, 67, 910–925. Ware, J. E. (1993). Sf-36 health survey: Manual and interpretation guide. Boston, MA: The Health Institute, New England Medical Center. Wessely, S., Hotopf, M., & Sharpe, M. (1998). Chronic fatigue and its syndromes. Oxford: Oxford University Press. Williamson, J. W., McColl, R., Mathews, D., Ginsburg, M., & Mitchell, J. H. (1999). Activation of the insular cortex is affected by the intensity of exercise. Journal of Applied Physiology, 87, 1213–1219. Woo, M., Kim, S., Kim, J., Petruzzello, S. J., & Hatfield, B. D. (2009). Examining the exercise-affect dose–response relationship: Does duration influence frontal EEG asymmetry? International Journal of Psychophysiology, 72, 166–172. Youngstedt, S. D., Dishman, R. K., Cureton, K. J., & Peacock, L. J. (1993). Does body temperature mediate anxiolytic effects of acute exercise? Journal of Applied Physiology, 74, 825–831. Yu, Y. H., & Blessing, W. W. (2000). Carotid and cardiopulmonary chemoreceptor activity increases hippocampal theta rhythm in conscious rabbits. American Journal of Physiology, 278, R973–R979. Zimmerman, M., & Mattia, J. I. (2001a). The psychiatric diagnostic screening questionnaire: Development, reliability and validity. Comprehensive Psychiatry, 42, 175–189. Zimmerman, M., & Mattia, J. I. (2001b). A self-report scale to help make psychiatric diagnoses the psychiatric diagnostic screening questionnaire. Archives of General Psychiatry, 58, 787–794. (Received June 26, 2009; Accepted November 12, 2009)

Psychophysiology, 47 (2010), 1075–1086. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01013.x

Neuronal generator patterns of olfactory event-related brain potentials in schizophrenia

JU¨RGEN KAYSER,a,b CRAIG E. TENKE,a,b DOLORES MALASPINA,c CHRISTOPHER J. KROPPMANN,a JENNIFER D. SCHALLER,a ANDREW DEPTULA,a NATHAN A. GATES,a JILL M. HARKAVY-FRIEDMAN,b ROBERTO GIL,b,d and GERARD E. BRUDERa,b a

Division of Cognitive Neuroscience, New York State Psychiatric Institute, New York, New York, USA Department of Psychiatry, Columbia University College of Physicians & Surgeons, New York, New York, USA c Department of Psychiatry, New York University School of Medicine, New York, New York, USA d Division of Translational Imaging, New York State Psychiatric Institute, New York, New York, USA b

Abstract To better characterize neurophysiologic processes underlying olfactory dysfunction in schizophrenia, nose–referenced 30–channel electroencephalogram was recorded from 32 patients and 35 healthy adults (18 and 18 male) during detection of hydrogen sulfide (constant-flow olfactometer, 200 ms unirhinal exposure). Event-related potentials (ERPs) were transformed to reference–free current source density (CSD) waveforms and analyzed by unrestricted Varimax–PCA. Participants indicated when they perceived a high (10 ppm) or low (50% dilution) odor concentration. Patients and controls did not differ in detection of high (23% misses) and low (43%) intensities and also had similar olfactory ERP waveforms. CSDs showed a greater bilateral frontotemporal N1 sink (305 ms) and mid-parietal P2 source (630 ms) for high than low intensities. N1 sink and P2 source were markedly reduced in patients for high intensity stimuli, providing further neurophysiological evidence of olfactory dysfunction in schizophrenia. Descriptors: Olfaction, Schizophrenia, Event-related potential, ERP, Current source density, CSD, Principal components analysis, PCA, Surface Laplacian

researchers have begun to advance the knowledge in basic mechanisms of olfactory perception (Lorig, 2000). The clinical significance of OERPs is evident in that stimulation with vanillin or hydrogen sulfide (H2S) yields no OERP components in anosmic patients (Kobal & Hummel, 1998), and OERPs are closely associated with odor thresholds, odor discrimination, and odor identification (Lo¨tsch & Hummel, 2006). Although there has been some disagreement about the naming of peaks in the OERP waveforms, when using a combined lateral-inferior electroencephalogram (EEG) recording reference (i.e., linked ears or mastoids), healthy adults typically show as the first distinctive deflection a negative peak at vertex between 300 and 500 ms, labeled N1 (e.g., Rombaux et al., 2006). This is followed by one or more positive deflections (e.g., P2, P3) peaking between 500 and 1500 ms (e.g., Pause, Sojka, Krauel, & Ferstl, 1996; Turetsky et al., 2003). Although significantly delayed compared to other modalities (approximate N1 peak latencies range between 100 and 200 ms for auditory or visual stimuli) because of a longer transduction time at the olfactory receptor level (e.g., Rombaux et al., 2006), the N1 component may have similar modality-specific properties (Pause & Krauel, 2000; Olofsson, Ericsson, & Nordin, 2008). The olfactory pathway, however, unlike all other sensory systems, does not include a thalamic relay, and it is unknown to what extent different anatomical structures and cortical regions within the olfactory system (e.g., olfactory bulb, orbital prefrontal cortex; cf. Martzke, Kopala, & Good, 1997)

The study of olfactory event-related potentials (OERP) requires a rapid onset of odor concentration (less than 50 ms rise time to 70% of maximum concentration; cf. Evans, Kobal, Lorig, & Prah, 1993; Rombaux, Mouraux, Bertrand, Guerit, & Hummel, 2006) and recording of olfactory responses that avoid concomitant trigeminal nerve stimulation (Lorig, 2000) and, depending on the research objective, potential confounds associated with active inhalation (Sobel et al., 1998; however, see Lorig, Matia, Peszka, & Bryant, 1996, for a balanced discussion on the relative merits of active vs. passive breathing techniques). This became possible through the development of an olfactometer capable of producing a rapid pulse of odor in a constant air stream (Kobal, 1982, 2003; cf. Rombaux et al., 2006). Using an olfactometer, The National Institute of Mental Health (NIMH) supported this research through grants MH066428, MH066597, and MH082393. We are grateful to Bruce Turetsky at the University of Pennsylvania for his help when we initially established an olfactory laboratory at New York State Psychiatric Institute. We thank Charles L. Brown, III, for developing a fine software for waveform plotting. Thanks are also due to Raymond Goetz and Deborah Goetz for their help with this project. We appreciate several constructive comments received during the review process by Tyler Lorig, Dean Salisbury, and two anonymous referees. Address reprint requests to: Ju¨rgen Kayser, New York State Psychiatric Institute, Division of Cognitive Neuroscience, Unit 50, 1051 Riverside Drive, New York, NY 10032, USA. E-mail: [email protected]. columbia.edu 1075

1076 contribute to early olfactory components. Nevertheless, both N1 and P2 vary with external odor characteristics; for example, their amplitudes increase with greater odor concentration (e.g., Tateyama, Hummel, Roscher, Post, & Kobal, 1998; Turetsky et al., 2003). In contrast, the P3 component, as in other stimulus modalities, appears to change as a function of subjective significance, stimulus probability, and emotional valence of odors (Pause et al., 1996, 2003; see also Laudien, Kuster, Sojka, Ferstl, & Pause, 2006). However, a direct comparison of chemosensory, auditory, and visual N1, P2, and P3 peak deflections at midline sites (Fz, Cz, Pz) revealed a clustering of chemosensory P2 and P3, which were, in turn, clustered with auditory and visual P3, suggesting that olfactory P2 may have functional properties typically attributed to P3 in other sensory systems (Olofsson et al., 2008). OERP components in healthy adults vary with age and gender, with younger adults or women having generally greater amplitude and shorter latency when compared to older adults or men (e.g., Covington, Geisler, Polich, & Murphy, 1999; Morgan, Geisler, Covington, Polich, & Murphy, 1999; Murphy et al., 2000; Olofsson & Nordin, 2004; Stuck et al., 2006). Very little is known about the current generators underlying the olfactory ERP components. Kettenmann, Hummel, Stefan, and Koba (1997), using magnetoencephalographic rather than EEG recordings, localized equivalent current dipoles or sources corresponding to P1, N1, and P2 components between the superior temporal plane, the parainsular cortex, central parts of the insular, and the superior temporal sulcus. Furthermore, Daniels et al. (2001) found that patients with right-sided lesions, primarily affecting the temporal or parietal lobe, showed deficits in odor discrimination and decreased amplitudes of P2 and P3 at parietal scalp locations, independent of stimulation side (left or right nostril).

Olfactory Deficits in Schizophrenia Evidence for olfactory dysfunction in schizophrenia has been reported in multiple studies using psychophysical measures of odor identification and detection thresholds. Studies have consistently found reduced ability to name or identify odors in schizophrenic patients compared to healthy controls, typically yielding large effect sizes (for a review, see Moberg et al., 1999). Findings for odor detection thresholds have been more mixed, with some studies reporting poorer odor thresholds in schizophrenia (Moberg et al., 1999) and others reporting normal or even superior olfactory acuity (Martzke et al., 1997; Moberg et al., 2006). Measuring unirhinal thresholds to n-butanol in 17 unmedicated patients and 17 well-matched healthy controls, Purdon and Flor-Henry (2000) found asymmetric thresholds in schizophrenia. Whereas controls revealed no nostril differences, patients had a greater deficit for the left compared to the right nostril, implicating a primarily left-lateralized impairment, given the predominantly ipsilateral afferent projections from the olfactory bulb to piriform cortex within the medial temporal lobe (e.g., Martzke et al., 1997; Moberg et al., 1999). Interestingly, this threshold asymmetry was reversed in another 10 patients after they received neuroleptic treatment, mostly because of left nostril improvements, which suggested that the effects of antipsychotic medication may differentially affect the two hemispheres (Purdon & Flor-Henry, 2000). Few studies, however, have been directed at the neurophysiologic processes underlying olfactory dysfunction in schizophrenia. In the first electrophysiologic study, Turetsky et al.

J. Kayser et al. (2003) measured OERPs in 21 patients with schizophrenia and 20 healthy controls to three concentrations of H2S. Patients and controls did not differ significantly in ratings of the perceived intensity of the odors, but, nonetheless, patients had reduced N1 and P2 amplitudes, with the largest difference for the strongest odor intensity. Turetsky, Kohler, Gur, and Moberg (2008) also found similar reductions of N1 and P2 amplitude in first degree relatives of patients with schizophrenia, suggesting that this represents a vulnerability marker for this disorder. Using odorants of different hedonic value (i.e., rose-like phenethyl alcohol and rotten butter-like isobutyraldehyde), Pause, Hellmann, Goder, Aldenhoff, and Ferstl (2008) reported shorter peak latencies across several ERP components in nine schizophrenic compared to nine depressed and nine healthy men, but these effects were evidently most robust for N1 during the presentation of negative odors. Unfortunately, no ERP waveforms were included in this report, making it difficult to evaluate the exact meaning of these findings or to relate them to other studies. Methodological Issues in Olfactory ERP Research Following early recommendations (Evans et al., 1993), most OERP studies have relied on peak and latency measures of ‘‘prominent’’ deflections in selected OERP waveforms, mostly at vertex (Cz) or neighboring midline (Fz, Pz) or lateral sites (C3/4) and usually referenced to linked ear lobes or linked mastoids (e.g., Kru¨ger, Frasnelli, Bra¨unig, & Hummel, 2006; Lundstro¨m, Seven, Olsson, Schaal, & Hummel, 2006; Murphy et al., 2000; Pause et al., 2003). The use of multichannel EEG montages has largely been limited to mapping ERP or CSD1 topographies (Laudien et al., 2006, 2008) or showing LORETA source localizations (Lorig, Rigdon, & Poor, 2006). However, whereas inverse source localization algorithms, such as LORETA or BESA, have the potential for data simplification and clarification, these approaches provide genuinely model-dependent solutions that need to be cautiously considered, pending independent validation. Statistical analyses have relied on ERP component measures employing a ‘‘region-of-interest’’ approach, in which the topographic ERP signal is reduced to a few spatially smeared sites, and is also subject to experimenter bias in the selection or grouping of electrodes (Kayser & Tenke, 2005). Although the need to systematically identify the olfactory ERP component structure (i.e., how many major components with what temporal, spatial, and functional characteristics) has long been recognized (Lorig, 2000), only preliminary efforts have been made to date. ERP components are classically conceived as an electrophysiologic correlate of the underlying neuronal generators associated with information processes (cf. Kayser & Tenke, 2003). This conceptual definition implies that an ERP component is characterized by (1) temporal (latency), (2) spatial (scalp topography), and (3) functional (task or condition) specificity (e.g., Donchin et al., 1977; Fabiani, Gratton, & Coles, 2000). However, the identification and measurements of ‘‘obvious’’ peaks and troughs in the ERP waveforms as meaningful entities can be misleading. Specifying peaks in noisy waveforms (a problem not resolved but rather aggravated by using an automated computer algorithm) 1 There appears to be considerable confusion about the meaning of sources and sinks and their relationship to ERP waveforms. CSD estimates represent the current flow entering (sinks) and leaving (sources) the scalp from the underlying brain tissue and are therefore equally important in characterizing neuronal generator activity. As such, these estimates must be fully compatible with the ERP topography from which they are derived in order to be of empirical or descriptive value.

Olfactory ERP generator patterns in schizophrenia and determining area integration limits for deflections that invert and shift across scalp locations are subject to experimenter bias and raise questions of statistical independence, which will crucially affect their statistical analysis. Moreover, these ERP component measures depend directly on the recording reference, because the timing, topography, and amplitude of these ERP deflections will change with any other reference (e.g., Dien, 1998; Kayser, Fong, Tenke, & Bruder, 2003), thereby affecting component interpretation (e.g., polarity, topography, generator). Thus, the definition and measurement of appropriate ERP components and the dependency of surface potentials on a reference location (e.g., linked ears or mastoids, nose, average) are two problems that have plagued ERP research (e.g., Kayser & Tenke, 2003, 2005; Nunez & Srinivasan, 2006; Tenke & Kayser, 2005). We have proposed that these limitations can be overcome without sacrificing information by combining current source density (CSD; surface Laplacian) and temporal principal components analysis (PCA) to identify relevant, data-driven components (Kayser & Tenke, 2006a, 2006b; Kayser et al., 2006, 2010; Kayser, Tenke, Gates, & Bruder, 2007; Kayser, Tenke, Gil, & Bruder, 2009; Tenke et al., 2008; Tenke, Kayser, Stewart, & Bruder, 2010). CSD provides a representation of current generators that underlie ERPs and represent the magnitude of radial current flow entering (sink) and leaving (source) the scalp (e.g., Nunez & Srinivasan, 2006). CSD analysis is a reference-free technique (any EEG recording reference scheme will yield the same, unique CSD transform) that provides sharper topographies compared to those of scalp potentials while also reducing redundant contributions due to volume conduction (e.g., Tenke & Kayser, 2005). Often-raised concerns include the requirement of a high-density EEG montage for reliably computing CSDs, as well as their presumed insensitivity to deep sources. We have empirically addressed these concerns, demonstrating that no information is lost with the CSD transform when directly compared to the original ERPs, and deep or distributed sources, such as P3, are adequately and sufficiently represented (Kayser & Tenke, 2006a). A low-density EEG montage may be as efficient as a dense electrode montage in summarizing CSD activity for group data, because group averages effectively impose a spatial low-pass filter to the data (Kayser & Tenke, 2006b). In the specific context of olfactory ERPs, for which generators are presumably deep (i.e., with origins in olfactory, gustatory, or limbic structures), the corresponding fields and CSDs will be more diffuse at scalp, rendering a low–resolution surface Laplacian an advantage, rather than a liability. Thus, these conventional concerns have been overstated, and CSDs have not only been proven to be useful but may constitute an optimal analytic approach for many practical ERP applications. Compared to more complex EEG source localization methods (Michel et al., 2004), relying on surface Laplacian estimates as an analytic strategy is more conservative because it completely avoids additional (and unproven) biophysical assumptions (tissue conductivity and geometry, laminar orientation, number and independence of generators). Temporal PCA is one of the most frequently used multivariate decomposition approaches for ERP data and has been repeatedly shown to be superior to more traditional ERP measures, for instance, revealing more robust F statistics and better reliabilities (i.e., internal consistency and temporal stability) when directly compared with integrated time windows or baseline-to-peak measures (e.g., Beauducel, Debener, Brocke, & Kayser, 2000; Kayser et al., 1997; Kayser, Tenke, & Bruder, 1998). Often-cited

1077 limitations, such as misallocation of variance because of latency jitter, are not restricted to the use of temporal PCA but also affect traditional component measures and more severely (e.g., Beauducel & Debener, 2003; Chapman & McCrary, 1995; Donchin & Heffley, 1978; Wood & McCarthy, 1984). With careful consideration, temporal PCA can provide a concise and unbiased summary of ERP/CSD activity (Kayser & Tenke, 2003, 2006a) associated with generator patterns underlying stimulus processing, even for slow and long-lasting components (e.g., Kayser et al., 2006), and could therefore provide an answer to the question of relative statistical independence between putative olfactory components (Lorig, 2000). Moreover, because the extracted CSD factors are independent of the recording reference, they have an unambiguous component polarity and topography. A primary goal of this study was therefore to employ this new CSD-PCA approach for an improved characterization of OERPs (i.e., N1, P2) in schizophrenia patients and healthy adults. Following the findings of Turetsky et al. (2003), it was predicted that schizophrenia patients would show reduced N1 and P2 amplitudes (i.e., their CSD equivalents) when compared to healthy adults, and these OERP differences will be most evident at higher concentrations of H2S. Methods Participants As part of a study of olfaction and social function in schizophrenia, 35 healthy adults (ages 18–61 years, M ! SD 5 31.7 ! 12.0; 18 men; 6 smokers) without current or past psychopathology, neurological illness, or substance abuse (Nurnberger et al., 1994) were recruited for payment (US$10/h) from the New York metropolitan area. These controls were compared to 17 inpatients and 15 outpatients at New York State Psychiatric Institute (ages 18–54 years, M ! SD 5 33.3 ! 9.6; 18 men; 5 smokers) meeting DSM-IV (American Psychiatric Association, 1994) criteria for schizophrenia (n 5 26; 15 paranoid, 9 undifferentiated, 1 catatonic, 1 residual) or schizoaffective disorder (n 5 6; 3 bipolar type, 3 depressive type). Diagnoses were based on clinical interviews by psychiatrists and trained psychologists and a semistructured interview (Nurnberger et al., 1994) including items from commonly used instruments (e.g., Andreasen 1983, 1984). Symptom ratings were obtained using the Positive and Negative Syndrome Scale (PANSS; Kay, Opler, & Fishbein, 1992). The mean total Brief Psychiatric Rating Scale (BPRS) score available for 31 patients was 28.1 ! 6.6, with about equal scores for positive (10.8 ! 4.9) and negative (11.7 ! 3.9) symptoms (general 23.7 ! 5.8), indicating that patients were mildly disturbed. Mean age of onset available was 23.7 ! 6.3 years with an average illness duration of 9.8 ! 8.9 years. Most patients (n 5 29) were treated with antipsychotic medications (9 aripriprazole, 7 risperidone, 5 olanzapine, 4 ziprasidone, 2 perphenazine, 1 clozapine, 1 quetiapine; chlorpromazine equivalents 25–800 mg/day; Woods, 2003). All participants were right-handed (Oldfield, 1971; laterality quotient, controls vs. patients, 73.6 ! 29.2 vs. 84.0 ! 18.3). Patients tended to have less education than control participants, but this difference was of only marginal significance (14.2 ! 2.7 vs. 15.5 ! 1.7 years), F(1,63) 5 3.77, p 5 .06. Participants were instructed to refrain from smoking on the day of test. OERP recording sessions were scheduled between 9 a.m. and 5 p.m. and lasted about 1.5 h. Time of testing did not differ between groups, F(1,63)o1.0, n.s., thereby controlling for putative circadian in-

1078 fluences on OERP amplitudes (Nordin, Lo¨tsch, Murphy, Hummel, & Kobal, 2003). The experimental protocol had been approved by the institutional review board and was undertaken with the understanding and written consent of each participant. Stimuli and Procedure Participants were seated in an IAC sound–attenuated booth using a chin and forehead rest, with a video camera monitoring participants’ compliance and behavior. While focusing on a fixation cross and breathing normally through the nose,2 H2S stimuli (10 ppm, Scott Speciality Gases, Plumsteadville, PA) at high (undiluted) and low (diluted to 50%) concentrations were delivered to the left or right nostril by a constant-flow olfactometer (OM2s, Heinrich Burghart GmbH, Wedel, Germany) through a Teflon tube inserted approximately 1 cm into the naris. Stimulus duration was 200 ms (not more than 50 ms rise time according to manufacturer specification). For any given session, the air stream at the exit of the olfactometer had a constant flow rate (about 8 l/ min), temperature (the measured range was 381–391C at the olfactometers head to approximate 371C body temperature in the nasal cavity), and relative humidity (above 80%). Odors were presented in four blocks of 24 trials each, with a variable interstimulus interval (15–25 s). White noise of approximately 75 dB SPL was presented binaurally via Telephonics TDH–49P earphones to preclude hearing the switching valves. Participants responded as to whether they perceived a low or high intensity odor by raising their left or right hand, which was visually monitored and recorded by the experimenter. Therefore, the present paradigm consisted of an active odor intensity detection task, requiring conscious processing of and responding to perceived hydrogen sulfide stimuli. Because the time of odor stimulation was not cued, participants could fail to respond (miss). Nostril order and response hand assignment were counterbalanced across blocks and participants. Data Recording and Artifact Procedures All data recording and preprocessing closely followed the procedures detailed elsewhere (e.g., Kayser et al., 2007). Briefly, nose-referenced EEG (30 channels) and bipolar EOG activity were continuously recorded at 200 samples/s with a gain of 10k (5k horizontal, 2k vertical EOG) within 0.1–30 Hz (! 6 dB/ octave). Volume-conducted blink artifacts were removed from the raw EEG by spatial PCA. Recording epochs of 2000 ms (250 ms prestimulus baseline) were extracted off-line, tagged for A/D saturation, and low-pass filtered at 20 Hz (! 24 dB/octave). A reference-free approach identified residual artifacts on a channelby-channel and trial-by-trial basis (Kayser & Tenke, 2006d). A trial was rejected if it contained artifact in more than eight channels; otherwise, artifactual data were replaced by spherical spline 2 Although OERP studies typically trained participants to perform velopharyngeal closure as an active breathing technique to prevent intranasal respiratory airflow and interference during odor presentation, these potential benefits may be offset by the dual-task demands, resulting in divided attention that may alter odor processing. Comparisons of different breathing conditions with rather small sample sizes yielded conflicting results as to whether and how OERP amplitudes are affected (Lorig et al., 1996; Pause, Krauel, Sojka, & Ferstl, 1999; Thesen & Murphy, 2001). Given the likelihood of differences between healthy adults and schizophrenia patients in compliance with and capability of performing the velopharyngeal closure technique and that its associated systematic confounds (vigilance, attention) are more likely to affect odor detection and OERPs than the uncontrolled nasal air flow (cf. Laudien, Wencker, Ferstl, & Pause, 2008; Mainland & Sobel, 2006), a natural, spontaneous breathing condition seemed to be the preferred choice.

J. Kayser et al. interpolation (Perrin, Pernier, Bertrand, & Echallier, 1989) from artifact-free channels. These procedures for artifact detection and reduction were originally developed to optimize the signalto-noise ratio in problematic ERP recordings, such as those stemming from artifact-prone psychiatric samples, but these routines also help in reducing the problem of latency jitter in olfactory ERPs (Lorig, 2000). Excluding trials on which the participant ‘‘missed’’ the odor, and disregarding the participant’s high versus low intensity response, separate olfactory ERPs for high and low odor intensity were averaged from correctly detected, artifact-free trials using the entire 2-s epoch. To obtain more stable waveforms, ERPs were pooled across nostrils because of their blocked presentation order, and preliminary analyses did not reveal any effects of interest; furthermore, previous research has suggested that side of odor stimulation is of subordinate importance for measuring OERPs (e.g., Olofsson et al., 2006; Stuck et al., 2006). The mean number of trials (" SD) used to compute OERP averages were 30.7 " 8.4 and 23.1 " 8.9 (high vs. low intensity, respectively) for healthy controls and 30.0 " 8.0 and 23.6 " 8.5 for patients. As expected, more trials entered into high than low intensity ERP averages, F(1,63) 5 45.3, po.0001, but there were no differences between patients and controls. Visual inspections of the individual ERP waveforms also confirmed an acceptable signal-to-noise ratio for each participant. ERP waveforms were screened for electrolyte bridges (Tenke & Kayser, 2001), low-pass filtered at 12.5 Hz (! 24 dB/octave), and finally baseline corrected using the 100 ms preceding stimulus onset. ERPs were re-referenced to linked mastoids (TP9/10) for comparison to prior OERP studies using linked ear lobes or mastoids as reference. CSD Transform, Temporal PCA, and Statistical Analyses All OERP waveforms at each electrode were transformed into reference-free CSD estimates (mV/cm2 units; 10 cm head radius; 50 iterations; m 5 4; smoothing constant l 5 10 ! 5) using a spherical spline surface Laplacian (Perrin et al., 1989). To determine their common sources of variance, CSD waveforms were submitted to temporal PCA derived from the covariance matrix, followed by unrestricted Varimax rotation of the covariance loadings (Kayser & Tenke, 2003, 2006c). The input data matrix consisted of 401 variables (time interval ! 250 to 1750 ms) and 4,154 observations stemming from 67 participants, two intensities, and 31 electrode sites, including the nose. Data from two meaningful, high-variance CSD factors corresponding to N1 and P2 were submitted to repeated measures analysis of variance (ANOVA) with group (patients, controls) and gender (male, female) as between-subjects factors and odor intensity (high, low) as a within-subjects factor. The ANOVA designs also included subsets of lateral, homologous recording sites over both hemispheres at which PCA factor scores were largest and most representative of the associated CSD components (cf. Kayser & Tenke, 2006a; Kayser et al., 2006), adding hemisphere and site as within-subjects factors to the design. However, because subsets were selected on the premise that they collectively represent sink or source activity targeted in these statistical analyses, site effects were of secondary interest and will not be reported. It appears to be a fairly common, although incorrect, assumption that CSD methods necessarily identify equivalent current dipoles. Because multiple, overlapping generators with different geometries, time courses, and signal-to-noise ratios likely contribute to the ERP signal, it is not clear whether a

Olfactory ERP generator patterns in schizophrenia prominent sink–source pattern represents opposite poles of a single dipole or several dipoles with different orientations. This uncertainty is not resolved by inverse models that identify putative current dipoles to simplify these generators patterns. In the case of the auditory N1, which consists of bilateral medial-central sinks and inferior-temporal sources having corresponding time courses and spanning the Sylvian fissure, thereby matching the orientation of the well-known underlying generator (e.g., Kayser & Tenke, 2006a, 2006b; Kayser et al., 2007, 2009), the ventral source may be larger than the central sink and subject to greater EMG noise from the neck. Another example would be a midline closed-field generator as described for a novelty vertex source (Tenke et al., 2010), where bilateral dipole orientations yield local field cancellations. The point is that CSD does not provide a single dipole measure, nor does it require one. For its quantification, we are adopting a pragmatic approach by analyzing CSD activity at regions or sites associated with distinct sinks or sources. For analyses of the behavioral data, percentages of missed responses were submitted to a similar repeated measures ANOVA without the electrode factors. Sources of interactions and main effects were explored with simple effects (BMDP-4V; Dixon, 1992). When appropriate, Greenhouse–Geisser epsilon correction was used to compensate for violations of sphericity (e.g., Keselman, 1998). A conventional significance level (po.05) was applied for all effects. Results Behavioral Data The mean percentages of H2S stimuli that were missed (! SD) were 23.4 ! 17.5 and 44.9 ! 19.3 (high vs. low intensity, respectively) for healthy controls, and 22.5 ! 16.1 and 41.1 ! 20.2 for patients, yielding a highly significant main effect of odor intensity, F(1,63) 5 77.2, po.0001, but no effects involving group, all F(1,63)o1.0, n.s.

1079 & Tenke, 2006a, 2006b; Tenke & Kayser, 2005). In contrast, the reference-free CSD waveforms (black dashed lines) identified robust sink activity at these sites, which was not compromised by the choice of reference. Although the observed N1 sink and P2 source in the CSD waveforms directly corresponded to the N1 and P2 potentials in the OERP waveforms, marked topographic distinctions were evident, particularly with respect to the frontotemporal N1 sink and lateral frontal sinks associated with the mid-parietal P2 source.4 N1 sink and P2 source amplitudes were greater to high- than low-intensity H2S stimuli in both patients and healthy adults, further confirming their relationship to olfactory processing (Figure 2). Schizophrenia patients showed similar olfactory ERP and CSD waveforms when compared to controls, but their N1 sink and P2 source amplitudes were smaller. PCA Component Waveforms and Topographies The first four PCA factors effectively explained all of the systematic CSD variance (82.6% after rotation). The time courses of the factor loadings (Figure 3A) and the corresponding factor score topographies (Figure 3B) identified two factors corresponding to N1 sink (305 ms peak latency, lateral frontotemporal maximum) and P2 source (630 ms peak latency, mid-parietal maximum). Two later factors had a frontocentral (1015 ms) or parietal (1750 ms) midline sink maximum, suggesting a close correspondence to the response requirements in this task (i.e., raising left or right hand; cf. Kayser et al., 2007) and were therefore not further analyzed. Both healthy adults and schizophrenia patients had bilateral N1 sinks for the high odor concentration over the lateral temporal sites (Figure 3B, top, first column) and a corresponding mid-frontopolar source. Similarly, both controls and patients showed a medial parietal P2 source topography to both low and high odor concentrations, with current sinks maximal over lateral frontotemporal regions (Figure 3B, bottom, Columns 1 and 2). The reduced amplitude of the N1 sink and P2 source in patients was most evident to the high concentration of H2S.

Average ERP and CSD Waveforms To the best of our knowledge, no complete ERP waveform topography for olfactory stimuli has yet been published, except for selected midline ‘‘topographies’’ (Fz, Cz, Pz), probably because of concerns about individual specificity (Lorig, 2000). By overlaying individual ERPs and CSDs, we established that the grand means accurately summarized temporal and spatial properties of the observed OERP components. Figure 1 compares the grand mean olfactory ERP and CSD component structure for all 67 participants at all 31 scalp locations (averaged across intensities).3 The OERP waveforms (solid gray lines) showed a typical negative–positive component sequence, including an N1 potential (approximate peak latency 300 ms) believed to reflect initial sensory processing of olfactory stimuli followed by a P2 potential (600 ms), which is comparable to those reported in prior studies (Pause et al., 1996; Turetsky et al., 2003). By explicitly including the mastoid reference sites (TP9/10), however, it becomes obvious that recording sites along the reference-dependent isopotential line (e.g., T7/8, FT9/10, P9/10) showed little or no ERP activity. Thus, ERP activity at these sites is severely attenuated because of the arbitrary choice of the recording reference (Kayser

N1 sink. At lateral centrotemporal sites (T7/8, C3/4, FC5/6, CP5/6) for factor 305, there was a highly significant main effect of intensity, F(1,63) 5 131.7, po.0001, confirming the presence of the N1 sink for high but not low odor intensities (Figure 3B, top; for detailed ANOVA means, see supplementary Table A1). A significant Group " Intensity interaction, F(1,63) 5 6.11, p 5 .02, resulted from a reduction of N1 sink amplitude in schizophrenia for high- but not low-intensity stimuli: simple group main effects at high intensity, F(1,63) 5 5.87, p 5 .02, at low intensity, F(1,63) o1.0, n.s. There were also a significant interactions of Group " Gender, F(1,63) 5 4.15, p 5 .05, and of Group " Gender " Intensity, F(1,63) 5 4.87, p 5 .03, which originated from greater high intensity N1 sinks for healthy women compared to healthy men (M ! SD, # 1.42 ! 1.57 vs. # 0.97 ! 0.92), with patients showing the opposite gender effect (# 0.51 ! 0.95 vs. # 0.95 ! 1.02); simple Group " Gender interaction effects, at high intensity, F(1,63) 5 5.27, p 5 .03, at low intensity, F(1,63)o1.0, n.s.

3 The ERP/CSD component structure was highly comparable for healthy adults and schizophrenia patients (see Figures A1 and A2 in the supplementary material).

4 Animated ERP (linked-mastoids reference) and CSD topographies comparing groups and intensities can be obtained at URL http:// psychophysiology.cpmc.columbia.edu/oerp2008.html.

Repeated Measures ANOVA of PCA Factor Scores

1080

J. Kayser et al. – 20 µV

HEOG

+

VEOG

NOSE FP2

FP1

F7

F8 0

500 1000 1500

F3

F4

0

500 1000 1500

FZ FT10

FT9 0

FC5

500 1000 1500

FC6

N1 sink T7

C3

0

CZ

C4

T8

0

500 1000 1500 0

CP5

500 1000 1500

500 1000 1500

CP6

TP9

TP10 PZ P3

P4

P7

P8

P2 source 0

P9

500 1000 1500

O1

P10

O2 OZ

– 5 µV +

ERP Waveforms (Linked Mastoids)

Sink – 0

500 1000 1500

Latency [ms]

0.1 µV/cm2

CSD Waveforms (Laplacian)

Source +

Figure 1. Grand mean olfactory ERPs referenced to linked mastoids and reference–free CSD waveforms for the total sample (N 5 67) at all 31 recording sites (averaged across intensity). Horizontal and vertical electrooculograms (EOG), which are shown at a smaller scale before blink correction, indicate no eye artifact concerns. Two prominent CSD components are labeled at sites T7 (N1 sink) and Pz (P2 source), where they closely corresponded to their ERP counterparts.

The analysis for the frontopolar source (Fp1/2) accompanying the bilateral centrotemporal sinks for factor 305 revealed highly significant intensity, F(1,63) 5 27.2, po.0001, and Group ! Intensity effects, F(1,63) 5 7.71, p 5 .007, stemming from a greater high-larger-than-low-intensity amplitude difference for controls compared with patients (Figure 3B, top). Across groups, this source was also greater over the right than left frontopolar site: hemisphere main effect, F(1,63) 5 4.28, p 5 .04. P2 source. At medial-lateral centroparietal sites (P3/4, P7/8, CP5/6, C3/4) for factor 630, there was also a highly significant main effect of intensity, F(1,63) 5 74.5, po.0001, stemming from a greater P2 source for high than low odor concentration (Figure 3B, bottom; for detailed ANOVA means, see supplementary Table A2). A significant group main effect, F(1,63) 5 6.48, p 5 .01, and a highly significant Group ! Intensity interaction, F(1,63) 5 14.0, p 5 .0004, indicated smaller P2 source

in patients compared to healthy adults, which was significant for high (simple group main effect, F(1,63) 5 16.3, p 5 .0001) but not low intensity stimuli, F(1,63)o1.0, n.s. A significant hemisphere main effect, F(1,63) 5 5.99, p 5 .02, resulted from rightlarger-than-left P2 source across groups. A greater P2 source in women compared with men for both controls (M " SD, 0.71 " 0.94 vs. 0.40 " 0.81) and patients (0.38 " 0.77 vs. 0.31 " 0.72) yielded a significant gender main effect, F(1,63) 5 5.41, p 5 .02. The analysis for the lateral frontotemporal sinks (FT9/10, F7/ 8) accompanying the parietal P2 for factor 630 revealed a highly significant main effects of intensity, F(1,63) 5 16.8, p 5 .0001, hemisphere, F(1,63) 5 13.2, p 5 .0006, and gender, F(1,63) 5 14.1, p 5 .0004, which resulted from greater sinks for high compared to low intensity and right-larger-than-left hemisphere sinks (Figure 3B, bottom), and greater sinks in women than men (M " SD, # 0.97 " 0.89 vs. # 0.47 " 0.84). However, there were no significant effects involving group.

Olfactory ERP generator patterns in schizophrenia

1081

Healthy Adults (n = 35)

Schizophrenia Patients (n = 32) N1 sink T7

T7

0

500

1000

1500

0

500

1000

1500

Sink – 0.05 µV/cm2 Source + PZ

PZ

Odor Intensity Low High

P2 source

0

500 1000 Latency [ms]

1500

0

500 1000 Latency [ms]

1500

Figure 2. Grand mean olfactory CSDs for 35 healthy adults and 32 schizophrenia patients comparing low- and high-intensity H2S stimuli at sites T7 and Pz.

Discussion The application of the CSD-PCA approach identified factors corresponding to the N1 and P2 potentials, which have been consistently observed in OERP studies (Lorig, 2000; Pause & Krauel, 2000). Schizophrenia patients and healthy controls showed a prominent N1 sink over frontotemporal sites and a corresponding mid-frontopolar source. This topography is fully compatible with postulated generators within the medial temporal lobe and/or basal cortical regions (e.g., orbital frontal cortex; cf. Martzke et al., 1997). In addition, the observed N1 sink topography was distinctly unique, that is, it did not match generator patterns previously described for early visual (e.g., Kayser et al., 2006, 2007, 2009) or auditory components (e.g., Kayser & Tenke, 2006a, 2006b; Kayser et al., 2007, 2009; Tenke et al., 2008, 2010), which strongly suggests that the underlying neuronal activity may indeed reflect an early, modality-specific processing stage during odor perception. In contrast, the P2 source had a mid-parietal topography, with current sinks over lateral frontotemporal sites, which is compatible with the notion of a close association of olfactory P2 with a classical P3b potential (e.g., Lorig, 2000; Olofsson et al., 2008). Moreover, the observed P2 source topography was highly similar to P3 source topographies repeatedly found during working and recognition memory paradigms using visual or auditory word stimuli (e.g., Kayser et al., 2006, 2007, 2009, 2010) or during auditory oddball paradigms (e.g., Kayser & Tenke, 2006a, 2006b; Tenke et al., 2010). The corresponding generators of olfactory P2 are therefore consistent with those of P3 in other modalities, rather than with regions unique to olfaction. Although this agrees with the P3-like interpretation of the P2 source, the likeness of the olfactory N1 sink to N1 activity observed for other modalities may be challenged by the suggestion that the olfactory bulbs themselves may

be closer homologs to the primary sensory cortices of other modalities than are piriform cortex and related olfactory cortical regions (Haberly, 2001). In this scenario, it is unlikely that neuronal activity of primary olfactory processing, equivalent to calcarine or Heschl’s gyrus activation within the visual or auditory pathways, will propagate to scalp and may therefore not register as an ERP component. Another consideration is that the completely different organization of the olfactory system (e.g., lack of thalamocortical projections, afferent and efferent projections of primary sensory cortex vs. limbic cortex) makes a homology with N1 from other modalities improbable. Rather, olfactory N1 sink activity peaking around 300 ms may instead reflect functional activation of secondary olfactory regions, including piriform cortex, analogous to inferior-temporal visual association cortex (see Figure 13 in Haberly, 2001). The implication of this proposition is that N1 sink could be regarded as an olfactory N2, analogous to an auditory or visual N2. In this case, the olfactory N1 should be associated with stimulus categorization and classification, and the sequence of olfactory N1 sink and P2 source in the present odor detection paradigm would be the olfactory equivalent of an N2/P3 complex typically observed during many ERP paradigms, including an oddball task. Although it is not impossible that an olfactory N1 originates in basal cortex, and the observed bilateral temporal N1 sink pattern is not necessarily inconsistent with this assumption, the preferential access of olfaction to evaluative (also limbic) processes would suggest a functional correlate that is consistent with N2–like categorization. The N1 sink and P2 source were greater to high than low concentrations of H2S, which is in accord with prior studies (Stuck et al., 2006; Turetsky et al., 2003) and supports their relation to olfactory processing. It is also compatible with the idea the N1 sink reflects N2-like categorization processes, although future studies have to pursue this hypothesis with a more ap-

1082

J. Kayser et al.

A

CSD Factor Loading Waveforms

– 0.02 µV/cm2 + 305 (6.4%)

1015 (21.3%)

630 (20.2%) 0

250

500

1750 (34.7%)

750 1000 Latency [ms]

1250

1500

1750

CSD Factor Score Topographies

B

High

Low

High - Low

Controls

N1 sink (305)

–1.8

–1.8

0.0

0.0

+1.8

+1.8

Patients

–0.9

Controls – Patients

0.0 +0.9

High

Low

High - Low

Controls P2 source (630)

–1.6

–0.9

0.0

0.0

+1.6

+0.9

Patients

–0.6

Controls – Patients

0.0 +0.6

Figure 3. (a) Factor loadings of the first four PCA factors (labels indicate peak latency [with variance explained]) extracted from olfactory CSD waveforms (N 5 67). (b) CSD factor score topographies corresponding to N1 sink (top) and P2 source (bottom) for 35 healthy controls and 32 schizophrenia patients comparing low- and high-intensity H2S stimuli. Margins show difference topographies for intensity (high minus low) and group (controls minus patients).

Olfactory ERP generator patterns in schizophrenia propriate design, for instance, by including a broader parametric manipulation or different odors. Notably, as the current data were based on 12–16 trials per intensity level, it is evident that viable and meaningful olfactory ERP/CSD averages can be obtained with a relatively small number of trials. Schizophrenic patients had reduced N1 sink and P2 source amplitudes to the higher concentration of H2S, replicating the findings of Turetsky et al. (2003). The reduced OERPs in schizophrenia patients were present in the absence of behavioral differences between patients and controls. Schizophrenia patients showed considerable success in performing the olfaction task, and their behavioral performance was on a par with that for healthy controls. This indicates that the OERP reductions in schizophrenia are not due to a failure to attend to stimuli or overall poorer task performance. Instead, it is more parsimonious to presume that the OERP differences reflect an abnormality in obligatory processing of odors in cortical regions related to olfaction. Similarly, the lack of an association of olfactory identification and neurocognitive test performance (Continuous Performance Test and Wisconsin Card Sorting Test) has been cited as evidence that reduced olfactory function in schizophrenia is not secondary to deficits in attention or executive function (Seidman et al., 1997). It still remains to be demonstrated, however, whether the OERP deficits in schizophrenia are specific to olfactory processing or stem from a frontotemporal dysfunction that affects ERPs in multiple modalities. Given our N2-like interpretation of the olfactory N1 sink, its marked reduction in schizophrenia is in striking accordance with ERP evidence documenting profound reductions of N2 amplitudes across processing modalities and paradigms (e.g., Alain, Bernstein, He, Cortese, & Zipursky, 2002; Alain, Cortese, Bernstein, He, & Zipursky, 2001; Bruder et al., 1998, 1999; Kayser et al., 1999, 2001, 2009; O’Donnell et al., 1993; Umbricht, Bates, Lieberman, Kane, & Javitt, 2006). The reduction of N1 sink over lateral temporal lobe sites and P2 source over medial parietal sites in schizophrenia patients was bilateral and not dependent on hemisphere. However, the P2 source and lateral frontotemporal sink, as well as the frontopolar source accompanying N1, were greater over right than left hemisphere sites across both patients and healthy adults. In this regard, brain-damaged patients with lesions to the temporal lobe or orbitofrontal cortex, particularly in the right hemisphere, showed deficits in higher-order odor processing (Jones-Gotman & Zatorre, 1993), and patients with right-sided lesions of the frontal or temporal lobe showed decreased amplitudes of P2 and P3 potentials to odors at parietal sites (Daniels et al., 2001). Positron emission tomography (PET) studies measuring regional cerebral blood flow (rCBF) in healthy adults judging the pleasantness and intensity of odors have provided additional evidence supporting the important role of right orbitofrontal cortex in olfactory processing (Zatorre, Jones-Gotman, & Rouby, 2000). Malaspina et al. (1998) measured rCBF (using SPECT scans) in 6 schizophrenia patients and 7 controls during an odor identification task, and the patients showed hypometabolism in right cortical regions, including the inferior frontal area, superior temporal lobe, and supramarginal and angular gyrus. A review of hemodynamic evidence of lateralized olfactory processes suggested that olfactory stimuli differentially activate left or right brain regions, including medial temporal lobe and orbitofrontal cortex, but the inconsistent nature of this asymmetry has prompted suggestions that hemispheric differences depend on the cognitive or emotional processing demands (Royet & Plailly, 2004). Also, a study

1083 of laterality of OERPs during monorhinal stimulation with amyl acetate in 28 healthy adults found generally larger N1/P2 amplitudes for left than right nostril stimulation and at left than right hemisphere sites for left nostril stimuli (Olofsson et al., 2006). A related issue that has attracted less attention in this context is the potential confound of blocking left or right stimulus presentations as mandated by use of an olfactometer, such as the one used in the current study. Blocked unilateral odor presentations could lead to corresponding contralateral shifts in attention (cf. Kinsbourne, 1970), which may interfere with the predominantly ipsilateral organization of the olfactory system (e.g., Martzke et al., 1997). Thus, additional research is needed to clarify the nature of hemispheric asymmetries of OERPs and their relation to stimulus and task demands. A gender effect was found for the N1 sink that differed across groups. Namely, healthy women showed greater N1 for the high concentration of H2S compared to healthy men, whereas schizophrenia patients showed the opposite gender effect. P2 also showed a gender effect, with women showing greater source and sink activity than men, but this was not dependent on group. Although Kopala, Clark, and Hurwitz (1989) originally reported that men with schizophrenia had greater olfactory impairment than women for smell identification, more recent studies by this and other groups have not replicated this gender effect (Kopala, Good, Martzke, & Hurwitz., 1995; Moberg et al., 1999; Seidman et al., 1997). Although we know of no reports examining gender differences in OERPs of schizophrenia patients, Becker et al. (1993) found larger P1/N1 and N1/P2 amplitudes for vanillin and H2S odorants in women compared to men in a sample of healthy and psychosis-prone subjects (i.e., gender differences were unaffected by group classification), and Stuck et al. (2006) also found larger P2 amplitudes to H2S in healthy women than men. Lundstro¨m and Hummel (2006), measuring ERPs of healthy adults to peppermint, which activates both olfactory and trigeminal systems, did not find a gender effect for P2 amplitude but did report that women had larger amplitude of N1 over the left than right hemisphere, whereas men had larger P1 amplitude over the right than left hemisphere. Although these studies suggest possible gender effects in OERPs, the extent to which they differ in schizophrenia patients and healthy adults needs further study. There are several limitations of this study that should be noted. First, participants responded to the odors by raising their hand. Although this is unlikely to have affected the earlier OERP components (N1 or P2), it may have interfered with the measurement of later components (cf. Kayser et al., 2007). Second, subjects were not cued as to the time of odor presentation, and there was also no control of their breathing technique (i.e., natural breathing through mouth and nose). Although this could well have increased the variability of OERP measurements, leading to overall reduced OERP amplitudes compared to controlled breathing procedures (cf. Pause et al., 1999; Thesen & Murphy, 2001), there is no reason to believe that it would have differentially affected the schizophrenia patients and healthy adults. Third, OERPs were measured only to the unpleasant smelling odor of H2S. One of the distinguishing features of olfactory stimuli is their strong affective associations and the brain regions mediating olfaction overlap with those mediating emotional processing. The extent to which deficits in OERPs in schizophrenia are related to the emotional valence of the odors is an important issue for future research (cf. Pause et al., 2008). Fourth, although the lack of antipsychotic medication control is

1084

J. Kayser et al.

also a limitation, there is little evidence that medication status is related to performance on psychophysical measures of olfactory function (Moberg et al., 1999); however, the reported relation of neuroleptic treatment to asymmetrical olfactory thresholds (Purdon & Flor-Henry, 2000) may imply a more complex moderating influence of drug treatment on olfactory function. Lastly, this study compared schizophrenia patients and healthy controls, but there were marked individual differences in the OERPs among patients, which raises the possibility that only a subgroup of schizophrenia patients have OERP deficits. Further study should be given to examining clinical, neurophysiological, and neuroanatomical correlates of olfactory deficits in schizophrenia. Apart from replicating the original findings of Turetsky et al. (2003) with a considerably larger sample, the current study ad-

vances olfactory ERP research by providing a complete, comparative topographic analysis of reference-independent current source densities underlying reference-dependent surface potentials. The PCA-based summary of orthogonal variance contributions identified a distinct, bilateral temporal N1 sink that appears to be unique to olfactory stimuli. This PCA-CSD component has a subtle ERP counterpart with similar topography that has not yet been reported in the literature, presumably because the common choice of a linked–mastoids reference attenuates the visibility of this topographic effect. In contrast, the topography of P2 source, the second prominent PCA-CSD component, was found to be highly similar to P3 source topographies observed for other stimulus modalities. The topographic CSD findings and insights for olfactory N1 and P2 are unique and may help stimulate methodological and theoretical advancements in the field.

REFERENCES Alain, C., Bernstein, L. J., He, Y., Cortese, F., & Zipursky, R. B. (2002). Visual feature conjunction in patients with schizophrenia: An event– related brain potential study. Schizophrenia Research, 57, 69–79. Alain, C., Cortese, F., Bernstein, L. J., He, Y., & Zipursky, R. B. (2001). Auditory feature conjunction in patients with schizophrenia. Schizophrenia Research, 49, 179–191. American Psychiatric Association. (1994). Diagnostic and Statistical Manual of the Mental Disorders (4th ed). Washington, DC: American Psychiatric Association. Andreasen, N. C. (1983). The Scale for the Assessment of Negative Symptoms (SANS). Iowa City: The University of Iowa. Andreasen, N. C. (1984). The Scale for the Assessment of Positive Symptoms (SAPS). Iowa City: The University of Iowa. Beauducel, A., & Debener, S. (2003). Misallocation of variance in event– related potentials: Simulation studies on the effects of test power, topography, and baseline–to–peak versus principal component quantifications. Journal of Neuroscience Methods, 124, 103–112. Beauducel, A., Debener, S., Brocke, B., & Kayser, J. (2000). On the reliability of augmenting/reducing: Peak amplitudes and principal components analysis of auditory evoked potentials. Journal of Psychophysiology, 14, 226–240. Becker, E., Hummel, T., Piel, E., Pauli, E., Kobal, G., & Hautzinger, M. (1993). Olfactory event–related potentials in psychosis–prone subjects. International Journal of Psychophysiology, 15, 51–58. Bruder, G., Kayser, J., Tenke, C., Amador, X., Friedman, M., Sharif, Z., et al. (1999). Left temporal lobe dysfunction in schizophrenia: Eventrelated potential and behavioral evidence from phonetic and tonal dichotic listening tasks. Archives of General Psychiatry, 56, 267–276. Bruder, G., Kayser, J., Tenke, C., Rabinowicz, E., Friedman, M., Amador, X., et al. (1998). The time course of visuospatial processing deficits in schizophrenia: An event-related brain potential study. Journal of Abnormal Psychology, 107, 399–411. Chapman, R. M., & McCrary, J. W. (1995). EP component identification and measurement by principal components analysis. Brain and Cognition, 27, 288–310. Covington, J. W., Geisler, M. W., Polich, J., & Murphy, C. (1999). Normal aging and odor intensity effects on the olfactory event-related potential. International Journal of Psychophysiology, 32, 205–214. Daniels, C., Gottwald, B., Pause, B. M., Sojka, B., Mehdorn, H. M., & Ferstl, R. (2001). Olfactory event-related potentials in patients with brain tumors. Clinical Neurophysiology, 112, 1523–1530. Dien, J. (1998). Issues in the application of the average reference: Review, critiques, and recommendations. Behavior Research Methods, Instruments, & Computers, 30, 34–43. Dixon, W. J. (Ed.). (1992). BMDP Statistical Software Manual: to Accompany the 7.0 Software Release. Berkeley, CA: University of California Press. Donchin, E., Callaway, R., Cooper, R., Desmedt, J. E., Goff, W. R., Hillyard, S. A., et al. (1977). Publication criteria for studies of evoked potentials (EP) in man. In J. E. Desmedt (Ed.), Progress in clinical neurophysiology: Vol. 1. Attention, voluntary contraction and event– related cerebral potentials (pp. 1–11). Basel: Karger.

Donchin, E., & Heffley, E. F. (1978). Multivariate analysis of event– related potential data: A tutorial review. In D. A. Otto (Ed.), Multidisciplinary perspectives in event–related brain potential research. Proceedings of the fourth international congress on event–related slow potentials of the brain (EPIC IV), Hendersonville, NC, April 4–10, 1976 (pp. 555–572). Washington, DC: The Office. Evans, W. J., Kobal, G., Lorig, T. S., & Prah, J. D. (1993). Suggestions for collection and reporting of chemosensory (olfactory) event-related potentials. Chemical Senses, 18, 751–756. Fabiani, M., Gratton, G., & Coles, M. G. H. (2000). Event–related brain potentials: Methods, theory, and applications. In J. T. Cacioppo, L. G. Tassinary, & G. G. Bernston (Eds.), Handbook of psychophysiology (2nd ed, pp. 53–84). Cambridge, UK: Cambridge University Press. Haberly, L. B. (2001). Parallel–distributed processing in olfactory cortex: New insights from morphological and physiological analysis of neuronal circuitry. Chemical Senses, 26, 551–576. Jones–Gotman, M., & Zatorre, R. J. (1993). Odor recognition memory in humans: Role of right temporal and orbitofrontal regions. Brain and Cognition, 22, 182–198. Kay, S. R., Opler, L. A., & Fishbein, A. (1992). Positive and negative syndrome scale (PANSS) rating manual. Toronto, Canada: Multihealth System Inc. Kayser, J., Bruder, G. E., Friedman, D., Tenke, C. E., Amador, X. F., Clark, S. C., et al. (1999). Brain event-related potentials (ERPs) in schizophrenia during a word recognition memory task. International Journal of Psychophysiology, 34, 249–265. Kayser, J., Bruder, G. E., Tenke, C. E., Stuart, B. K., Amador, X. F., & Gorman, J. M. (2001). Event-related brain potentials (ERPs) in schizophrenia for tonal and phonetic oddball tasks. Biological Psychiatry, 49, 832–847. Kayser, J., Fong, R., Tenke, C. E., & Bruder, G. E. (2003). Event-related brain potentials during auditory and visual word recognition memory tasks. Cognitive Brain Research, 16, 11–25. Kayser, J., & Tenke, C. E. (2003). Optimizing PCA methodology for ERP component identification and measurement: Theoretical rationale and empirical evaluation. Clinical Neurophysiology, 114, 2307– 2325. Kayser, J., & Tenke, C. E. (2005). Trusting in or breaking with convention: Towards a renaissance of principal components analysis in electrophysiology. Clinical Neurophysiology, 116, 1747– 1753. Kayser, J., & Tenke, C. E. (2006a). Consensus on PCA for ERP data, and sensibility of unrestricted solutions. Clinical Neurophysiology, 117, 703–707. Kayser, J., & Tenke, C. E. (2006b). Electrical distance as a reference-free measure for identifying artifacts in multichannel electroencephalogram (EEG) recordings. Psychophysiology, 43, S51. Kayser, J., & Tenke, C. E. (2006c). Principal components analysis of Laplacian waveforms as a generic method for identifying ERP generator patterns: I. Evaluation with auditory oddball tasks. Clinical Neurophysiology, 117, 348–368.

Olfactory ERP generator patterns in schizophrenia Kayser, J., & Tenke, C. E. (2006d). Principal components analysis of Laplacian waveforms as a generic method for identifying ERP generator patterns: II. Adequacy of low-density estimates. Clinical Neurophysiology, 117, 369–380. Kayser, J., Tenke, C. E., & Bruder, G. E. (1998). Dissociation of brain ERP topographies for tonal and phonetic oddball tasks. Psychophysiology, 35, 576–590. Kayser, J., Tenke, C. E., Gates, N. A., & Bruder, G. E. (2007). Reference-independent ERP old/new effects of auditory and visual word recognition memory: Joint extraction of stimulus- and response-locked neuronal generator patterns. Psychophysiology, 44, 949–967. Kayser, J., Tenke, C. E., Gates, N. A., Kroppmann, C. J., Gil, R. B., & Bruder, G. E. (2006). ERP/CSD indices of impaired verbal working memory subprocesses in schizophrenia. Psychophysiology, 43, 237– 252. Kayser, J., Tenke, C. E., Gil, R. B., & Bruder, G. E. (2009). Stimulus– and response–locked neuronal generator patterns of auditory and visual word recognition memory in schizophrenia. International Journal of Psychophysiology, 73, 186–206. Kayser, J., Tenke, C. E., Kroppmann, C. J., Fekri, S., Alschuler, D. M., Gates, N. A., et al. (2010). Current source density (CSD) old/new effects during recognition memory for words and faces in schizophrenia and in healthy adults. International Journal of Psychophysiology, 75, 194–210. Kayser, J., Tenke, C., Nordby, H., Hammerborg, D., Hugdahl, K., & Erdmann, G. (1997). Event–related potential (ERP) asymmetries to emotional stimuli in a visual half–field paradigm. Psychophysiology, 34, 414–426. Keselman, H. J. (1998). Testing treatment effects in repeated measures designs: An update for psychophysiological researchers. Psychophysiology, 35, 470–478. Kettenmann, B., Hummel, C., Stefan, H., & Kobal, G. (1997). Multiple olfactory activity in the human neocortex identified by magnetic source imaging. Chemical Senses, 22, 493–502. Kinsbourne, M. (1970). The cerebral basis of lateral asymmetries in attention. Acta Psychologica, 33, 193–201. Kobal, G. (1982). A new method for determination of the olfactory and the trigeminal nerve’s dysfunction: Olfactory (OEP) and chemical somatosensensory (CSEP) evoked potentials. In A. Rothenberger (Ed.), Event-related potentials in children (pp. 455–461). Amsterdam: Elsevier Biomedical Press. Kobal, G. (2003). Electrophysiological measurement of olfactory function. In R. L. Doty (Ed.), Handbook of olfaction and gustation (2nd ed, pp. 229–249). New York: Marcel Dekker. Kobal, G., & Hummel, T. (1998). Olfactory and intranasal trigeminal event-related potentials in anosmic patients. Laryngoscope, 108, 1033–1035. Kopala, L., Clark, C., & Hurwitz, T. A. (1989). Sex differences in olfactory function in schizophrenia. American Journal of Psychiatry, 146, 1320–1322. Kopala, L., Good, K., Martzke, J., & Hurwitz, T. (1995). Olfactory deficits in schizophrenia are not a function of task complexity. Schizophrenia Research, 17, 195–199. Kru¨ger, S., Frasnelli, J., Bra¨unig, P., & Hummel, T. (2006). Increased olfactory sensitivity in euthymic patients with bipolar disorder with event–related episodes compared with patients with bipolar disorder without such episodes. Journal of Psychiatry & Neuroscience, 31, 263– 270. Laudien, J. H., Kuster, D., Sojka, B., Ferstl, R., & Pause, B. M. (2006). Central odor processing in subjects experiencing helplessness. Brain Research, 1120, 141–150. Laudien, J. H., Wencker, S., Ferstl, R., & Pause, B. M. (2008). Context effects on odor processing: An event-related potential study. NeuroImage, 41, 1426–1436. Lorig, T. S. (2000). The application of electroencephalographic techniques to the study of human olfaction: A review and tutorial. International Journal of Psychophysiology, 36, 91–104. Lorig, T. S., Matia, D. C., Peszka, J., & Bryant, D. N. (1996). The effects of active and passive stimulation on chemosensory event-related potentials. International Journal of Psychophysiology, 23, 199–205. Lorig, T. S., Rigdon, M., & Poor, A. (2006). Temporal pattern of odor administration alters hemispheric processing in humans. NeuroReport, 17, 231–234.

1085 Lo¨tsch, J., & Hummel, T. (2006). The clinical significance of electrophysiological measures of olfactory function. Behavioral Brain Research, 170, 78–83. Lundstro¨m, J. N., & Hummel, T. (2006). Sex–specific hemispheric differences in cortical activation to a bimodal odor. Behavioral Brain Research, 166, 197–203. Lundstro¨m, J. N., Seven, S., Olsson, M. J., Schaal, B., & Hummel, T. (2006). Olfactory event–related potentials reflect individual differences in odor valence perception. Chemical Senses, 31, 705–711. Mainland, J., & Sobel, N. (2006). The sniff is part of the olfactory percept. Chemical Senses, 31, 181–196. Malaspina, D., Perera, G. M., Lignelli, A., Marshall, R. S., Esser, P. D., Storer, S., et al. (1998). SPECT imaging of odor identification in schizophrenia. Psychiatry Research, 82, 53–61. Martzke, J. S., Kopala, L. C., & Good, K. P. (1997). Olfactory dysfunction in neuropsychiatric disorders: Review and methodological considerations. Biological Psychiatry, 42, 721–732. Michel, C. M., Murray, M. M., Lantz, G., Gonzalez, S., Spinelli, L., & Grave de Peralta, R. (2004). EEG source imaging. Clinical Neurophysiology, 115, 2195–2222. Moberg, P. J., Agrin, R., Gur, R. E., Gur, R. C., Turetsky, B. I., & Doty, R. L. (1999). Olfactory dysfunction in schizophrenia: A qualitative and quantitative review. Neuropsychopharmacology, 21, 325–340. Moberg, P. J., Arnold, S. E., Doty, R. L., Gur, R. E., Balderston, C. C., Roalf, D. R., et al. (2006). Olfactory functioning in schizophrenia: Relationship to clinical, neuropsychological, and volumetric MRI measures. Journal of Clinical and Experimental Neuropsychology, 28, 1444–1461. Morgan, C. D., Geisler, M. W., Covington, J. W., Polich, J., & Murphy, C. (1999). Olfactory P3 in young and older adults. Psychophysiology, 36, 281–287. Murphy, C., Morgan, C. D., Geisler, M. W., Wetter, S., Covington, J. W., Madowitz, M. D., et al. (2000). Olfactory event-related potentials and aging: Normative data. International Journal of Psychophysiology, 36, 133–145. Nordin, S., Lo¨tsch, J., Murphy, C., Hummel, T., & Kobal, G. (2003). Circadian rhythm and desensitization in chemosensory event-related potentials in response to odorous and painful stimuli. Psychophysiology, 40, 612–619. Nunez, P. L., & Srinivasan, R. (2006). Electric fields of the brain: The neurophysics of EEG. New York: Oxford University Press. Nurnberger, J. I. Jr., Blehar, M. C., Kaufmann, C. A., York Cooler, C., Simpson, S. G., Harkavy-Friedman, J., et al. (1994). Diagnostic interview for genetic studies. Rationale, unique features, and training. NIMH Genetics Initiative. Archives of General Psychiatry, 51, 849–859. O’Donnell, B. F., Shenton, M. E., McCarley, R. W., Faux, S. F., Smith, R. S., Salisbury, D. F., et al. (1993). The auditory N2 component in schizophrenia: Relationship to MRI temporal lobe gray matter and to other ERP abnormalities. Biological Psychiatry, 34, 26–40. Oldfield, R. C. (1971). The assessment and analysis of handedness: The Edinburgh inventory. Neuropsychologia, 9, 97–113. Olofsson, J. K., Broman, D. A., Gilbert, P. E., Dean, P., Nordin, S., & Murphy, C. (2006). Laterality of the olfactory event-related potential response. Chemical Senses, 31, 699–704. Olofsson, J. K., Ericsson, E., & Nordin, S. (2008). Comparison of chemosensory, auditory and visual event-related potential amplitudes. Scandinavian Journal of Psychology, 49, 231–237. Olofsson, J. K., & Nordin, S. (2004). Gender differences in chemosensory perception and event-related potentials. Chemical Senses, 29, 629–637. Pause, B. M., Hellmann, G., Goder, R., Aldenhoff, J. B., & Ferstl, R. (2008). Increased processing speed for emotionally negative odors in schizophrenia. International Journal of Psychophysiology, 70, 16–22. Pause, B. M., & Krauel, K. (2000). Chemosensory event-related potentials (CSERP) as a key to the psychology of odors. International Journal of Psychophysiology, 36, 105–122. Pause, B. M., Krauel, K., Sojka, B., & Ferstl, R. (1999). Is odor processing related to oral breathing? International Journal of Psychophysiology, 32, 251–260. Pause, B. M., Raack, N., Sojka, B., Goder, R., Aldenhoff, J. B., & Ferstl, R. (2003). Convergent and divergent effects of odors and emotions in depression. Psychophysiology, 40, 209–225.

1086 Pause, B. M., Sojka, B., Krauel, K., & Ferstl, R. (1996). The nature of the late positive complex within the olfactory event-related potential (OERP). Psychophysiology, 33, 376–384. Perrin, F., Pernier, J., Bertrand, O., & Echallier, J. F. (1989). Spherical splines for scalp potential and current density mapping [Corrigenda EEG 02274, EEG Clin. Neurophysiol., 1990, 76, 565]. Electroencephalography and Clinical Neurophysiology, 72, 184–187. Purdon, S. E., & Flor–Henry, P. (2000). Asymmetrical olfactory acuity and neuroleptic treatment in schizophrenia. Schizophrenia Research, 44, 221–232. Rombaux, P., Mouraux, A., Bertrand, B., Guerit, J. M., & Hummel, T. (2006). Assessment of olfactory and trigeminal function using chemosensory event-related potentials. Neurophysiologie Clinique, 36, 53– 62. Royet, J. P., & Plailly, J. (2004). Lateralization of olfactory processes. Chemical Senses, 29, 731–745. Seidman, L. J., Goldstein, J. M., Goodman, J. M., Koren, D., Turner, W. M., Faraone, S. V., et al. (1997). Sex differences in olfactory identification and Wisconsin Card Sorting performance in schizophrenia: Relationship to attention and verbal ability. Biological Psychiatry, 42, 104–115. Sobel, N., Prabhakaran, V., Desmond, J. E., Glover, G. H., Goode, R. L., Sullivan, E. V., et al. (1998). Sniffing and smelling: Separate subsystems in the human olfactory cortex. Nature, 392, 282–286. Stuck, B. A., Frey, S., Freiburg, C., Hormann, K., Zahnert, T., & Hummel, T. (2006). Chemosensory event-related potentials in relation to side of stimulation, age, sex, and stimulus concentration. Clinical Neurophysiology, 117, 1367–1375. Tateyama, T., Hummel, T., Roscher, S., Post, H., & Kobal, G. (1998). Relation of olfactory event-related potentials to changes in stimulus concentration. Electroencephalography and Clinical Neurophysiology, 108, 449–455. Tenke, C. E., & Kayser, J. (2001). A convenient method for detecting electrolyte bridges in multichannel electroencephalogram and event-related potential recordings. Clinical Neurophysiology, 112, 545–550. Tenke, C. E., & Kayser, J. (2005). Reference-free quantification of EEG spectra: Combining current source density (CSD) and frequency principal components analysis (fPCA). Clinical Neurophysiology, 116, 2826–2846. Tenke, C. E., Kayser, J., Shankman, S. A., Griggs, C. B., Leite, P., Stewart, J. W., et al. (2008). Hemispatial PCA dissociates temporal from parietal ERP generator patterns: CSD components in healthy adults and depressed patients during a dichotic oddball task. International Journal of Psychophysiology, 67, 1–16. Tenke, C. E., Kayser, J., Stewart, J. W., & Bruder, G. E. (2010). Novelty P3 reductions in depression: Characterization using principal components analysis (PCA) of current source density (CSD) waveforms. Psychophysiology, 47, 133–146. Thesen, T., & Murphy, C. (2001). Age-related changes in olfactory processing detected with olfactory event-related brain potentials using velopharyngeal closure and natural breathing. International Journal of Psychophysiology, 40, 119–127.

J. Kayser et al. Turetsky, B. I., Kohler, C. G., Gur, R. E., & Moberg, P. J. (2008). Olfactory physiological impairment in first–degree relatives of schizophrenia patients. Schizophrenia Research, 102, 220–229. Turetsky, B. I., Moberg, P. J., Owzar, K., Johnson, S. C., Doty, R. L., & Gur, R. E. (2003). Physiologic impairment of olfactory stimulus processing in schizophrenia. Biological Psychiatry, 53, 403–411. Umbricht, D. S., Bates, J. A., Lieberman, J. A., Kane, J. M., & Javitt, D. C. (2006). Electrophysiological indices of automatic and controlled auditory information processing in first–episode, recent–onset and chronic schizophrenia. Biological Psychiatry, 59, 762–772. Wood, C. C., & McCarthy, G. (1984). Principal component analysis of event–related potentials: Simulation studies demonstrate misallocation of variance across components. Electroencephalography and Clinical Neurophysiology, 59, 249–260. Woods, S. W. (2003). Chlorpromazine equivalent doses for the newer atypical antipsychotics. Journal of Clinical Psychiatry, 64, 663–667. Zatorre, R. J., Jones–Gotman, M., & Rouby, C. (2000). Neural mechanisms involved in odor pleasantness and intensity judgments. NeuroReport, 11, 2711–2716.

Supplementary Material The following supplementary material is available for this article (all figures provided in PDF format): Figure A1. Grand mean olfactory ERP (in mV) waveforms referenced to linked mastoids for 35 healthy adults and 32 schizophrenia patients at all 31 recording sites (averaged across intensity). Horizontal and vertical electrooculograms (EOG), which are shown at a smaller scale before blink correction, indicate no eye artifact concerns. Two prominent ERP components are labeled at sites T7 (N1) and Pz (P2). Figure A2. Reference-free CSD (mV/cm2) waveforms for 35 healthy adults and 32 schizophrenia patients at all 31 recording sites (averaged across intensity). Two prominent CSD components are labeled at sites T7 (N1 sink) and Pz (P2 source). Table A1. Means (! SD) of N1 sink (factor 305) Table A2. Means (! SD) of P2 source (factor 630) This material is available as part of the online article from: http://www.blackwell-synergy.com/doi/10.1111/j.1469-8986. 2010.01013.x. (This link will take you to the article abstract). Please note: Blackwell Publishing is not responsible for the content or functionality of any supplementary materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article. (Received August 25, 2009; Accepted November 13, 2009)

Psychophysiology, 47 (2010), 1087–1093. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01010.x

Comparison of a three-quarter electrode band configuration with a full electrode band configuration for impedance cardiography

SIMON L. BACON,a,b,c AVRIL J. KELLER,d KIM L. LAVOIE,b,c,e and TAVIS S. CAMPBELLd a

Department of Exercise Science, Concordia University, Montreal, Quebec, Canada Montreal Behavioural Medicine Centre, Hopital du Sacre-Coeur de MontrealFA University of Montreal affiliated hospital, Montreal, Quebec, Canada c Montreal Behavioural Medicine Centre, Montreal Heart InstituteFA University of Montreal affiliated hospital, Montreal, Quebec, Canada d Department of Psychology, University of Calgary, Calgary, Alberta, Canada e Department of Psychology, University of Quebec at Montreal (UQAM), Montreal, Quebec, Canada b

Abstract Impedance cardiography is a technique commonly used in psychophysiological studies. However, concerns about the utility of full circumferential band electrodes (FB) have been raised. The current study was designed to compare FB with a three-quarter circumferential band configuration (PB). A total of 47 participants (66% female, mean [SD] age 5 20.4 [3.0] years) underwent 2 testing sessions, once using FB and once using PB. Session order was randomized and balanced. Each session consisted of 5 min of rest, math task, recovery, and cold pressor test. Average baseline and task pre-ejection period (PEP), stroke volume (SV), cardiac output (CO), heart rate (HR), blood pressure (BP), and total peripheral resistance (TPR) was calculated from impedance cardiography and blood pressure monitoring. Participants were are asked to rate measures of comfort after each session. There were no significant difference between the mean levels of PEP, SV, CO, HR, and TPR for the PB versus the FB configurations. However, both systolic BP and diastolic BP were higher during the FB session. Intraclass correlations were high (ricc 5 .63–.93) between PB and FB. Bland-Altman analyses revealed a low level of bias (! 5%) between the configurations. Based on limits of agreement between " 30%, there was equivalence in PEP between the 2 configurations, and SV, CO, and TPR were close to reaching equivalence. Participants clearly indicated greater comfort with the PB configuration compared to the FB. The current study provides incremental evidence that suggests a three-quarter PB configuration may be utilized for standard psychophysiological testing instead of the standard FB configuration. However, further studies are needed to validate the PB configuration against other techniques. Descriptors: Band configurations, Impedance cardiography, Bias, Limits of agreement

band electrodes (Kubicek, Karnegis, Patterson, Witsoe, & Mattson, 1966). However, this setup has limitations, as participants tend to find it uncomfortable, potentially inducing anxiety and feelings of restricted breathing (McGrath, O’Brien, Hassinger, & Shah, 2005; Sherwood et al., 1990). To overcome these limitations, several alternate electrode setups have been examined. Since the introduction of the first spot electrode configuration (Penney, Patwardhan, & Wheeler, 1985), there have been a number of proposed variations in both the placement of the spot electrodes and in the number of electrodes used (Ikarashi, Nogawa, Tanaka, & Yamakoshi, 2007; Ikarashi, Nogawa, Yamakoshi, Tanaka, & Yamakoshi, 2006; Wang, Haynor, & Kim, 2001). Several of these have been validated against more invasive methods of assessing cardiac output (e.g., thermodilution; Bernstein & Lemmens, 2005; Doering, Lum, Dracup, & Friedman, 1995), and, in comparison to band configurations, there is the potential for better signal-to-noise ratios with spot electrodes (Qu, Zhang, Webster, & Tompkins, 1986).

Impedance cardiography is a technique commonly used to measure non-invasively hemodynamic parameters in psychophysiological studies (Sherwood et al., 1990). Since its introduction over 50 years ago (Nyboer, 1959; Nyboer, Bagno, Barnett, & Halsey, 1940), there have been many advances in impedance cardiography methodology, including measuring devices, analytical equations, and electrode types and configurations. The original commercial impedance cardiography setup utilized a tetra-polar band electrode configuration with two full circumferential bands placed around the neck and two around the thorax, with lead connections positioned at either end of the circumferential Salary support was provided by the Fonds de la recherche en sante´ du Que´bec (FRSQ) (KLL & SLB) and the Canadian Institutes of Health Research (SLB). Address reprint requests to: Tavis S. Campbell, Ph.D., University of Calgary, Department of Psychology, 2500 University Dr. NW, Calgary, Alberta T2N 1N4, Canada. E-mail: [email protected] 1087

1088 In spite of the advances made with spot electrode configurations, there are remaining problems, and some researchers still default to using band configurations. Such researchers cite concerns associated with reduced reproducibility and poor interstudy comparisons, both within studies using spot configurations and across studies comparing spot and band configurations (McGrath et al., 2005). These problems generally arise because of the need for greater care with spot electrode site preparation (Sherwood et al., 1990), greater variations in electrode placement with spot configurations (Ikarashi et al., 2006, 2007; Sherwood, Royal, Hutcheson, & Turner, 1992), and systematic differences in absolute values of impedance measures between spot and band configurations (Gotshall & Sexson, 1994; Sherwood et al., 1992; Woltjer, van der Meer, Bogaard, & de Vries, 1995). In comparison to the literature on spot electrodes, there have been relatively few studies that have utilized alternate band configurations. For example, half band (Ring, Liu, & Brener, 1994; Watanabe, Kamide, Torii, & Ochiai, 1981) and three-quarter band (Bacon et al., 2006) configurations have been used, though these have not necessarily been applied to all four bands (Bacon et al., 2006) and have not necessarily been formally compared to full band configurations. Of note, one preliminary study previously assessed five diminishing alternative band configurations against the standard full band array (Wilmers & Brener, 1991). Each of these configurations covered progressively less of the circumferential distance around the body. This study found that a configuration that spanned three quarters of the circumferential distance was the shortest configuration to provide impedance measures that were comparable to the full band. However, it should be noted that this study was only published in abstract form and, as such, provided minimal methodological detail and modest statistical support for such a configuration. Specifically, the study did not provide any details regarding potential bias or problems with interband limits of agreement, the reporting of which has become standard when comparing different techniques (Bland & Altman, 1986). The objective of the current study was to compare an altered three-quarter circumferential partial band configuration (PB) with the original full circumferential band configuration (FB) on hemodynamic variables at rest and in response to a math task and a cold pressor test. In addition, we assessed subjective comfort levels of participants with each configuration. It was hypothesized that measures derived from the PB would have a high level of correlation and agreement with measures obtained from the FB. It was also hypothesized that the PB would be more comfortable and would not feel as restrictive to participants compared to the FB.

Methods Participants A total of 47 participants were recruited from students attending the University of Calgary (66% female, mean [SD] age 5 20.4 [3.0]) years; see Table 1 for full demographics), via the Department of Psychology’s Undergraduate Research Pool. Students received course credit in exchange for their participation. Participants were excluded if they currently used analgesics or antihypertensive medications, had arrhythmias or known cardiovascular disease, or were currently pregnant, all of which were assessed by a standard health history questionnaire. Current

S.L. Bacon et al. Table 1. Participant Demographics for the 47 Participants n (%) or M (SD) Age (years) Sex (male) Ethnicity White Asian Handedness (right) BMI (kg/m2) Resting SBP (mmHg) Resting DBP (mmHg) Smoking status (current) Asthma (yes)

20.4 (3.0) 16 (34%) 33 (70%) 33 (70%) 14 (30%) 42 (89%) 23.06 (0.09) 115.0 (8.3) 69.0 (5.3) 4 (9%) 7 (15%)

smokers were asked to refrain from smoking for 4 h prior to testing. All participants provided written and informed consent, and the study was approved by the University of Calgary Conjoint Faculties Research Ethics Board. Physiological Measures The impedance cardiogram was recorded from each participant during two identical testing protocols, described below, one using FB and one using an alternate PB, the order of which was randomized. The FB utilized the standard tetra-polar configuration (Kubicek et al., 1966; Sherwood et al., 1990). Disposable mylar band electrodes (T8001 Instrumentation for Medicine, Inc., Greenwich, CT) were applied circumferentially with one voltage electrode around the base of the neck and one encircling the thorax over the tip of the xiphoid process. One current electrode was placed around the neck and one placed around the thorax at a minimum distance of 3 cm above and below the upper and lower voltage electrodes respectably (see Figure 1a). The distance between the recording electrodes was defined as the mean of the minimum distances between the electrically conductive aluminum portions of the voltage electrodes as measured over the sternum and over the spine. For the PB (Figure 1b), four adhesive silver–silver chloride electrocardiogram (ECG) electrodes (105–2249, Stat Healthcare Corp., Calgary, AB) were placed on the back side of the participant over sites prepared by prior application of rubbing alcohol, followed by ECG gel. The two voltage electrodes were placed on the neck over the sixth cervical vertebrae and on the back over the seventh thoracic vertebrae. The two current electrodes were placed one over the fourth cervical vertebrae and one over the ninth thoracic vertebrae. Mylar band electrodes that spanned three quarters of the circumference of the participant’s neck or thorax were attached to the ECG electrodes. The distance between the recording electrodes was defined as the minimum distance between the conductive portions of the two voltage silver–silver chloride electrodes. Impedance cardiography signals (500 samples per second) were converted into digital signals using a Bio-electric Impedance Cardiograph (HIC-2000, Bio-Impedance Technology, Chapel Hill, NC). The impedance cardiographs were imported to a standard PC (IBM). Recording and analysis were carried out using the Cardiac Output Program (Bio-Impedance Technology, Chapel Hill, NC). Three ECG electrodes were used to independently measure the electrocardiogram signal. Two electrodes were placed bilaterally on the upper rib cage, with a ground electrode placed on the right hipbone. The ECG, basal thoracic impedance (Z0), and first derivative of the impedance signal (dZ/

Alternate impedance band configuration

a Full band

Ie

1089 peripheral resistance (TPR; d-s/cm5) and cardiac output (CO; liters/minute) were calculated using standard equations (Bacon et al., 2006; Sherwood et al., 1990).

Ve

Ve Ie

b Partial band

Ie Ve

Ve Ie Figure 1. Electrode placements for the full band configuration (a) and the three-quarter band configuration (b). Voltage (Ve) and current (Ie) electrodes and the position of the spot electrodes ( ! ) are indicated on the figures.

dt) were sampled and processed over designated 60-s sample periods. The software system uses the ensemble averaging procedure to filter respiratory and motion artifact from the impedance cardiogram and identifies the B point, which occurs at the onset of the rapid upslope of dZ/dt, via a standard algorithm (Sherwood, Allen, Hutcheson, & Obrist, 1986). All waveforms were examined and verified or edited using a postacquisition interactive-graphics editing facility. These recordings provided measurements of heart rate (HR; in beats per minute [bpm]), preejection period (PEP; in milliseconds, calculated from Q wave onset), and stroke volume (SV; in milliliters, computed by the COP-WIN program using the Kubichek equation: SV 5 (r(dZ/ dt)(L2)(LVET))/Z20). The algorithm used by the COP-WIN software essentially employs the same protocol as that required for the parameters of the Kubichek equation. The Q point is defined by the software as the onset of the Q wave, just before the Q-wave trough. The B wave is the point where the dZ/dt waveform crosses the zero axis between the Q wave and the peak of the dZ/ dt wave. The X-wave point is the maximum trough of the dZ/dt wave following both the QRS complex of the ECG waveform and the peak of the dZ/dt waveform based upon the first derivative of both the dZ/dt waveform and the ECG waveform and corresponding to the second heart sound. Measurements of systolic blood pressure (SBP; in millimeters of mercury [mm Hg]) and diastolic blood pressure (DBP; in mmHg) were obtained using an automatic blood pressure monitor (Accutor Plus, Data Scope, Mont Vale, NJ) and a blood pressure cuff on the upper part of the nondominant arm. From these measures both total

Procedure Design The present study used a randomized, repeated measures design. Each testing session lasted for approximately 2 h. Prior to testing, participants’ height and weight were measured. All stress tasks were conducted in a temperature-controlled (211C) room. The order in which the participants were instrumented with each of the two configurations was randomly established prior to testing and counterbalanced. Following instrumentation of the initial impedance cardiography configuration, participants were instructed to sit quietly for a 5-min baseline rest period. Following the baseline period, participants engaged in two 5-min laboratory stress tasks, separated by a 5-min recovery period. Impedance and electrocardiograph measurements were recorded continuously during baseline, during each task, and during the subsequent recovery periods. Blood pressure was measured at 1min intervals throughout the test session. Once one session was completed, participants were then reinstrumented with the alternate band configuration and the battery of tests was repeated. Stress Tasks Math task. Participants were presented with a series of mathematical subtraction equations with the answers presented on a computer screen. They were asked to determine whether the answer provided for each equation was correct or incorrect. A researcher informed the participants that they were being judged on their response time and accuracy. Each correct response was followed by a beep emitted from a speaker located beneath the participant’s chair. Each incorrect response was followed by a noxious blare emitted from the same speaker. The program was designed to increase or decrease the difficulty of the next equations based on the participant’s performance on the current equation, resulting in a constant correct score of approximately 60%. Cold pressor test. Participants were instructed to immerse their right hand, up to the wrist fold, into approximately 41C circulating ice water for up to 5 min. They were required to keep their hand motionless and fingers spread apart for the duration of the test. Subjective Discomfort Previous research has noted anecdotal reports of discomfort and distress associated with the use of the full-circumferential band configuration. Using standard visual analog scales (from 0 5 not at all to 10 5 extremely), participants were asked to rate six dimensions of their experience for each of the configurations. These scales were administered at the end of each of the testing sessions. These included the following questions: (1) How much tension or constriction do you feel around your neck? (2) How do the bands on your neck affect your ability to breathe? (3) How much tension or constriction do you feel around your rib cage? (4) How do the bands around your rib cage affect your ability to breathe? (5) Overall, how uncomfortable are you right now? (6) How anxious do you feel right now? Data Reduction and Analyses The 1-min ensemble-averaged data for impedance generated data were averaged across each 5-min condition (baseline, math

1090

S.L. Bacon et al.

Table 2. Absolute Levels of Cardiovascular Variables at Baseline and during the Math Task and the Cold Pressor Test for Both Full and Partial Band Sessions Baseline

PEP (ms) SV (ml) CO (l/min) HR (bmp) SBP (mmHg) DBP (mmHg) TPR (d-s/cm5)

Math

Cold pressor

Full band

Partial band

Full band

Partial band

Full band

Partial band

123.04 (11.48) 83.99 (25.16) 5.76 (1.42) 66.75 (9.56) 116.64 (8.70) 70.26 (6.15) 1258.26 (315.31)

124.85 (12.63) 88.28 (25.34) 5.42 (1.26) 66.28 (9.42) 115.27 (8.46) 69.09 (6.15) 1326.80 (367.66)

121.75 (11.93) 84.03 (23.39) 5.89 (1.37) 70.76 (10.19) 119.28 (9.06) 72.47 (6.11) 1259.94 (313.37)

123.16 (12.55) 84.97 (22.27) 5.68 (1.22) 69.17 (9.32) 117.47 (9.30) 72.50 (6.36) 1296.51 (317.18)

117.90 (10.82) 78.30 (20.55) 5.62 (1.33) 76.15 (10.70) 131.95 (10.12) 85.06 (7.46) 1509.14 (331.91)

118.72 (11.99) 76.19 (23.37) 5.78 (1.34) 75.33 (9.31) 130.59 (9.64) 83.63 (7.03) 1461.85 (370.96)

Table 3. Main Effects from the ANOVA Models Order of presentationa

PEP SV CO HR SBP DBP TPR

Band configurationa

Stress

F

p

F

p

F

df

p

0.39 0.37 0.23 0.01 0.02 0.05 0.14

.845 .548 .635 .999 .888 .831 .706

3.29 0.35 0.88 3.38 15.63 5.27 0.27

.077 .560 .353 .072 o.001 .026 .609

25.26 21.24 3.37 61.92 256.02 200.97 58.21

1.72,77.22 1.35,60.93 1.37,61.47 1.41,63.24 1.51,68.10 1.34,60.38 1.36,61.11

o.001 o.001 .059 o.001 o.001 o.001 o.001

a

df 5 1,45.

task, recovery, and cold pressor task). Similarly, blood pressure measurements were used to calculate average SBP and DBP for each condition. Recorded minute-by-minute data for both mean arterial pressure and cardiac output were used in the calculation of total peripheral resistance. To evaluate mean differences for each cardiovascular measure across electrode configurations, a series of repeated measures analyses of variance (ANOVAs) were conducted. The models consisted of two within-subject factors: configuration (FB, PB) and stress (baseline, math task, cold pressor) and one betweensubject factor: order of presentation (FB then PB or PB then FB). Because of the likelihood of a lack of sphericity in the data as a result of multiple recordings taken over closely spaced intervals in time (Jennings, 1987; Keselman, 1998), the Greenhouse–Geisser correction was employed. The intraclass correlation analysis, using a two-way mixed model assessing consistency of a single measure with the SPSS reliability analysis algorithm, were calculated for all cardiovascular measures at baseline and during both tasks to determine the strength of association between FB and PB configurations. Bias, which is defined as the absolute difference between comparative measures (in this case FB vs. PB), and the limits of agreement, defined as the variability in the difference between the comparative measures, between the FB and the PB configurations were assessed using a series of Bland-Altman analyses (Bland & Altman, 1986). Variables were transformed to a percentage of the mean of each variable using the following equation: ((FB ! PB)/((FB1PB)/2)) " 100 (Bernstein & Lemmens, 2005; Critchley & Critchley, 1999) for these analyses. Data analyses were conducted using SPSS (Version 15.0, SPSS Inc., Chicago, IL). Significance for all tests was set at the .05 alpha level.

Table 4. Intraclass Correlations for Absolute Levels of Baseline, Math Task, and Cold Pressor Test Baseline

Math

Cold Pressor

.89 .80 .71 .80 .93 .70 .63

.86 .83 .75 .73 .86 .76 .71

.79 .73 .67 .84 .91 .71 .71

PEP SV CO HR SBP DBP TPR Note: All pso.001.

Results ANOVA Analyses Mean values for each variable (HR, SBP, DBP, SV, CO, TPR, and PEP) recorded during the FB and PB sessions are presented in Table 2. As described in Table 3, there were no main effects for order of presentation for any of the measures. In contrast, there was a main effect of stress for all the measures, except CO. Post hoc analyses revealed that both the math and the cold pressor tasks increased HR, SBP, DBP, CO, and TPR and decreased SV and PEP (all pso.001 except the SV response to the math task where p 5 .046). Overall, the changes were more substantial during the cold pressor task. There was no main effect of band configuration on PEP, SV, CO, HR, or TPR. However, there was a significant difference between band configuration type for SBP and DBP, where both SBP and DBP were lower when the participant was instrumented with the PB configuration compared with the FB configuration.

Alternate impedance band configuration

10.00

a

Baseline

5.00 0.00 –5.00 –10.00 –15.00 100.00

b

110.00

120.00 130.00 140.00 Average PEP baseline

150.00

Math

b

10.00

Full - partial % bias in math CO

Full - partial % bias in math PEP

40.00 30.00 20.00 10.00 0.00 –10.00 –20.00 –30.00 4.00

5.00 6.00 Average baseline CO

7.00

8.00

Math 50.00

5.00 0.00 –5.00 –10.00 –15.00 –20.00 100.00

110.00

120.00

130.00

140.00

150.00

20.00 15.00 10.00 5.00 0.00 –5.00 –10.00 –15.00 –20.00 90.00 100.00 110.00 120.00 130.00 140.00 150.00 160.00 Average cold pressor PEP

Figure 2. Bland-Altman plots for preejection period (PEP) at baseline (a) and during the math task (b) and the cold pressor test (c).

Intraclass Correlations As reported in Table 4, there were positive correlations between PB and FB values for all measures for baseline levels (ricc range 5 .63–.93), math task levels (ricc range 5 .71–.86), and cold pressor test levels (ricc range 5 .67–.91). Bland-Altman Analyses The full results of the Band-Altman analyses are presented in Table 5. All measures had a bias within 5% between the two band configurations, indicating that the group averages were similar between the two configurations. In a meta-analysis of methods to

c Full - partial % bias in cold pressor CO

Cold pressor 25.00

40.00 30.00 20.00 10.00 0.00 –10.00 –20.00 –30.00 –40.00 –50.00 3.00

160.00

Average math PEP

Full - partial % bias in cold pressor PEP

50.00

–40.00 3.00

160.00

15.00

c

Baseline 60.00

Full - partial % bias in baseline CO

Full - partial % bias in baseline PEP

a

1091

4.00

5.00

6.00 7.00 8.00 Average math CO

9.00

10.00

5.00 6.00 7.00 8.00 Average cold pressor CO

9.00

10.00

Cold pressor 50.00 40.00 30.00 20.00 10.00 0.00 –10.00 –20.00 –30.00 –40.00 –50.00 3.00

4.00

Figure 3. Bland-Altman plots for cardiac output (CO) at baseline (a) and during the math task (b) and the cold pressor test (c).

assess CO, a range of limits of agreement between ! 30% had been proposed to define an acceptable level of agreement to indicate equivalence of a new technique to the current reference method, that is, PB versus FB (Critchley & Critchley, 1999). Using this definition, PEP (Figure 2) fell within these limits and SV, CO (Figure 3), and TPR were close to reaching this level (Table 5). Given that the two configurations were not assessed at the same time, which may lead to differences in the two results because of placement issues, we also assessed three non-banddependent measures (i.e., measures that are not derived from the impedance measures). The rationale for this was to provide

1092

S.L. Bacon et al.

Table 5. Bland-Altman Analyses for Absolute Levels during Baseline, the Math Task, and the Cold Pressor Test Baseline

Math

95% limits of agreement

PEP SV CO HR SBP DBP TPR

Cold pressor

95% limits of agreement

95% limits of agreement

% bias

% upper

% lower

% bias

% upper

% lower

% bias

% upper

% lower

" 1.36 3.02 5.06 " 0.86 1.10 1.46 " 3.74

7.63 40.70 42.62 16.72 6.54 14.89 34.47

" 10.34 " 34.66 " 32.51 " 18.43 " 4.34 " 11.98 " 41.95

" 0.89 1.96 2.40 1.45 1.15 " 0.12 " 1.95

9.45 33.93 34.24 20.76 9.29 11.55 30.77

" 11.23 " 30.01 " 29.44 " 17.86 " 6.98 " 11.79 " 34.68

" 0.73 " 3.42 " 3.53 0.01 0.45 1.14 4.37

12.27 31.04 30.48 15.77 7.00 14.14 37.55

" 13.73 " 37.89 " 37.53 " 15.75 " 6.10 " 11.86 " 28.82

Note: Bias calculated as the difference as percentage of mean: (FB " PB)/((FB1BP)/2) # 100.

comparative data on the ‘‘natural’’ deviations between setups. HR, SBP, and DBP all had a low level of bias and fell within the ! 30% range for the limits of agreement. Subjective Discomfort: Visual Analog Scale As shown in Table 6, participants reported that, relative to the FB configuration, the PB configuration was associated with less tension or constriction around the neck and thorax, less interference with their ability to breathe, and greater overall comfort. In contrast, participants reported no difference in subjective level of anxiety when instrumented with PB versus FB. When asked to indicate their preferred configuration, 38 (81%) of the participants preferred the PB setup. Discussion The current study was designed to evaluate if a three-quarter partial band configuration was equivalent to a full band configuration for the assessment of impedance cardiography in the context of a typical laboratory stress paradigm. Furthermore, the current study also assessed the subjective comfort/discomfort that participants experienced with each of the electrode configurations. There seemed to be a high level of correlation between measures obtained using the two different configurations with minimal bias. Though PEP fell within acceptable limits of agreement between the FB and PB, other variables, that is, SVand CO, were just outside the boundaries of acceptable levels. There was a clear preference for the PB configuration by the participants, with the PB configuration being associated with less constriction around their neck and thorax, less impact on breathing, and overall greater comfort. In addition to collecting and reporting impedance-generated data, we also reported levels of HR, SBP, and DBP for the two

band configurations. Given the independence of these measures and the consistency of the tasks performed, we expected to see no difference in these variables. However, SBP and DBP values were significantly higher for the FB compared to the BP. Although only speculative, it is possible that these higher values may reflect the greater discomfort experienced by participants during the FB compared to the PB. This is consistent with previous suggestions that the FB configurations may create artifactual readings (Wilmers & Brener, 1991). As such, the lack of agreement between the configurations for FB and BP for SV and CO might actually reflect some degree of difference in discomfort and not solely measurement error. Limitations Though the current study has many strengths, such as the repeated measures design, the use of commonly used stress tasks, and the inclusion of both men and women, the results need to be interpreted in light of a certain limitations. Most notably, the current comparisons are made between two different band configurations rather than against an external measure, for example, thermodilution or the direct fick technique. Given the generally positive results obtained in the current study, this comparison would be the logical next step. The current study also only compared two configurations. No comparisons were made between shorter band configurations or against spot configurations. However, previous studies have found that band configurations less than three-quarter circumferential length tend to deviate from full band configurations (Wilmers & Brener, 1991), and there is still much debate over the optimal spot electrode configuration (Hoetink et al., 2002; Ikarashi, Nogawa, Tanaka, & Yamakoshi, 2006, 2007), making it difficult to find an appropriate comparison configuration. Given the nature of the sample used (generally healthy university students), the current results

Table 6. Subjective Discomfort Data for Each Question from the Visual Analog Scale Questionnaire Questiona Neck tension Neck breathing Rib cage tension Rib cage breathing Uncomfortable Anxious

Mean millimeters (SD) for FB

Mean millimeters (SD) for PB

F

P

Z2

49 (23) 20 (18) 36 (25) 23 (22) 31 (20) 14 (14)

27 (19) 7 (9) 18 (20) 9 (13) 20 (19) 13 (14)

54.61 28.03 25.78 21.55 16.77 0.84

o.001 o.001 o.001 o.001 o.001 .365

.55 .38 .36 .32 .27 .02

Note: FB 5 Full band configuration; PB 5 partial band configuration. a Lower score indicate less discomfort.

Alternate impedance band configuration

1093

may not be generalizable to nonstudent populations. However, it would be expected that issues related to comfort might be more pronounced in older and sicker populations, thus making the PB configuration more attractive. Finally, the current study assessed the different configurations over the course of two separate testing sessions, which may have introduced some measurement error. Unfortunately, because of the nature of the band configurations, it was not possible to test both configurations at the same time, as they occupy the same positions on the body.

Conclusion The current study provides incremental evidence suggesting that a three-quarter PB configuration may be used for standard psychophysiological testing instead of the standard FB configuration. However, researchers need to balance potential variability with SV and CO measures with increased participant comfort. Further studies are needed to validate the PB configuration against other techniques and to ensure that these results are consistent in older and diseased populations.

REFERENCES Bacon, S. L., Ring, C., Li Saw Hee, F., Lip, G. Y. H., Blann, A. D., Lavoie, K. L., et al. (2006). Hemodynamic, hemostatic, and endothelial reactions to psychological and physical stress in coronary artery disease patients. Biological Psychology, 72, 162–170. Bernstein, D. P., & Lemmens, H. J. M. (2005). Stroke volume equation for impedance cardiography. Medical & Biological Engineering & Computing, 43, 443–450. Bland, J. M., & Altman, D. G. (1986). Statistical methods for assessing agreement between two methods of clinical measurement. Lancet, 327, 307–310. Critchley, L. A., & Critchley, J. A. (1999). A meta-analysis of studies using bias and precision statistics to compare cardiac output measurement techniques. Journal of Clinical Monitoring & Computing, 15, 85–91. Doering, L., Lum, E., Dracup, K., & Friedman, A. (1995). Predictors of between-method differences in cardiac output measurement using thoracic electrical bioimpedance and thermodilution. Critical Care Medicine, 23, 1667–1673. Gotshall, R. W., & Sexson, W. R. (1994). Comparison of band and spot electrodes for the measurement of stroke volume by the bioelectric impedance technique. Critical Care Medicine, 22, 420–425. Hoetink, A. E., Faes, T. J. C., Schuur, E. H., Gorkink, R., Goovaerts, H. G., Meijer, J. H., et al. (2002). Comparing spot electrode arrangements for electric impedance cardiography. Physiological Measurement, 23, 457–467. Ikarashi, A., Nogawa, M., Tanaka, S., & Yamakoshi, K.-i. (2007). Experimental and numerical study on optimal spot-electrodes arrays in transthoracic electrical impedance cardiography. Conference Proceedings: Annual International Conference of the IEEE Engineering in Medicine & Biology Society, 2007, 4580–4583. Ikarashi, A., Nogawa, M., Yamakoshi, T., Tanaka, S., & Yamakoshi, K.-i. (2006). An optimal spot-electrodes array for electrical impedance cardiography through determination of impedance mapping of a regional area along the medial line on the thorax. Conference Proceedings: Annual International Conference of the IEEE Engineering in Medicine & Biology Society, 1, 3202–3205. Jennings, J. R. (1987). Editorial policy on analyses of variance with repeated measures. Psychophysiology, 24, 474–475. Keselman, H. J. (1998). Testing treatment effects in repeated measures designs: An update for psychophysiological researchers. Psychophysiology, 35, 470–478. Kubicek, W. G., Karnegis, J. N., Patterson, R. P., Witsoe, D. A., & Mattson, R. H. (1966). Development and evaluation of an impedance cardiac output system. Aerospace Medicine, 37, 1208–1212. McGrath, J. J., O’Brien, W. H., Hassinger, H. J., & Shah, P. (2005). Comparability of spot versus band electrodes for impedance cardiography. Journal of Psychophysiology, 19, 195–203.

Nyboer, J. (1959). Electrical impedance plethysmography. Springfield, IL: Thomas Publishers. Nyboer, J., Bagno, S., Barnett, A., & Halsey, R. H. (1940). Radiocardiograms: Electrical impedance changes in the heart in relation to electrocardiograms and heart sounds. Journal of Clinical Investigation, 2, 263–270. Penney, B. C., Patwardhan, N. A., & Wheeler, H. B. (1985). Simplified electrode array for impedance cardiography. Medical & Biological Engineering & Computing, 23, 1–7. Qu, M. H., Zhang, Y. J., Webster, J. G., & Tompkins, W. J. (1986). Motion artifact from spot and band electrodes during impedance cardiography. IEEE Transactions on Biomedical Engineering, 33, 1029–1036. Ring, C., Liu, X., & Brener, J. (1994). Cardiac stimulus and heartbeat detection: Effects of tilt-induced changes in stroke volume. Psychophysiology, 31, 553–564. Sherwood, A., Allen, M. T., Fahrenberg, J., Kelsey, R. M., Lovallo, W. R., & van Doornen, L. J. (1990). Methodological guidelines for impedance cardiography. Psychophysiology, 27, 1–23. Sherwood, A., Allen, M. T., Hutcheson, J. S., & Obrist, P. A. (1986). Ensemble averaging of the impedance cardiogram [Abstract]. Psychophysiology, 23, 461. Sherwood, A., Royal, S. A., Hutcheson, J. S., & Turner, J. R. (1992). Comparison of impedance cardiographic measurements using band and spot electrodes. Psychophysiology, 29, 734–741. Wang, Y., Haynor, D. R., & Kim, Y. (2001). A finite-element study of the effects of electrode position on the measured impedance change in impedance cardiography. IEEE Transactions on Biomedical Engineering, 48, 1390–1401. Watanabe, T., Kamide, T., Torii, Y., & Ochiai, M. (1981). A convenient measurement of cardiac output by half-taped impedance cardiography. Japanese Journal of Medical Electronics and Biological Engineering, 19, 30–34. Wilmers, F. E., & Brener, J. M. (1991). Test of alternative band electrode arrays for impedance cardiography (ZKG). Psychophysiology, 28, S62. Woltjer, H. H., van der Meer, B. J., Bogaard, H. J., & de Vries, P. M. (1995). Comparison between spot and band electrodes and between two equations for calculations of stroke volume by means of impedance cardiography. Medical & Biological Engineering & Computing, 33, 330–334.

(Received February 18, 2009; Accepted October 4, 2009)

Psychophysiology, 47 (2010), 1094–1101. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01011.x

Ethnicity, gender, genotype, and anger as related to nocturnal dipping

GEORGE D. BISHOP, DANIEL P.K. NG, FRANCIS NGAU, and SITI NURBAYA Department of Psychology, National University of Singapore, Singapore

Abstract Bishop, Pek, and Ngau (2005) found a significant interaction in Singapore between anger and nocturnal dipping among Indians but not Chinese and Malays. The current study examines the role of 5-HTTLPR genotype in this relationship. Two hundred thirty-one undergraduates participated in up to 4 days of 24-h ambulatory monitoring, completed the State-Trait Anger Expression Inventory, and provided blood samples for genotyping of 5-HTTLPR. Results indicated individuals with two copies of the short allele (SS) showed reduced dipping when they were high in Outward Anger (OA) but increased dipping when they were low in OA. Further, for Indian men only, dipping was reduced for individuals having the SS genotype when they were low on Anger In and increased when they were high on Anger In. These data provide further evidence for the role of 5-HTTLPR in cardiovascular risk as well as ethnic differences in the 5-HTTLPR–phenotype relationship. They also provide further evidence for 5-HTTLPR as a ‘‘plasticity gene.’’ Descriptors: Nocturnal dipping, 5-HTTLPR, Ethnicity, Anger, Singapore, Plasticity

the relationship between anger and dipping may vary by race. Bishop, Pek, and Ngau (2005) found that higher levels of trait anger were associated with reduced blood pressure dipping for Indians in Singapore whereas this was not true for Chinese and Malays. This reduced dipping appeared to be the result of continued vasoconstriction among nondippers during the night. The reduced blood pressure dipping among Indians high in trait anger is particularly interesting in light of the fact that Indians are at significantly higher risk of cardiovascular disease (CVD) as compared with Chinese and Malays in Singapore (Hughes, Lun, & Yeo, 1990). Indeed, high rates of CVD have been found for South Asians relative to other groups in the United Kingdom (Marmot, Adelstein, & Bulusu, 1984), South Africa (Walker, 1980), Trinidad (Miller, Beckles, Alexis, Byam, & Price, 1982), and Canada (Anand et al., 2000). The present study was undertaken to further examine the interaction of race with dispositional anger and to examine the potential role of the serotonin transporter gene (5-HTTLPR) in this relationship. As such, the first purpose of the current study was to replicate the findings of Bishop et al. (2005) showing reduced SPB dipping for Indians with high trait anger using a larger sample that underwent multiple 24-h periods of BP monitoring. In line with this, we hypothesized that Indians would show reduced BP dipping when they were high in trait anger. The second purpose of this study was to explore the role of 5HTTLPR in the relationship of BP dipping with race and trait anger. Although there is now evidence that blood pressure dipping is a heritable trait (Fava et al., 2005), there has been little research on the specific genetic markers involved with what little research that has been done focusing on the angiotensin

Nocturnal blood pressure (BP) dipping is a well-established phenomenon related to the diurnal cycle of blood pressure and related functions (O’Brien, Sheridan, & O’Malley, 1988). Research evidence points to marked individual differences in the extent of BP dipping (Staessen et al., 1997), and there is accumulating evidence that these individual differences are predictive of cardiovascular risk. Specifically, individuals showing less than a 10% reduction of systolic blood pressure (SBP) from daytime to nighttime have been found to be at elevated risk for hypertension (Timio et al., 1995), left ventricular hypertrophy (Verdecchia et al., 1995), myocardial infarction, heart failure, stroke, and sudden death (Staessen et al., 1999) as well as overall cardiovascular mortality (Ohkubo et al., 1997; Palatini et al., 1992; Staessen et al., 1999). There is also evidence for racial differences in nocturnal dipping that parallel differences in cardiovascular risk. Several studies have found that African Americans show a reduction in nocturnal dipping that parallels their higher rates of hypertension (Fumo et al., 1992; Herbert et al., 1996; Ituarte, Kamarck, Thompson, & Bacanu, 1999). Further, there is evidence that psychosocial variables such as anger and hostility play a role in reduced dipping (Thomas, Nelesen, & Dimsdale, 2004) and that

This research was supported by grants RP-581-000-041-112 and RP581-000-054-112 from the National University of Singapore Academic Research Fund and RP-581- 000-083-750 from the NUS Faculty of Arts and Social Sciences. Address reprint requests to: George D. Bishop, Ph.D., Department of Psychology, National University of Singapore, 9 Arts Link, Singapore 117570. E-mail: [email protected] 1094

5-HTTLPR and nocturnal dipping converting enzyme (ACE) gene insertion/deletion polymorphism (Czupryniak et al., 2008; Spiering, Zwaan, Kroon, & de Leeuw, 2005). The findings reported here are from a larger study of the role of serotonergic genes in emotion and cardiovascular reactivity. The serotonin system plays a key role in emotions as well as in emotion-related disorders (Murphy et al., 2008), and there is growing evidence for the role of 5-HTTLPR in cardiovascular responses to stress (Williams et al., 2003, 2008). Further, there is emerging evidence that 5-HTTLPR is associated with sleep quality (Brummett et al., 2007), with racial differences in sleep quality shown to accompany racial differences in nocturnal dipping (Hughes, Kobayashi, & Deichert, 2007). On this basis, there appeared to be reason to investigate the role that 5-HTTLPR might play in nocturnal dipping. 5-HTTLPR is a 44-base pair insertion/deletion polymorphism in the promoter region of the serotonin transporter gene. Two variants of this gene have been identified, the ‘‘short’’ (S) and ‘‘long’’ (L) alleles, with the L allele associated with increased transcriptional efficiency and lower neuroticism in Caucasians (Lesch et al., 1996). With each person having two copies of the gene, this results in three genotype groups (SS, LS, and LL). In this study, we examined differences in nocturnal dipping as a function of genotype group.

Methods Participants The total sample consisted of 328 undergraduates (93 Indians, 154 Chinese, 81 Malays; 49.7% female) at the National University of Singapore. Of these, 74 did not have genotyping data and 5 lacked valid State-Trait Anger Expression Inventory (STAXI) scores. To ensure high quality data, to be included in the analyses for any 24-h period, participants were required to have gone to bed between 10 p.m. and 6 a.m., slept at least 4 hours, and have at least six valid daytime blood pressures and three valid night time readings. Also because daytime posture was used as a covariate, participants were required to have completed at least six diaries during waking hours. An additional 18 participants who did not meet these criteria for at least one monitoring period were eliminated from the analyses, resulting in a final sample size of 231 (62 Indians, 116 Chinese, and 53 Malays; 51.9% female). Age ranged from 18 to 27 years (M 5 21.1). Genotyping Genotyping was done from blood samples using procedures described by Lesch et al. (1996). Following DNA extraction, the 5HTT-linked polymorphism was amplified from genomic DNA as a 484/528bp polymerase chain reaction (PCR) product. PCR amplification was carried out with primers 5 0 -GGCGTTG CCGCTCTGAATTGC-3 0 (forward) and 5 0 -GAGGGACTGA GCTGGACAACCCAC-3 0 (reverse). In the total sample, 36 (13.9%) participants were LL, 115 (44.4%) were LS, and 108 (41.7%) were SS. For the final sample, the distribution was 31 (13.6%) LL, 102 (44.2%) LS, and 98 (42.4%) SS. Analysis of these distributions showed that the allele frequencies did not significantly depart from Hardy-Weinberg equilibrium and were equivalent between ethnic groups. Given the rarity of the LL genotype and that fact that preliminary analyses suggested no readily discernable allele dominance pattern, individuals with the LL genotype were not included in analyses.

1095 Measurement of Anger The STAXI (Spielberger, 1988; Spielberger et al., 1985) was used to measure dispositional anger. The STAXI is a 44-item inventory with five subscales: State Anger, Trait Anger (TA), Anger In (AI), Anger Out (AO), and Anger Control (AC). As we were interested in dispositional anger, we used only the TA, AI, AO, and AC subscales. An initial principal components analysis of these scales indicated two components accounting for 79.5% of the variance. Varimax rotation of these components showed high loadings on the first component for TA (.78), AO (.84), and AC ( ! .72). These three subscales were combined into an index labeled Outward Anger (OA) by first taking the z scores for each component and then averaging them after reversing the z scores for AC. The second component consisted of only AI (.96) and was analyzed separately. Sleep Questionnaire To obtain information on sleeping times and quality of sleep upon returning to the laboratoty after each 24-h monitoring period, participants were asked to complete a questionnaire that asked what time they went to bed the night before, the time they got up that morning, whether they had trouble falling asleep, if they indicated trouble falling asleep how long it took them to fall asleep, how many times they woke up in the night, when they woke up did they get out of bed, when they woke up during the night did they have difficulty going back to sleep, and whether they felt rested on waking in the morning. The question on difficulty falling asleep was combined with the time taken to get to sleep and coded as 0 5 No or o15 min to get to sleep, 1 5 Yes, 15– 30 mins, 3 5 Yes, 31–60 min, 4 5 Yes, 460 min. Waking up in the night was coded as 0 5 No, 1 5 1–2 times, 2 5 3–4 times, and 3 5 44 times. Getting out of bed was coded as 0 5 Did not wake up during the night, 1 5 Woke up but did not get out of bed, 2 5 Got out of bed 1–2 times, 3 5 Got out of bed42 times. Difficulty getting back to sleep was coded as 0 5 Did not wake up, 1 5 Woke up but no difficulty going back to sleep, 2 5 Difficulty getting back to sleep. Finally the question on feeling rested on awaking was coded as 0 5 No, 1 5 Yes. Blood Pressure Measurement Participants underwent 24-h BP monitoring for up to 4 days using Spacelabs 90217 ambulatory blood pressure monitors (Spacelabs Medical, Redmond, WA). The Spacelabs 90217 is a lightweight ambulatory blood pressure monitor using the oscillometric method for BP determination. SBP and DBP were obtained every 30 min. When the monitor was unable to obtain a valid reading, a second attempt was made 2 min later. Participants were told to avoid movement if possible during the time the blood pressure cuff was inflated so as to reduce movement artifact. Participants also wore an AIM-8 ambulatory impedance monitor (Bio-Impedance Technology, Chapel Hill, NC) in this study. However, because of technical problems with the AIM-8 and data produced by it, data from the AIM-8 are not reported here. Procedures Study procedures were approved by the National University of Singapore Institutional Review Board. Participants reported to a psychophysiology laboratory, where they were briefed on the procedures for the ambulatory monitoring. After the procedures of the study were explained, participants signed an informed consent form. Participants were then instrumented with the am-

1096 bulatory monitors and given instruction on the use of the Palm Zire palmtop computer for filling out a computerized diary that they were to complete after each waking BP measurement. As part of their orientation to the ambulatory monitoring, participants were instructed that when the blood pressure cuff began to inflate they should try to move as little as possible until the blood pressure cuff was completely deflated. They were also instructed to begin filling out the questionnaire on the palmtop computer only once the blood pressure cuff had completely deflated. This was done to reduce arm movements that might interfere with the blood pressure readings. Participants were instructed not to get the equipment wet and were provided with written instructions to remind them of various aspects of the use of the palmtop computer along with the researcher’s contact number should there be any problems. Finally, the participant was given an appointment for returning approximately 24 h later. To encourage filling out the diaries, participants were paid on a graduated scale according to their cooperation in completing the diaries. For wearing the monitors and completing up to 50% of the diaries they were paid S$10 (US$6.06) for each day of monitoring. They were then paid S$2.00 (US$1.21) for each percentage above 50% of diaries completed across all days up to a maximum payment of S$140 (US$84.80). In order to be counted, each diary was required to be started within 5 min of the time recorded for the BP reading. In cases where the first BP reading was inconclusive the 5 min was counted from the time of the second reading, which was automatically initiated 2 min following an inconclusive reading. Data Screening and Reduction The Spacelabs 90217 automatically checks readings for possible artifacts and eliminates those determined to be erroneous. To further test for possible artifacts, the criteria proposed by Marler, Jacob, Lehoczky, and Shapiro (1988) were used to eliminate likely artifactual blood pressure readings. Both SBP and DBP values were excluded from analyses if SBP 4250 mmHg or o70 mmHg, DBP 4150 mmHg or o45 mmHg, or SBP/DBP 43 or o[1.0651(.00125 ! DBP)]. Waking and sleeping hours were determined by asking each participant upon returning the monitor what time she or he had gone to bed the night before and gotten up that morning. To verify sleep times, all diary entries were checked to be sure that they did not occur during a time the participant claimed to be sleeping. In cases where diaries were completed after the participant indicated she or he had gone to sleep or before the time given for rising, sleeping times were adjusted to exclude those times. Altogether, attempts were made to obtain blood pressure measurements for a total of 37,383 time periods, with valid blood pressure readings obtained for 32,965 periods (88.2%). The number of valid blood pressure readings during the day for each participant ranged from 0 to 49, with an average of 30.4. For nighttime readings, the range was 0 to 23, with an average of 9.4. Based on the criteria stated above for amount of sleep, sleep times, and number of valid BP measurements, usable data were obtained for 549 monitoring days from 231 participants for an average of 2.4 monitoring days per participant. Daytime averages for each day for each participant were then obtained by taking the average of readings from the beginning of the monitoring period until the time the participant retired for the night and then those taken from the time the participant got up the next morning until the end of the monitoring period. Sleeping

G.D. Bishop et al. averages for each monitoring day were obtained by taking the average of all readings taken from the time the participant went to bed until getting up the next morning. Percentage of dipping was then computed by subtracting the sleeping average from the daytime average, dividing by the daytime average, and then multiplying by 100. For each participant, the mean of the averages for all days was then obtained. Participants who showed a drop in SBP of 10% or more were defined as dippers, whereas those exhibiting a drop of less than 10% were categorized as nondippers. By this definition, 153 (66.2%) participants were classified as dippers and 78 (33.8%) were classified as nondippers. This criterion is the one most commonly used in the research literature on nocturnal dipping, as individuals showing less than 10% reduction in nocturnal blood pressure have been shown to be at increased cardiovascular risk (O’Brien et al., 1988; Ra¨ikko¨nen et al., 2004; Verdecchia et al., 1995). Statistical Analysis Comparison of dippers and nondippers was accomplished through the use of w2 analysis for categorical dependent variables and t tests for continuous dependent variables with the exception of cardiovascular parameters. For daytime cardiovascular parameters, an analysis of covariance (ANCOVA) was performed using body mass index (BMI) and percentage of time lying down as covariates. For nighttime cardiovascular parameters, an ANCOVA was performed using BMI and daytime values as covariates. The main analyses reported below used percentage of day to night systolic and diastolic dipping (as computed above) as the dependent variables. These data were analyzed using a 3 (Ethnicity) ! 2 (Gender) ! 2 (Genotype) ! Anger factorial ANCOVA model with AI and OA entered as continuous variables in separate analyses. To ensure that results obtained were not biased by differences in demographics, anthropometric variables, daytime activities, or stress, each of these variables was individually tested for their relationship to percentage of systolic dipping in preliminary analyses. Only BMI and percentage of time lying down showed significant relationships with percentage of dipping and were thus used as covariates in the main analyses. In addition, analyses of dipping and nighttime values included daytime values as a covariate, as daytime values showed significant relationships with both dipping and nighttime values. Significant interactions were followed up using simple effects analysis. Results Comparison of Dippers and Nondippers Table 1 shows comparisons between dippers and nondippers. As can be seen from this table, nondippers were more likely to be female (64.1%) than was the case for dippers (45.8%), w2(1, N 5 231) 5 6.97, p 5 .008. Also, dippers spent less time lying down (M 5 9.91% vs. 12.66%), t(229) 5 " 2.07, p 5 .039, Z2p 5 .018, and had higher scores on AC (23.3) than did nondippers (22.0), t(229) 5 " 2.06, p 5 .034, Z2p 5 .018. As expected, dippers and nondippers differed significantly on cardiovascular parameters. Waking values for dippers were significantly higher than those for nondippers for heart rate (78.52 vs. 76.03), F(1,227) 5 4.13, p 5 .043, Z2p 5 .018, SBP (115.62 vs. 112.50), F(1,227) 5 7.60, p 5 .006, Z2p 5 .032, DBP (72.01 vs. 70.15), F(1,227) 5 7.62, p 5 .006, Z2p 5 .032, and mean arterial pressure (MAP; 86.54 vs. 84.27), F(1,227) 5 8.98, p 5 .003, Z2p 5 .038. By

5-HTTLPR and nocturnal dipping

1097

Table 1. Comparison of Dippers and Nondippers Variable Demographics % Indian % Chinese % Malay % Female Age (years) % with family history of CHD or hypertension Genotype % LL % LS % SS Anthropometric Height (m) Weight (kg) Body mass index Ambulatory cardiovascular values Awake heart rate (bpm)a Awake systolic blood pressure (mmHg)a Awake diastolic blood pressure (mmHg)a Awake mean arterial pressure (mmHg)a Sleep heart rate (bmp)b Sleep systolic blood pressure (mmHg)b Sleep diastolic blood pressure (mmHg)b Sleep mean arterial pressure (mmHg)b Waking posture Standing (%) Sitting (%) Lying down (%) Average daytime activity (1–4) Time slept (min) Self-rated stress (1–4) STAXI Trait Anger Anger In Anger Out Anger Control Outward Anger

Dippers (n 5 153) (# 10% drop in SBP) 30.1 47.7 22.2 45.8 21.2 $ 0.15 48.7% 14.4 43.1 42.5 1.68 $ 0.01 59.76 $ 0.91 21.09 $ 0.27

Nondippers (n 5 78) (o10% drop in SBP) 20.5 55.1 24.4 64.1 21.0 $ 0.19 51.6%

p n.s. .008 n.s. n.s.

11.5 46.2 42.3

n.s.

1.66 $ 0.01 57.82 $ 1.28 20.87 $ 0.37

.065 n.s. n.s.

78.52 115.62 72.01 86.54 58.91 98.50 57.68 71.31

$ $ $ $ $ $ $ $

0.71 0.65 0.39 0.44 0.36 0.27 0.25 0.22

76.03 112.50 70.15 84.27 59.74 106.52 62.14 76.90

$ $ $ $ $ $ $ $

1.00 0.92 0.55 0.62 0.51 0.37 0.35 0.31

.043 .006 .006 .003 n.s. o.001 o.001 o.001

22.98 67.11 9.91 1.56 364.22 1.73

$ $ $ $ $ $

0.83 0.92 0.70 0.02 4.89 0.05

20.99 66.36 12.66 1.49 358.28 1.67

$ $ $ $ $ $

1.14 1.45 1.25 0.03 6.87 0.07

n.s. n.s. .058 .058 n.s. n.s.

19.96 18.07 15.24 23.40 " 0.07

$ $ $ $ $

0.36 0.33 0.27 0.38 0.07

20.68 18.20 15.21 21.99 0.07

$ $ $ $ $

0.56 0.43 0.42 0.54 0.09

n.s. n.s. n.s. 034 n.s.

Note: Numbers in table are means and standard errors for all variables except ethnicity and sex, which are percentages. a Values adjusted for body mass index and percentage of time lying down. b Values adjust for body mass index and daytime values.

contrast, sleep values for dippers were lower than those for nondippers for SBP (98.50 vs. 106.52), F(1,227) 5 301.56, po.001, Z2p 5 .571, DBP (57.68 vs. 61.14), F(1,227) 5 108.65, po.001, Z2p 5 .324, and MAP (71.31 vs. 76.90), F(1,227) 5 214.44, po.001, Z2p 5 .486. Main Analyses On the basis of findings by Bishop et al. (2005), we hypothesized a significant interaction between Ethnicity and OA. However, this hypothesized interaction was not obtained for either systolic dipping, F(2,173) 5 0.33, p 5 .7184, or diastolic dipping, F(2,173) 5 1.65, p 5 .1946. As expected, 5-HTTLPR showed a significant relationship to nocturnal dipping. However, this relationship differed by Ethnicity, Gender, and Anger. Significant Ethnicity ! Gender ! Genotype ! AI interactions were obtained for both systolic, F(2,173) 5 3.13, p 5 .0460, Z2p 5 .035, and diastolic, F(2,173) 5 4.78, p 5 .0096, Z2p 5 .052, dipping. For systolic dipping, simple interaction effects analysis showed that the only Ethnicity ! Gender group that showed a significant Genotype ! AI interaction was Indian men, F(1,21) 5 23.39, po.0001, Z2p 5 0.527, with all other groups showing nonsignificant inter-

actions, all ps4.15. The pattern of this interaction, shown in Figure 1, indicates that AI shows a positive relationship with systolic dipping for those with the SS genotype, b 5 0.91, F(1,9) 5 44.59, po.0001, Z2p 5 0.832, but has nonsignificant negative relationship for those with the LS genotype, b 5 " 0.38, F(1,9) 5 2.54, p 5 .1454, Z2p 5 0.220. Examination of daytime and nighttime SBP indicated a significant four-way interaction for nighttime SBP, F(2,174) 5 3.40, p 5 .0357, Z2p 5 0.038. Simple effects analysis showed that the reduced dipping by SS Indian men who are low on AI was due to higher nighttime SBP for these individuals, b 5 " 0.57, F(1,10) 5 17.53, p 5 .0019, Z2p 5 0.637 (see Figure 2). The parallel four-way interaction for daytime SBP was not statistically significant, F(2,174) 5 0.59, p 5 .5545. The pattern for diastolic dipping was slightly different (see Figure 3). As with systolic dipping, the interaction of AI and genotype was significant only for Indian men, F(1,21) 5 8.87, p 5 .0072, Z2p 5 .297. For other groups, this relationship was nonsignificant, all ps4.31. Further, as with systolic dipping, the relationship between AI and dipping was positive for Indian men with the SS genotype, b 5 0.49, F(1,9) 5 7.45, p 5 .0232, Z2p 5 .453. However, different from the results for systolic dip-

1098

G.D. Bishop et al.

25

15 10

p ns

p < .04

p < .0001

5

LS (N=14)

SS (N = 14)

0 Low

High

Percentage Dipping

Percentage Dipping

20

20 15

p < .03

10 5

LS

Anger In 0

Figure 1. Genotype " Anger In interaction for systolic dipping, Indian men only. Percentage dipping refers to the percentage drop in blood pressure from day to night.

ping, the relationship between AI and dipping was significantly negative for those with the LS genotype, b 5 ! 0.54, F(1,9) 5 6.53, p 5 .0310, Z2p 5 .420. Examination of daytime and nighttime DBP levels indicated that, as with systolic dipping, the differences in dipping were because of higher nighttime BP on the part of those showing reduced dipping. In addition to these findings for AI, there is evidence that the relationship of OA with systolic dipping is moderated by genotype, as seen in the significant Genotype " OA interaction, F(1,173) 5 5.72, p 5 .0179, Z2p 5 0.032. The pattern of this interaction is shown in Figure 4. Simple effects analysis indicated that whereas the relationship between OA and SPB dipping was significantly negative for those with the SS genotype, b 5 ! 0.23, F(1,83) 5 6.79, p 5 .0109, Z2p 5 0.076, it was not significant for those with the LS genotype, b 5 0.10, F(1,87) 5 0.90, p 5 .3466, Z2p 5 0.010. Parallel to the Ethnicity " Gender " Genotype " AI interaction, this was because of a positive relationship between OA and nighttime SBP for those with the SS genotype, b 5 0.25, F(1,83) 5 6.60, p 5 .0120, Z2p 5 0.074. To assess the possibility that the differences in dipping patterns, as well as the related differences for nighttime BP, might be related to participants’ quality of sleep, we analyzed responses to the questions on the sleep questionnaires completed by participants when they returned the monitoring equipment. Specifically, we examined participants’ responses to the questions of

110

SBP

105

p < .002

100 95

p ns

90 85

LS

SS

High

Low

Anger In Figure 3. Genotype " Anger In interaction for diastolic dipping, Indian men only. Percentage dipping refers to the percentage drop in blood pressure from day to night.

whether they had difficulty falling asleep, whether they woke up during the night, and, if they woke up, did they get out of bed, whether they had difficulty going back to sleep, and whether they awoke feeling rested. We also looked at their reported sleeping time adjusted for the presence of diaries filled out during the period they indicated they were sleeping. In only a few cases did the effects found to be significant for nocturnal dipping come close to being significant for these sleep variables, and in no case did the shape of the interaction obtained match that found for dipping. As such, there is no evidence that the effects for dipping and nighttime BP in this study were the result of differences in sleep quality.

Discussion To summarize, the key findings of this study indicated significant interactions between Ethnicity, Genotype, Gender, and AI for both systolic and diastolic dipping, in which the significant effects were found only among Indian men. Specifically, Ethnicity "

15

Percentage Dipping

115

SS

p < .02 p ns 10

5 LS

SS

0 Low High

Low

Anger In Figure 2. Genotype " Anger In interaction for nighttime SBP, Indian men only.

High

Outward Anger Figure 4. Genotype " Outward Anger interaction for systolic dipping. Percentage dipping refers to the percentage drop in blood pressure from day to night.

5-HTTLPR and nocturnal dipping Genotype ! Gender ! AI interactions were obtained in which Indian men with the SS genotype showed reduced dipping when they were low in AI but greater dipping when they were high in AI, an effect that was not found in any other group. In addition an interaction was obtained between Genotype and OA for systolic dipping, in which individuals with the SS genotype showed reduced dipping when they were high on OA. In all cases, these effects appear to be the result of BP remaining high at night for the groups with reduced dipping as compared to those showing more dipping. Analyses of quality of sleep found no evidence that the reduced dipping or related higher nighttime BP was a function of poorer sleep quality. Finally, there was no evidence that reduced dipping was related to daytime BP. The results of this study do not replicate the findings of Bishop at al. (2005) showing that Indians high in trait anger (TA) showed reduced SBP nocturnal dipping. It is not clear why the results of Bishop et al. were not replicated. This is something that needs to be examined in future research. One possibility is that the differences between the results of Bishop et al. and the findings here come from differences in the methodology in that in Bishop et al. dipping was measured for only one 24-h period whereas in the present study participants were monitored for up to 4 days. The greater stability obtained from multiday monitoring would argue that the results from Bishop et al. may have been a chance finding and hence can be discounted. Although the Bishop et al. (2005) finding was not replicated, a conceptually related finding was obtained. Specifically, a different pattern of dipping was obtained for Indians than for Chinese and Malays, one that fits with the higher CHD rates found among Indians. Specifically, the Ethnic ! Gender ! Genotype ! AI interactions showed that the relationship between anger and dipping depended on ethnicity, gender, and genotype, with the relationship between AI and dipping found only for Indians men with the SS genotype. This finding needs to be replicated, but it provides the first evidence we know of linking 5HTTLPR to nocturnal dipping and to ethnic differences in CVD risk. At this point, it is not known whether Indian men with CHD are more likely than other groups to have low Anger In combined with the SS genotype. Follow-up analyses of the participants in the present study, however, are consistent with this possibility in showing that Indian men with the SS genotype did tend to have low AI scores (M 5 16.1) relative to the mean across all participants (18.12). This is consistent with higher CHD risk for this group, because the lowest dipping and hence higher CHD risk was found with low AI scores. Future research should examine whether this pattern is found in relationship to diagnosed CHD. Another question that needs to be addressed in future research is why higher levels of AI are protective for Indian men with the SS genotype. One possibility is that, because AI can be interpreted as representing a suppression of anger, individuals with low AI may have more of a tendency to express angry feelings, with this expression of angry feelings then being associated with greater CHD risk (cf. Siegman, 1994). This interpretation suggests that Indian men with the SS genotype would tend to be more expressive of their anger, which would then lead to greater CHD risk. In addition, a Genotype ! OA interaction was obtained that showed that systolic dipping was reduced for individuals high in OA but only when they had the SS genotype. As this effect was not qualified by any higher order interaction, this is an effect that applies equally to the three ethnic groups and also

1099 equally to men and women. Because OA reflects the extent to which a person experiences and expresses anger outwardly, this is a further suggestion that it is the tendency to express angry feelings that increases CHD risk. When interpreting these interactions between genotype and anger, it is important to note that, for both the Genotype ! AI interaction for Indian men and the Genotype ! OA interaction, it is principally individuals with the SS genotype who show a relationship between anger and dipping. The exception to this is that for diastolic dipping Indian men with the LS genotype showed a negative relationship between AI and dipping. For all results involving systolic dipping, however, no relationship was obtained between anger and dipping for those with the LS genotype. In addition, individuals with the SS genotype show extremes of dipping in that Indian men who are high on AI show the greatest percentage of dipping in that ethnic-gender group whereas those low on AI show the least dipping. This fits a pattern recently identified by Belsky et al. (2009) as representing ‘‘for-better-or-for-worse’’ or plasticity. In other words, individuals with the SS genotype do not just show greater CVD risk (in this case reduced nocturnal dipping) when they are high in OA or, if Indian and male, low in AI, but also show reduced risk when they are at the other end of the anger scale. This suggests that for individuals with the SS genotype intervention for helping them deal better with anger may go beyond simply reducing risk to conferring greater resilience than would be the case for individuals with the LS genotype. The fact that relationships between anger and dipping were only found for individuals with the SS genotype fits with other evidence related to 5-HTTLPR. Characteristically, it is individuals with the S allele that show enhanced vulnerability or, in many cases, plasticity in their responses to the environment (Belsky et al., 2009). This has been found to be the case for depression (Brummett et al., 2008; Caspi et al., 2003), anxiety (Gunthert et al., 2007), and ADHD (Retz et al., 2008). The results of the present study add additional evidence for considering the low activity allele of 5-HTTLPR to be a plasticity gene, that is, a gene that renders the person more responsive to environmental influences whether for better or for worse. At this point in time, the pathway connecting the SS genotype to variations in nocturnal dipping is not clear. One possibility is that the differences in nocturnal dipping may be because of differences occurring during sleep. Differences in dipping were not associated with daytime SBP values but appeared to be a direct reflection of nighttime SBP, with individuals low in dipping showing sustained nighttime SBP. Although we found no evidence that differences in sleep quality accounted for the higher levels of nighttime SBP for those with reduced dipping, it is quite possible that the ad hoc sleep questionnaire used was not sensitive enough to detect relevant differences. It is instructive to note that Brummett et al. (2007) found sleep quality to reflect an interaction between 5-HTTLPR genotype and stress, with individuals homozygous for the S allele showing reduced sleep quality when experiencing the stress of being a caregiver for a spouse or parent with dementia as compared to controls or individuals with at least one L allele. The differences observed by Brummett et al. (2007) could easily be reflected in higher SBP at night. This is a possibility that needs to be explored in future research. The findings of this study need to be interpreted in light of its limitations. First the participants in this study were all young adults within a limited age range. As such, it is important to replicate these results in an older sample with a broader age

1100

G.D. Bishop et al.

range. Second, all of the participants were Asian and living in Singapore. It is thus possible that these findings apply only to Asians. Future studies should replicate these findings in other ethnic groups so as to test for their generalizability. Third, the small number of individuals with the LL genotype and the dissimilarity of this group with the LS group during preliminary analyses made it impossible to include them in the reported analyses, as this would have implied an L-dominant model for which there is no evidence. As such, these analyses are of necessity incomplete, and there is no evidence for the LL genotype concerning the relationship of anger to dipping. Given the relative rarity of the L allele in Asian populations, future research with Asians will require substantially larger samples in order to

include individuals with the LL genotype in analyses. Finally, the use of an ad hoc sleep questionnaire reduces our confidence in the null findings obtained from this instrument and argues for the use of a more sensitive and preferably well-validated instrument for measuring sleep variables in future research. In conclusion, this study obtained evidence for interactions between anger and 5-HTTLPR genotype such that, across ethnicity and gender, individuals with the SS genotype who were high in OA showed reduced systolic dipping. Further Indian men with the SS genotype showed reduced dipping when they were low in AI but greater dipping when they were high in AI. These findings fit the interpretation of the low activity allele of the serotonin transporter gene as being a plasticity gene.

REFERENCES Anand, S. S., Yusuf, S., Vuksan, V., Devanesen, S., Teo, K. K., Montague, P. A., et al. (2000). Differences in risk factors, atherosclerosis and cardiovascular disease among ethnic groups in Canada: The study of health assessment and risk in ethnic groups (SHARE). Indian Heart Journal, 52, S35–S43. Belsky, J., Jonassaint, C., Pluess, M., Stanton, M., Brummett, B., & Williams, R. (2009). Vulnerability genes or plasticity genes? Molecular Psychiatry, 14, 746–754. Bishop, G. D., Pek, J., & Ngau, F. (2005). Ethnicity, sex, trait anger and nocturnal blood pressure decline. Psychophysiology, 42, 290–297. Brummett, B. H., Boyle, S. H., Siegler, I. C., Kuhn, C. M., Ashley-Koch, A., Jonassaint, C. R., et al. (2008). Effects of environmental stress and gender on asociations among symptoms of depression and the serotonin transporter gene linked polymorphic region (5-HTTLPR). Behavior Genetics, 38, 34–43. Brummett, B. H., Krystal, A. D., Ashley-Koch, A., Kuhn, C. M., Zu¨chner, S., Siegler, I. C., et al. (2007). Sleep quality varies as a function of 5-HTTLPR genotype and stress. Psychosomatic Medicine, 69, 621–624. Caspi, A., Sugden, K., Moffitt, T. E., Taylor, A., Craig, I. W., Harrington, H., et al. (2003). Influence of life stress on depression: Moderation by a polymorphism in the 5-HTT gene. Science, 301, 386–389. Czupryniak, L., Mlynarski, W., Pawlowski, M., Saryusz-Wolska, M., Borkowska, A., Bodalski, J., et al. (2008). Circadian blood pressure variation in normotensive type 2 diabetes patients and angiotensin converting enzyme polymorphism. Diabetes Research and Clinical Practice, 80, 386–391. Fava, C., Burri, P., Almgren, P., Arcaro, G., Groop, L., Hulthe´n, U. L., et al. (2005). Dipping and variability of blood pressure and heart rate at night are heritable traits. American Journal of Hypertension, 18, 1402–1407. Fumo, M. T., Teeter, S., Lang, R. M., Bednarz, J., Sareli, P., & Murphy, M. B. (1992). Diurnal blood pressure variation and cariac mass in American Blacks and Whites and South African Blacks. American Journal of Hypertension, 5, 111–116. Gunthert, K. C., Conner, T. S., Armeli, S., Tennen, H., Covault, J., & Kranzler, H. R. (2007). Serotonin transporter gene polymorphism (5-HTTLPR) and anxiety reactivity in daily life; A daily process approach to gene-environment interaction. Psychosomatic Medicine, 69, 762–768. Herbert, L. A., Agarwal, G., Ladson-Wofford, S. E., Reif, M., Hiremath, L., Carlton, S. G., et al. (1996). Nocturnal blood pressure in treated hypertensive African Americans compared to treated hypertensive European Americans. Journal of the American Society of Nephrology, 7, 2130–2134. Hughes, J. W., Kobayashi, I., & Deichert, N. T. (2007). Ethnic differences in sleep quality accompany ethnic differences in night-time blood pressure dipping. American Journal of Hypertension, 20, 1104– 1110. Hughes, K., Lun, K. C., & Yeo, P. P. B. (1990). Cardiovascular diseases in Chinese, Malays, and Indians in Singapore. I. Differences in mortality. Journal of Epidemiology and Community Health, 44, 24–28. Ituarte, P. H. G., Kamarck, T. W., Thompson, H. S., & Bacanu, S. (1999). Psychosocial mediators of racial differences in nighttime

blood pressure dipping among normotensive adults. Health Psychology, 18, 393–402. Lesch, K. P., Bengel, D., Heils, A., Sabol, S. Z., Greenberg, B. D., Petri, S., et al. (1996). Association of anxiety-related traits with a polymorphism in the serotonin transporter gene regulatory region. Science, 274, 1527–1531. Marler, M. R., Jacob, R. G., Lehoczky, J. P., & Shapiro, A. P. (1988). The statistical analysis of treatment effects in 24-hour ambulatory blood pressure recordings. Statistics in Medicine, 7, 697–716. Marmot, M. G., Adelstein, A. M., & Bulusu, L. (1984). Lessons from the study of immigrant mortality. Lancet, i, 1455–1458. Miller, G. J., Beckles, G. L. A., Alexis, S. D., Byam, N. T. A., & Price, S. G. L. (1982). Serum lipoproteins and susceptibility of men of Indian descent to coronary heart disease: The St. James survey, Trinidad. Lancet, ii, 200–203. Murphy, D. L., Fox, M. A., Timpano, K. R., Moya, P. R., Ren-Patterson, R., Andrews, A. M., et al. (2008). How the serotonin story is being rewritten by new gene-based discoveries prinicipally related to SLC6A4, the serotonin transporter gene, which functions to influence all cellular serotonin systems. Neuropharmacology, 55, 932–960. O’Brien, E., Sheridan, J., & O’Malley, K. (1988). Dippers and nondippers. Lancet, 2, 397. Ohkubo, T., Imai, Y., Tsuji, I., Nagai, K., Watanabe, N., Minamai, N., et al. (1997). Relations between nocturnal decline in blood pressure and mortality: The Ohasama Study. American Journal of Hypertension, 10, 1201–1207. Palatini, P., Penzo, M., Raccioppa, A., Zugno, E., Guzzardi, G., Anaclerio, M., et al. (1992). Clinical relevance of nighttime blood pressure and of daytime blood pressure variability. Archives of Internal Medicine, 152, 1855–1860. Ra¨ikko¨nen, K., Matthews, K. A., Kondwani, K. A., Bunker, C. H., Melhem, N. M., Ukoli, F. A. M., et al. (2004). Does nondipping of blood pressure at night reflect a trait of blunted cardiovascular responses to daily activities? Annals of Behavioral Medicine, 27, 131– 137. Retz, W., Freitag, C. M., Retz-Junginger, P., Wenzler, D., Schneider, M., Kissling, C., et al. (2008). A functional serotonin transporter promoter gene polymorphism increases ADHD symptoms in delinquents: Interaction with adverse childhood environment. Psychiatry Research, 158, 123–131. Siegman, A. W. (1994). Cardiovascular consequences of expressing and repressing anger. In A. W. Siegman & T. W. Smith (Eds.), Anger, hostility, and the heart (pp. 173–197). Hillsdale, NJ: Erlbaum. Spielberger, C. D. (1988). State-Trait Anger Expression Inventory (revised research edition): STAXI professional manual. Odessa, FL: Psychological Assessment Resources. Spielberger, C. D., Johnson, E. H., Russell, S. F., Crane, R. J., Jacobs, G. A., & Worden, T. J. (1985). The experience and expression of anger: Construction and validation of an anger expression scale. In M. A. Chesney & R. H. Rosenman (Eds.), Anger and hostility in cardiovascular and behavioral disorders (pp. 5–28). Washington, DC: Hemisphere Publishing. Spiering, W., Zwaan, I. M., Kroon, A. A., & de Leeuw, P. W. (2005). Genetic influences on 24 h blood pressure profiles in a hypertensive population: Role of the angiotensin-converting enzyme insertion/de-

5-HTTLPR and nocturnal dipping letion and angiotensin II type 1 receptor A1166C gene polymorphisms. Blood Pressure Monitoring, 10, 135–141. Staessen, J. A., Bieniaszewski, L., O’Brien, E., Gosse, P., Hayashi, H., Imai, Y., et al. (1997). Nocturnal blood pressure fall on ambulatory monitoring in a large international database. Hypertension, 29, 30–39. Staessen, J. A., Thijs, L., Fagard, R., O’Brien, E. T., Clement, D., de Leeuw, P. W., et al. (1999). Predicting cardiovascular risk using conventional vs ambulatory blood pressure in older patients with systolic hypertension. Systolic Hypertension in Europe Trial Investigators. Journal of the American Medical Association, 282, 539–546. Thomas, K. S., Nelesen, R. A., & Dimsdale, J. E. (2004). Relationships between hostility, anger expression, and blood pressure dipping in an ethnically diverse sample. Psychosomatic Medicine, 66, 298–304. Timio, M., Venanzi, S., Lolli, S., Lippi, G., Verdura, C., Monarca, C., et al. (1995). ‘‘Non-dipper’’ hypertensive patients and progressive renal insufficiency: A 3-year longitudinal study. Clinical Nephology, 43, 382–387. Verdecchia, P., Schillaci, G., Borgioni, C., Ciucii, A., Sacchi, M., Battistelli, M., et al. (1995). Gender, day-night blood pressure changes,

1101 and left ventricular mass in essential hypertension: Dippers and peakers. American Journal of Hypertension, 8, 193–196. Walker, A. R. P. (1980). The epidemiology of ischemic heart disease in the different ethnic populations in Johannesburg. South African Medical Journal, 57, 748–752. Williams, R. B., Marchuk, D. A., Gadde, K. M., Barefoot, J. C., Grichnik, K., Helms, M. J., et al. (2003). Serotonin-related gene polymorphisms and central nervous system serotonin function. Neuropsychopharmacology, 28, 533–541. Williams, R. B., Marchuk, D. A., Siegler, I. C., Barefoot, J. C., Helms, M. J., Brummett, B. H., et al. (2008). Childhood socioeconomic status and serotonin transporter gene polymorphism enhance cardiovascular reactivity to mental stress. Psychosomatic Medicine, 70, 32– 39.

(Received May 30, 2009; Accepted November 6, 2009)

Psychophysiology, 47 (2010), 1102–1108. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01018.x

Comparison of baroreceptor cardiac reflex sensitivity estimates from inter-systolic and ECG R-R intervals

GUSTAVO A. REYES DEL PASO, M. ISABEL GONZA´LEZ and JOSE´ A. HERNA´NDEZ Departamento de Psicologı´ a. Universidad de Jae´n, Jae´n, Spain

Abstract Baroreceptor reflex sensitivity (BRS) is frequently evaluated using the spontaneous sequence method. Many of these studies use the inter-systolic interval (ISI) derived from a blood pressure monitor (e.g., Finapres) as interbeat interval measure instead of the traditionally recommended R-R series derived from the ECG. In this study, we examine possible differences between estimates of BRS from ISI and ECG R-R intervals. BRS was evaluated in 35 participants under three conditions: rest, mental arithmetic, and recovery periods. Although correlations between the two estimates are very high (all rs4.9), small but significant differences were found: the measures from ISI systematically yield higher BRS values and result in the detection of a greater number of reflex sequences. The higher BRS values from measures of ISI are due to the effects of pulse transit time fluctuations associated with the sequences of change in blood pressure. Descriptors: Baroreceptor cardiac reflex sensitivity, Inter-systolic interval, ECG R-R, Agreement analysis

et al., 2001), where the modulus or gain function between the variations in SBP and IBI generates an estimate of baroreceptor reflex sensitivity (BRS, i.e., the ms of change in IBI per mmHg change in SBP). More complex frequency domain techniques employ exogenous autoregressive causal parametric modelling (Persson et al., 2001). These techniques make it possible to consider other specific influences on IBI, such as respiratory-related sources (Porta, Baselli, Rimoldi, Malliani, & Pagani, 2000), independently of the baroreflex. Although these techniques commonly involve the use of an indirect measure of blood pressure at finger level (which under some circumstances may not be equivalent to an invasive measure near the heart, Imholz, Wieling, van Montfrans, & Wesseling, 1998), they do not require any external intervention on the subjects under evaluation and are thus more suitable for psychophysiological research. Indices of spontaneous BRS are frequently evaluated in psychophysiological research and clinical practice since they are relatively easy to obtain and are sensitive to psychological and health processes. Some previous studies have compared various techniques to estimate BRS (Laude, Elghozi, Girard, Bellard, Bouhaddi, et al., 2004; Reyes del Paso, 1994; Vallais, Baselli, Lucini, Pagani, & Porta, 2009). Despite a significant correlation, the different methods lead to different results as a function of factors such as the specific experimental condition, the different procedures implementing each technique, etc. (Laude et al., 2004; Vallais et al., 2009). Instead of comparing different methods, in this study we analyze estimates of baroreflex function from IBI time series obtained through blood pressure and electrocardiogram (ECG) recordings. The first step in calculating BRS is the accurate determination of individual SBP points and IBI. BRS measures vary within a range of a few ms/mmHg (e.g., 12 ! 7 ms/mmHg, in healthy

The baroreceptor reflex, which relates blood pressure to interbeat interval (IBI), is a basic mechanism for the short-term regulation of blood pressure, a relevant source of autonomic cardiac regulation, and a powerful prognostic factor for cardiovascular health (Parati, di Rienzo, & Mancia, 2000; Persson, di Rienzo, Castiglioni, Cerutti, Pagani, et al., 2001; Reyes del Paso, Langewitz, Robles, & Pe´rez, 1996). Traditional methods to assess baroreceptor function are based on laboratory tests that require the application of an external stimulus to the subjects to alter blood pressure while the reflex heart rate response is measured (e.g., see Parati et al., 2000 for a review). With the availability of modern beat-to-beat blood pressure measurement devices, techniques have been developed for the non-invasive analysis of spontaneous baroreflex control of cardiac activity. These techniques are based on the analysis of spontaneous covariation of systolic blood pressure (SBP) and IBI, both in the time and frequency domain (Parati et al., 2000; Persson et al., 2001). The application of this methodology in the time domain, usually known as the sequence method (Parati et al., 2000; Persson et al., 2001; Steptoe & Sawada, 1989) locates spontaneous cardiac sequences in which the baroreflex is operating. The most commonly used frequency domain techniques are based on spectral analysis or auto-regressive modelling (Parati et al., 2000; Persson This research was supported by grants from the Spanish Commission of Science and Technology (Project No. SEJ2006-09808) and Junta de Andalucı´ a (Research Group HUM338). We would like to thank the anonymous reviewers of Psychophysiology for their valuable comments and suggestions that significantly improved the final version of the manuscript. Address reprint requests to: Gustavo A. Reyes del Paso, Departamento de Psicologı´ a, Universidad de Jae´n, 23071 Jae´n, Spain. E-mail: [email protected] 1102

BRS from inter-systolic and ECG R-R intervals young individuals) so calculation to the nearest millisecond is essential for the IBI measure. The ECG is the ‘‘gold standard’’ in cardiovascular psychophysiology (Berntson, Bigger, Eckberg, Grossman, Kaufmann, et al., 1997; Jennings, Berg, Hutcheson, Obrist, & Turpin, 1981) as it is assumed to provide the greatest accuracy and is therefore the recommended method for obtaining the IBI. The use of IBI measures derived from arterial pressure pulse in distal locations such as the finger (i.e., plethysmography) can present two problems (Jennings et al., 1981). First, blood pressure pulses lack the clear sharp peak that can be observed in the ECG, making it more difficult to determine a clear point in the recording for peak detection. Specifically, the critical time interval for systolic peak detection in plethysmographic recording is longer, while the ECG R-spike is clearly recognizable. Second, the propagation of the pulse arterial pressure is influenced by ventricular and vascular factors. Especially important in this concern is the influence of pulse transit time (Giardino, Lehrer, & Edelberg, 2002; Sharpley, 1993). Average measures of IBI from ECG and plethysmography generally yield similar results. However, during periods of cardiac acceleration and deceleration, some differences can be found (Sharpley, 1993). These observations have led to the conclusion that the plethysmographic procedure is less acceptable for beat-by-beat or second-by-second measures of heart rate (Jennings et al., 1981). Despite the recommended use of the ECG, many of the studies conducted on BRS (e.g., Reyes del Paso, Herna´ndez, & Gonza´lez, 2004; Steptoe & Sawada, 1989; Steptoe & Vo¨gele, 1990; Watkins, Fainman, Dimsdale, & Ziegler, 1995) have used as IBI measure the inter-systolic interval (ISI) derived from a blood pressure monitor, assuming implicitly that the resulting indices would be comparable to those derived from ECG. The signal provided by these devices (e.g., finapress) is fairly stable, with very few artifacts, and is less affected by changes in vascular tone than are traditional plethysmograph measures (Imholz et al., 1998). Advantages of using the ISI are: 1) it makes BRS easier to obtain, especially for clinical use, given that only one single blood pressure channel (or single device) of physiological recording is needed; and 2) this procedure also simplifies the management of the run-off periods during periodic recalibration of the blood pressure device, facilitating the synchronization of the SBP and IBI time-series. This study aimed at validating estimates of BRS in the time domain obtained from plethysmography by examining possible differences between indices derived from blood pressure IBI-ISI and ECG IBI R-wave to R-wave intervals (R-R). Additionally, the influence of both procedures on the mean number of detected baroreflex sequences will be analyzed. Given the interest in psychophysiology to investigate alterations in cardiovascular parameters under psychological manipulations, this comparison is carried out under three conditions: rest, mental arithmetic, and recovery periods, and the relative sensitivity of the two indices to this manipulation will be also analyzed. We will also analyze the associations between pulse transit time and baroreceptor parameters since pulse transit time is the main candidate to explain possible differences between methods.

Method BRS was evaluated in thirty-five healthy participants, 7 men and 28 women, aged between 18 and 24 years. Blood pressure was

1103 recorded continuously and non-invasively with the Ohmeda 2300 BP monitor (Ohmeda, Louisville, KY) from the middle phalanx of the third finger of the right hand. The hand was positioned at the level of the heart. ECG was obtained from Einthoven Lead II using a Biopac ECG amplifier. Data acquisition and recording of both ECG and blood pressure were carried out at 500 Hz using a Biopac MP100 system (Biopac System Inc., Goleta, CA). It is necessary to take into account that in current devices (e.g., Finometer) the analog output is a reconstruction of a 200-Hz sampled digital recording, which limits the ‘‘true’’ time resolution of the blood pressure signal to 5 ms. If higher time resolution is desired, a resampling procedure is needed on the 200-Hz digital recording to increase sample rate to 500 or 1000 Hz. This can be done by means of spline (e.g., Duschek, Dietel, Shandry, & Reyes del Paso, 2008) or parabolic (e.g., Di Rienzo, Parati, Castiglioni, Tordi, Mancia, & Pedotti, 2001) interpolation, for example, with the Matlab software (The MathWorks, Inc., South Asheboro, NC). Each experimental session comprised of a 10-min rest period (of which the last 5 min were taken as baseline), a 3-min mental arithmetic task, and another 3-min recovery period. Participants added and/or subtracted three two-figure numbers which were presented for a maximum of 17 s on the computer screen. The subjects gave the response by typing the solution on the computer keyboard. If no answer was given in this time, the following trial was presented. Custom flexible algorithms were used to detect beat-to-beat maximum values during ECG R-spike and SBP peaks. Correct functioning of these algorithms was checked by visual inspection and manual editing of the automatic detection outcomes. IBI is taken as the interval between successive systolic peaks (i.e., IBIISI) or the interval between successive ECG R-R (IBI-RR). Baroreceptor function was evaluated using a program (Reyes del Paso, 1994) that searches for sequences of three to six consecutive cardiac cycles in which SBP increases (by at least 1 mmHg per beat) in combination with an increase in IBI (of at least 2 ms per beat) or sequences in which the decrease of the SBP is accompanied by a decrease in IBI (following the same criteria of minimum change). Each systolic value is paired with the IBI calculated from the heart beat immediately following itFa lag of one beat, which is associated with the better estimates of BRS (Steptoe & Vo¨gele, 1990). To consider one of these sequences as a baroreflex sequence, the correlation coefficient between the SBP and IBI values has to be larger than 0.87. When one of these sequences is detected, the corresponding regression line is computed across all pulses in that particular sequence, and the slope of the line is taken as an index of BRS (units ms/mmHg). The average value of the BRS index for the whole period considered is then computed. The number of detected reflex sequences is reported in terms of sequences per minute. Pulse transit time is taken as the time (ms) elapsed between the R-spike and the closest following systolic point. Differences between both estimates were evaluated using Pearson’s product-moment correlations and mean difference comparisons with paired Student t tests. Because the use of correlations may be misleading when comparing two measurement techniques, we used the agreement analysis methods suggested by Bland and Altman (1986), which involves the plotting of the difference between methods against the mean of both measures. The limits of agreement and confidence intervals were also computed (Bland & Altman, 1986). Effects of the experimental condition on physiological parameters were assessed through

1104

G. A. Reyes del Paso et al.

repeated measures analysis of variance (ANOVA) designs with two levels (baseline vs. task or task vs. recovery). To evaluate the sensitivity of the baroreceptor indices to the experimental conditions (if the above effects were significant in the whole sample for the two methods), we used a bootstrap approach to again evaluate the effects by randomly taking 1000 subsets of 12 participants. Comparisons within the bootstrap procedure were performed with paired t-tests to detect the percentage of cases with a significant difference. Results Table 1 shows the descriptive data both of BRS and the number of reflex sequences detected as well as the t statistics. Scatterplots are displayed in Figure 1. There is a very high linear association between the values from the two methods as all correlations are above 0.9 (pso.001). However, there were significant differences between the estimates from IBI-ISI and IBI-RR. For the three experimental periods, measures from IBI-ISI resulted in higher BRS estimates and in a larger number of reflex sequences. The mean differences between the two methods for BRS was 1.04 ! .63, .71 ! .80, and .44 ! .88 ms/mmHg for baseline, mental stress, and recovery, respectively. For the number of reflex sequences, the mean differences were 4.05 ! 5.78, 1.54 ! 4.58, and 2.20 ! 3.03 sequences/min for baseline, mental stress, and recovery, respectively. Figure 2 plots the differences between the methods against their mean (mean of the two methods). The figure shows a systematic bias: the estimates from IBI-ISI yield higher BRS values and a larger number of reflex sequences. Except for BRS during baseline (r 5 .424, p 5 .011), the bias between methods was not correlated with the mean magnitude of the estimate (r 5 .072 and r 5 " .281 for BRS during task and recovery, and r 5 .225, r 5 " .006, and r 5 .038, for the number of reflex sequences during baseline, task, and recovery, respectively). The lack of agreement can be summarized by calculating the bias from the mean differences and the standard deviation of the differences. Assuming a normal distribution of the differences, 95% of them will lie between the mean difference !2 standard deviations (i.e., 95% confidence interval). Table 1. Descriptive Statistics for BRS (ms/mmHg) and Number of Reflex Sequences (N1 Seq. in sequences/min) as a Function of the Methods and Experimental Periods

BRS-ISI (BL) BRS-RR (BL) BRS-ISI (Task) BRS-RR (Task) BRS-ISI (Recov.) BRS-RR (Recov.) N1 Seq.-ISI (BL) N1 Seq.-RR (BL) N1 Seq.-ISI (Task) N1 Seq.-RR (Task) N1 Seq.-ISI (Recov.) N1 Seq.-RR (Recov.)

M

SD

Min

Max

12.29 11.24 10.28 9.57 11.96 11.52 12.36 11.55 10.75 10.24 12.10 11.37

5.07 4.79 4.13 4.15 4.49 4.74 3.57 3.32 4.42 4.43 3.52 3.48

5.17 4.57 2.89 2.62 5.13 4.34 4.80 4.60 3.33 3.67 5.33 6.00

26.47 24.32 22.15 21.12 21.16 22.76 20.40 18.60 21.67 21.00 20.67 20.00

t

p

9.69

.0001

5.24

.0001

2.96

.006

3.54

.001

1.99

.055

4.28

.0001

Note: ISI: estimates obtained from IBI-ISI; RR: estimates obtained from IBI-RR, BL: baseline, Task: mental stress, Recov.: recovery, M 5 mean, SD 5 standard deviation, Min 5 minimal, Max 5 maximal. Results of the comparisons between methods with the paired t-test are also displayed.

The changes from baseline to task were significant for both BRS from IBI-ISI (F(1,34) 5 20.09, po.0001, m2 5 .40) and BRS from IBI-RR (F(1,34) 5 14.02, p 5 .011, m2 5 .29). The changes from task to recovery were also significant both for BRS from IBI-ISI (F(1,34) 5 15.02, po.0001, m2 5 .31) and BRS from IBI-RR (F(1,34) 5 19.23, po.0001, m2 5 .36). The results of the bootstrap analysis for the comparisons between baseline and task percentages of significant cases were 78.3% vs. 59.8% for BRS from IBI-ISI and IBI-RR, respectively. For the comparisons between task and recovery, the percentages obtained were 60.8% vs. 68.7% for BRS from IBI-ISI and IBI-RR, respectively. For the number of reflex sequences, the change from baseline to task was significant for the BRS IBI-ISI (12.36 ! 3.57 vs 10.75 ! 4.42 sequences/min, F(1,34) 5 5.73, po.05, m2 5 .14) but not for IBI-RR (11.55 ! 3.32 vs 10.23 ! 3.43 sequences/m, F(1,34) 5 3.77, p4.05, m2 5 .10). The change from task to recovery was also significant for IBI-ISI (10.75 ! 4.42 vs 12.10 ! 3.32 sequences/m, F(1,34) 5 6.56, po.05, m2 5 .16) but not for IBI-RR (10.23 ! 3.43 vs 11.37 ! 3.48 sequences/m, F(1,34) 5 3.24, p4.05, m2 5 .09). Pulse transit time decreased from baseline to task (268 ! 22 to 252 ! 25 ms, F(1,34) 5 93.29, po.0001, m2 5 .73) and increased from task to recovery (252 ! 25 to 260 ! 22 ms, F(1,34) 5 49.67, po.0001, m2 5 .59). Correlations between pulse transit time and baroreceptor indices are displayed in Table 2. Pulse transit time correlates positively with both BRS indices, but the associations do not reach significance for the differences between methods. For the number of reflex sequences, pulse transit time correlates positively with the differences between methods in the baseline and task period, but is not correlated with the number of reflex sequences for any of the methods (see Table 2).

Discussion The results of this study show a very high linear association between the baroreceptor indices provided by the IBI-ISI and IBI-RR (see Figure 1). However, this almost perfect correlation between methods does not mean that the resulting measures agree. Although the differences are small, the estimates from the IBI-ISI systematically produce a high BRS and lead to the detection of a greater number of reflex sequences than the estimates from IBI-RR (see Table 1 and Figure 2). Except for BRS during baseline, where a slight association was observed, the bias between the methods was uncorrelated with the mean magnitude of the estimate, which could be considered of crucial importance for the establishment of a basic equivalence between the two methods. As regard to sensitivity to experimental manipulation for the whole sample, the two BRS indices appear to be equivalent in terms of their ability to detect the decrease in BRS during mental stress-load conditions like our mental arithmetic task (Reyes del Paso, Gonza´lez, & Herna´ndez, 2004; Reyes del Paso, Gonza´lez, Herna´ndez, Duschek, & Gutie´rrez, 2009; Reyes del Paso, Langewitz, Robles, & Pe´rez, 1996; Steptoe & Sawada, 1989; Yasumasu, Reyes del Paso, Takahara, & Nakashima, 2006). When the bootstrap approach was applied, even the BRS index from IBI-ISI appeared to be superior in detecting the predicted decrease in BRS from baseline to task (78.3% vs. 59.8% of significant comparisons for BRS from IBI-ISI and IBI-RR, respectively). As regard to the number of detected reflex sequences, only

BRS from inter-systolic and ECG R-R intervals

1105

B

A

C 25.00

r = .99

15.00 10.00 5.00 0.00

25.00

20.00

r = .98

15.00 10.00 5.00

0.00

5.00

5.00

10.00 5.00

10.00

15.00

20.00

25.00

BRS-ISI (ms/mmHg)

20.00 r = .94 15.00 10.00 5.00

Number of Sequences (RR)

r = .95

15.00

10.00

C

Number of Sequences (RR)

Number of Sequences (RR)

20.00

10.00

15.00

5.00 10.00 15.00 20.00 25.00

B

A

5.00

r = .98

BRS-ISI (ms/mmHg)

BRS-ISI (ms/mmHg)

15.00

20.00

0.00

0.00 5.00 10.00 15.00 20.00 25.00 30.00

0.00

BRS-RR (ms/mmHg)

20.00

BRS-RR (ms/mmHg)

BRS-RR (ms/mmHg)

25.00

20.00 r = .96 15.00

10.00

5.00

20.00

0.00

Number of Sequences (ISI)

5.00 10.00 15.00 20.00 25.00

5.00

10.00

15.00

20.00

25.00

Number of Sequences (ISI)

Number of Sequences (ISI)

Figure 1. Scatterplot, regression lines, and r coefficients for the relationships between BRS (top) and number of detected reflex sequences (bottom) assessed from IBI-ISI and IBI-RR. A: baseline, B: task, C: recovery.

B

A 2.00 1.50 1.00 0.50 0.00

3.00 Differences Against Mean

Differences Against Mean

Differences Against Mean

C 3.00

2.50

2.00 1.00 0.00 –1.00

5.00

0.00

10.00 15.00 20.00 25.00

B Differences Against Mean

Differences Against Mean

4.00 3.00 2.00 1.00 0.00 –1.00

0.00

20.00

Average Number of Sequences

5.00 10.00 15.00 20.00 25.00 Average BRS (ms/mmHg)

C 4.00

3.00

2.00 0.00 –2.00

–2.00

15.00

–1.00

5.00 10.00 15.00 20.00 25.00

Difference Against Mean

A

10.00

0.00

Average BRS (ms/mmHg)

Average BRS (ms/mmHg)

5.00

1.00

–2.00

–2.00

–0.50

2.00

0.00

5.00 10.00 15.00 20.00 25.00

Average Nummber of Sequences

2.00 1.00 0.00 –1.00 8.00

12.00

16.00

20.00

Average Number of Sequences

Figure 2. Plot of the BRS (top) and number of detected reflex sequences (bottom) differences between the two methods against the respective means of the two methods. The central lines represent the mean difference, the upper lines the mean difference 12 SD, and the lower lines the mean difference ! 2 SD. A: baseline, B: task, C: recovery.

1106

G. A. Reyes del Paso et al.

Table 2. Correlations between Pulse Transit Time and BRS Indices

BRS-ISI BRS-RR BRS-Bias N1 Seq.-Bias

Baseline

Task

Recovery

.366 .353n .253 .335n

.436 .397n .269 .476nn

.362n .344n .214 .165

n

nn

Note: nnpo.01, npo.05. Correlations taken from IBI-ISI (BRS-ISI) and IBI-RR (BRS-RR), the difference between both BRS indices (BRS-Bias), and the difference between the two methods for the number of reflex sequences (No Seq.-Bias). Correlations for the number of sequences were lower than .2 and insignificant.

the index from IBI-ISI allows detecting significant differences between the experimental conditions. The number of reflex sequences is utilized in the time domain approach to calculate two secondary indices of baroreceptor cardiac functioning. The reflex is not always effective, and not all progressive changes in blood pressure are followed by compensatory reflex changes in IBI. This aspect of the baroreceptor function is evaluated by the baroreflex effectiveness index, defined as the frequency in which the progressive SBP changes are followed by the expected IBI reflex modulation (Di Rienzo et al., 2001; Duschek & Reyes del Paso, 2007; Reyes del Paso et al., 2004). Another index dependent on the number of reflex sequences is the baroreceptor power, defined as the proportion of cardiac cycles which form part of the reflex sequences with respect to the total number of cardiac cycles during the analyzed period (Duschek & Reyes del Paso, 2007; Reyes del Paso et al., 1996; Steptoe & Sawada, 1989). In light of the greater number of detected reflex sequences with the IBI-ISI method, both the baroreceptor effectiveness index and the baroreceptor power would be greater when obtained with the IBI-ISI procedure. The main factor to explain the differences we found between methods is pulse transit time. The time interval from the occurrence of the R-wave and the arrival of the blood pulse at the peripheral location are subject to two main factors. The most relevant determinants of pulse transit time are sympathetic b-adrenergic influences on myocardial contractility (Contrada, del Bo, Levy, & Weiss, 1995; Obrist, Light, McCubbin, Hutcheson, & Hoffer, 1979), with pulse transit time correlating negatively with measures of myocardial performance such as pre-ejection time. Second, but to a lesser extent, pulse transit time

248 ms

also correlates negatively with SBP, the correlations ranging from ! .49 to ! .7 (Newlin, 1981; Obrist et al., 1979). As expected, pulse transit time decreases during the mental arithmetic task, showing the increased sympathetic b-adrenergic influences on myocardial contractility under stress conditions (Duschek, Muckenthaler, Werner, & Reyes del Paso, 2009). Pulse transit time is positively associated with both BRS indices (see Table 2). These associations could be interpreted as the additive effects of the two factors determining pulse transit time. On the one hand, the cardiac baroreflex is one of the main determinants of vagal cardiac efferent activity (Reyes del Paso et al., 1996), and vagal activation exerts a significant inotropic control on the heart decreasing ventricular contractility (Contrada et al., 1995; Lewis, Al-Khalidi, Bonser, Clutton-Brock, Morton, et al., 2001). On the other hand, BRS is inversely related to tonic blood pressure (Hesse, Charkoudian, Liu, Joyner, & Eisenach, 2007). Both effects in turn induced greater pulse transit time values, and in such a way greater BRS could be expected to be related to longer pulse transit times. However, pulse transit time is not significantly associated with the differences between the two BRS methods. Despite being uncorrelated with the number of reflex sequences for any of the methods, greater pulse transit times increase the differences between methods, at least during baseline and task periods (see Table 2). The above analyses were performed with whole means, but to gain insight into the effect of pulse transit time in determining the differences between methods, the analysis of the individual IBI fluctuations could be more informative. It is known that progressive increases in pulse transit times are associated with increased differences between the IBI-ISI and IBI-RR time-series, while the opposite occurs during periods of progressive decreases in pulse transit times (Giardino et al., 2002; Sharpley, 1993). The differences we found between methods could be dependent on the oscillatory nature of cardiovascular parameters and the specific changes produced within the sequences of progressive increase or decrease in blood pressure. The calculation of BRS in the time domain relies on the detection of cardiac sequences of progressive blood pressure increases accompanied by cardiac deceleration and sequences of blood pressure decreases accompanied by cardiac acceleration. These progressive changes in blood pressure would produce pulse transit time fluctuations that ultimately have the effect of increasing the IBI oscillations. This effect is illustrated in Figure 3.

242 ms

908 ms

238 ms

1020 ms

902 ms

1110 ms

161 mmHg

158 mmHg

152 mmHg

240 ms

1016 ms

160 mmHg

1112 ms

Figure 3. Recording of ECG (top) and blood pressure (bottom) showing the IBI values associated with the ISI and R-R time series. Vertical lines are artificially inserted in the SBP peaks to facilitate the illustration of the resulting pulse transit time values (top file of numbers).

BRS from inter-systolic and ECG R-R intervals

1107

To understand the mechanism responsible for this interference, we need to take into account that for the calculation of regression slope each SBP value was paired with the IBI immediately subsequent to it. As mentioned above, pulse transit time is negatively associated with SBP and increases and decreases in association with SBP oscillations. The main sources of bias and contributors to the BRS differences are the first and last IBI-ISI values in the sequence. In a sequence of progressive increase in SBP, the first beat is associated with a decrease in pulse transit time, which, incorporated into the IBI-ISI measure, leads to a decrease in the real IBI. This makes the IBI-ISI time-series of the sequence begin with a small value. With respect to the last IBI-ISI value in the sequence, the second beat in its calculation is associated with a relative decrease in SBP, and consequently an increase in pulse transit time that artificially increases the IBI-ISI and makes this time series end with a greater value. Through this mechanism, the pulse transit time fluctuations associated with the progressive changes in blood pressure result in an amplification of the IBI oscillations. In the example in Figure 3, the IBI-RR change within the sequence is 1110–908 5 202 ms, while for the IBI-ISI it is 1112–902 5 210 ms. For the same SBP changes, the use of the IBI-ISI measure would result in greater BRS. Just the opposite occurs during the sequences of decreases in blood pressure, also leading to an increased BRS. The results obtained suggest that the incorporation of the pulse transit time oscillations into the IBI-ISI time series also favors the fulfillment of the criteria for progressive IBI changes, resulting in the detection of a greater number of reflex sequences. Another question is whether the bias we have found for the time domain procedure could be generalized to the frequency domain methods. Spectral techniques are based on the analysis of continuous covariation of SBP and IBI fluctuations rather than on the detection of specific reflex sequences. Given that fluctuations in SBP would generate parallel fluctuations in pulse transit time, thus increasing the amplitude of the IBI oscillations, the interfering mechanism affecting the time domain technique would very probably also affect the frequency domain procedure. Specifically, the assessment of BRS in frequency domain methods relies on calculations of the gain of the transfer function between the SBP and IBI changes (e.g., the square root of the ratio between the IBI and SBP variability powers or a coefficient, Parati et al., 2000; Persson et al., 2001) in frequency regions where SBP and IBI displayed a high coherence (i.e., linear coupling), as the high frequency (0.15 to 0.40 Hz) and low frequency

(0.04 to 0.15 Hz) bands. Giardino et al. (2002) compared finger plethysmography to ECG in the measurement of heart rate variability. They found that plethysmograph-derived heart rate variability was consistent and significantly higher than that obtained from ECG, both for the high and low frequency bands. The difference between methods was greater for the high frequency band, where respiratory-related blood pressure fluctuations could induce in-phase oscillations in pulse transit time. These results are congruent with those obtained in the present study and suggest that BRS estimates from IBI-ISI in the frequency domain would also result in greater BRS. If IBI variability power is higher when obtained through plethysmography, then BRS indices obtained in the frequency domain should also be greater than those obtained from ECG, for the low and especially for the high frequency bands. Results of the present study may depend on the specific method utilized to estimate spontaneous BRS, in this case the sequence method. Results might be different with other methodologies, especially those involving the use of complex autoregressive causal parametric modelling (Porta et al., 2000; Vallais et al., 2009). Another possible limitation of the study is the imbalance in our sample with respect to gender. Although there is no reason to think that gender could affect the mechanisms producing the differences between methods, this sample limitation should be taken into account. In conclusion, although both indices of BRS produce very similar results, a cautionary note is necessary: the index from plethysmography yields greater BRS values (with a mean of .73 ! .77 ms/mmHg during the three experimental periods in our study) and increases the probability of detecting reflex sequences. However, given the low magnitude of the differences and the very high linear correlations between the two methods, this bias is tolerable in most psychophysiological and clinical contexts, where interest is in the effects of certain manipulations (mental stress or load, relaxation, drug, biofeedback, physical maneuvers, etc.) on BRS through repeated measures designs (e.g., Reyes del Paso et al., 1996, 2004; Steptoe & Sawada, 1989), or the analysis of the relations between some variables (e.g., stress, pain, anxiety, depression, cognitive performance, cardiovascular disease, etc.) and BRS (e.g., Duschek & Reyes del Paso, 2007; Reyes del Paso et al., 2009; Yasumasu et al., 2006). The method also appears suitable for the comparison of two or more groups on baroreceptor parameters, as long as all groups are measured with the same IBI measure (e.g., Duschek et al., 2008).

REFERENCES Berntson, G. G., Bigger, J. T., Eckberg, D. L., Grossman, P., Kaufmann, P. G., Malik, M., et al. (1997). Heart rate variability: Origins, methods, and interpretative caveats. Psychophysiology, 34, 623–648. Bland, J. M., & Altman, D. G. (1986). Statistical methods for assessing agreement between two methods of clinical measurement. The Lancet, 8, 307–310. Contrada, R. J., del Bo, A., Levy, L., & Weiss, T. (1995). Form and magnitude of beta-sympathetic and parasympathetic influences on pulse transit time. Psychophysiology, 32, 329–334. Di Rienzo, M., Parati, G., Castiglioni, P., Tordi, R., Mancia, G., & Pedotti, A. (2001). Baroreflex effectiveness index: An additional measure of baroreflex control of heart rate in daily life. American Journal of Physiology. Regulatory Integrative and Comparative Physiology, 280, R744–R751. Duschek, S., Dietel, A., Schandry, R., & Reyes del Paso, G. A. (2008). Increased baroreflex sensitivity and reduced cardiovascular reactivity in chronic low blood pressure. Hypertension Research, 31, 1873–1878.

Duschek, S., Muckenthaler, M., Werner, N., & Reyes del Paso, G. A. (2009). Relationships between features of autonomic cardiovascular control and cognitive performance. Biological Psychology, 81, 110–117. Duschek, S., & Reyes del Paso, G. A. (2007). Quantification of cardiac baroreflex function at rest and during autonomic stimulation. Journal of Physiological Sciences, 57, 259–268. Giardino, N. D., Lehrer, P. M., & Edelberg, R. (2002). Comparison of finger plethysmograph to ECG in the measurement of heart rate variability. Psychophysiology, 39, 246–253. Hesse, C., Charkoudian, N., Liu, Z., Joyner, M. J., & Eisenach, J. H. (2007). Baroreflex sensitivity inversely correlates with ambulatory blood pressure in healthy normotensive humans. Hypertension, 50, 41–46. Imholz, B. D., Wieling, W., van Montfrans, G. A., & Wesseling, K. H. (1998). Fifteen years experience with finger arterial pressure monitoring: Assessment of the technology. Cardiovascular Research, 38, 605–616.

1108 Jennings, J. R., Berg, W. K., Hutcheson, J. S., Obrist, P., & Turpin, G. (1981). Publication guidelines for heart rate studies in man. Psychophysiology, 18, 226–231. Laude, D., Elghozi, J. L., Girard, A., Bellard, E., Bouhaddi, M., et al. for the EuroBaVar study (2004). Comparison of various techniques used to estimate spontaneous baroreflex sensitivity. American Journal of Physiology. Regulatory, Integrative and Comparative Physiology, 286, R226–R231. Lewis, M. E., Al-Khalidi, A. H., Bonser, R. S., Clutton-Brock, T., Morton, D., Paterson, D., et al. (2001). Vagus nerve stimulation decreases left ventricular contractility in vivo in the human and pig heart. Journal of Physiology, 534, 547–552. Newlin, D. B. (1981). Relationships of pulse transmission times to pre-ejection period and blood pressure. Psychophysiology, 18, 316– 321. Obrist, P. A., Light, K. C., McCubbin, J. A., Hutcheson, J. S., & Hoffer, J. L. (1979). Pulse transit time: Relationship to blood pressure and myocardial performance. Psychophysiology, 16, 292–301. Parati, G., di Rienzo, M., & Mancia, G. (2000). How to measure baroreflex sensitivity: From the cardiovascular laboratory to daily life. Journal of Hypertension, 18, 7–19. Persson, P. B., di Rienzo, M., Castiglioni, P., Cerutti, C., Pagani, M., Honzikova, N., et al. (2001). Time versus frequency domain techniques for assessing baroreflex sensitivity. Journal of Hypertension, 19, 1699–1705. Porta, A., Baselli, G., Rimoldi, O., Malliani, A., & Pagani, M. (2000). Assessing baroreflex gain from spontaneous variability in conscious dog: Role of causality and respiration. American Journal of Physiology. Heart and Circulation Physiology, 279, H2558– H2567. Reyes del Paso, G. A. (1994). A program to assess baroreceptor cardiac reflex function. Behavior Research Methods, Instruments, & Computers, 26, 62–64. Reyes del Paso, G. A., Gonza´lez, M. I., & Herna´ndez, J. A. (2004). Baroreceptor sensitivity and effectiveness varies differentially as a function of cognitive attentional demands. Biological Psychology, 67, 385–395.

G. A. Reyes del Paso et al. Reyes del Paso, G. A., Gonza´lez, M. I., Herna´ndez, J. A., Duschek, S., & Gutie´rrez, N. (2009). Tonic blood pressure modulated the relationships between baroreceptor cardiac reflex sensitivity and cognitive performance. Psychophysiology, 46, 932–938. Reyes del Paso, G. A., Herna´ndez, J. A., & Gonza´lez, M. I. (2004). Differential analysis in the time domain of the baroreceptor cardiac reflex sensitivity as a function of sequence length. Psychophysiology, 41, 483–488. Reyes del Paso, G. A., Langewitz, W., Robles, H., & Pe´rez, N. (1996). A between-subjects comparison of respiratory sinus arrhythmia and baroreceptor cardiac reflex sensitivity as non-invasive measures of tonic parasympathetic cardiac control. International Journal of Psychophysiology, 22, 163–171. Sharpley, C. F. (1993). Differences in pulse rate and heart rate and effects on the calculation of heart rate reactivity during periods of mental stress. Journal of Behavioral Medicine, 17, 99–109. Steptoe, A., & Sawada, Y. (1989). Assessment of baroreceptor reflex function during mental stress and relaxation. Psychophysiology, 26, 140–147. Steptoe, A., & Vo¨gele, C. (1990). Cardiac baroreflex function during postural change assessed using non-invasive spontaneous sequence analysis in young men. Cardiovascular Research, 24, 627–632. Vallais, F., Baselli, G., Lucini, D., Pagani, M., & Porta, A. (2009). Spontaneous baroreflex sensitivity estimates during graded bicycle exercise: A comparative study. Physiology Measurement, 30, 201– 213. Watkins, L. L., Fainman, C., Dimsdale, J., & Ziegler, M. G. (1995). Assessment of baroreflex control from beat-to-beat blood pressure and heart rate changes: A validation study. Psychophysiology, 32, 411–414. Yasumasu, T., Reyes del Paso, G. A., Takahara, K., & Nakashima, Y. (2006). Reduced baroreflex cadiac sensitivity predicts increased cognitive performance. Psychophysiology, 43, 41–45.

(Received July 15, 2009; Accepted November 18, 2009)

Psychophysiology, 47 (2010), 1109–1118. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01021.x

Psychological, muscular and kinematic factors mediate performance under pressure

ANDREW COOKE, MARIA KAVUSSANU, DAVID MCINTYRE, and CHRISTOPHER RING School of Sport & Exercise Sciences, University of Birmingham, Birmingham, UK

Abstract It is well established that performance is influenced by pressure, but the underlying mechanisms of the pressureperformance relationship are poorly understood. To address this important issue, the current experiment evaluated psychological, physiological, and kinematic factors as mediators of the pressure-performance relationship. Psychological, physiological, and kinematic responses to three levels of competitive pressure were measured in 23 males and 35 females during a golf putting task. Pressure manipulations impaired putting performance. Self-reported anxiety, effort, and perceived pressure were increased. Heart rate, heart rate variability, muscle activity, and lateral clubhead acceleration were also elevated. Mediation analyses revealed that effort, muscle activity, and lateral acceleration partially mediated the decline in performance. Results confirmed that pressure elicits effects on performance through multiple pathways. Descriptors: Anxiety, Attention, Competition, Golf putting, Movement kinematics, Processing efficiency

According to processing efficiency theory, anxiety has two effects on the central executive component of the attentional system. First, it consumes attentional capacity through worry. When attentional capacity is consumed to the extent that no auxiliary resources remain to retain on-task attention, performance is impaired. Second, it increases effort. Increased effort can enhance performance by mobilizing auxiliary processing resources that increase the amount of attention devoted to a task. A key distinction is made between performance effectiveness (i.e., the quality of performance) and efficiency (i.e., effectiveness divided by effort expended). As performance effectiveness can be maintained by compensatory increases in effort, anxiety is proposed to impair efficiency more than effectiveness. Williams, Vickers, and Rodrigues (2002) examined processing efficiency theory in sport. Table tennis players were required to execute shots towards targets under high and low states of anxiety, where high anxiety was created by evaluation and competition. Results indicated that performance effectiveness, indexed by shot accuracy, was impaired with increased anxiety. Moreover, increased self-reported effort, probe reaction time, and number of eye fixations during ball tracking were noted with high anxiety, providing evidence that anxiety reduced processing efficiency. A similar finding was made by Wilson, Smith, Chattington, Ford, and Marple-Horvat (2006). They demonstrated that increased anxiety had a detrimental effect on driving performance, and processing efficiency, indexed by self-report and eye-gaze measures. These studies are compatible with the explanation of impaired performance that is offered by processing efficiency theory. Support for processing efficiency theory’s account of maintained performance under pressure was recently offered by Wilson, Smith, and Holmes (2007). They assessed the putting

Pressure to perform well is elicited by situational incentives such as social comparison, evaluation, and rewards for success (Baumeister & Showers, 1986). As these factors are common features of sporting competition (Martens, 1975), the relationship between pressure and performance has been frequently examined in the sport domain (e.g., Beilock & Carr, 2001; Gucciardi & Dimmock, 2008; Jackson, Ashford, & Norsworthy, 2006). However, the precise mechanisms that underlie changes in motor performance under pressure remain a source of debate (Beilock & Gray, 2007). Processing Efficiency Theory Processing efficiency theory (Eysenck & Calvo, 1992) offers a psychological explanation of changes in performance under pressure in sport (Wilson, 2008; Woodman & Hardy, 2001). The theory attributes these changes to the effects of anxiety, which accompanies pressure (Mullen, Hardy, & Tattersall, 2005), on our limited attentional capacity. It is based on Baddeley’s (1986) tripartite model of working memory, which consists of a limited capacity control center (central executive), a subsystem for verbal information processing (phonological loop), and a subsystem for visual and spatial information processing (visuospatial sketchpad). The main effects of anxiety are purportedly on the central executive, which is responsible for active processing and selfregulatory functions (e.g., performance monitoring and strategy selection) (Eysenck, Derakshan, Ramos, & Calvo, 2007).

Address reprint requests to: Andrew Cooke, School of Sport & Exercise Sciences, University of Birmingham, Edgbaston, Birmingham, UK B15 2TT. E-mail: [email protected] 1109

1110 accuracy of golfers under high- and low-pressure conditions while measuring anxiety, effort, and heart rate variability. Heart rate variability is often assessed in low (0.02–0.06 Hz), mid (0.07– 0.14 Hz), and high (0.15–0.50 Hz) frequency bands of the heart rate spectrum (Mulder, 1992). Changes in heart rate variability are caused by variations in parasympathetic and/or sympathetic neural influences and primarily reflect the influence of temperature (low-frequency band), blood pressure (mid-frequency band), and respiration (high-frequency band) on cardiac control (Jorna, 1992). Reduced heart rate variability in the midfrequency band can also indicate increased mental effort (Mulder, 1992). Wilson et al. (2007) found that self-reported effort and anxiety increased in the high-pressure condition but heart rate variability in the mid-frequency band and putting accuracy remained unchanged. The absence of a reduction in heart rate variability may have been due to respiratory changes under pressure that were not assessed. Consistent with processing efficiency theory, increased self-reported effort allowed performance effectiveness to be maintained despite increased anxiety in the high-pressure condition. However, as greater effort was required to achieve the same level of performance in the high-pressure condition as was achieved in the low-pressure condition, processing efficiency was reduced. The aforementioned studies show that changes in performance under pressure are consistent with the predictions of processing efficiency theory (Wilson, 2008). Neuromuscular and Kinematic Effects of Pressure Beyond the boundaries of processing efficiency theory, other variables emerge as candidates that could further explain the effects of pressure on motor performance. For example, muscle tension can increase with increased pressure (e.g., Duffy, 1932), potentially disrupting motor performance. To test this suggestion, Weinberg and Hunt (1976) measured anxiety and electromyographic activity while participants threw a tennis ball at a target. After completing half of the trials (low-pressure condition), participants received failure feedback (high-pressure condition). Results showed that participants who experienced high levels of anxiety contracted their agonist (biceps) and antagonist (triceps) muscles for longer than the low-anxious participants during the high-pressure condition. This indicated reduced neuromuscular efficiency with elevated anxiety and pressure. Participants low in anxiety improved their performance, assessed by throwing accuracy, in the high-pressure condition. Performance did not change in those high in anxiety. Reduced neuromuscular efficiency appeared to be the factor that prevented the high-anxiety participants recording a performance improvement. This finding indicates that increased muscle tension under pressure could have an important influence on performance. Recent climbing studies have also adopted a multi-disciplinary approach to understanding the effects of pressure on performance by assessing the kinematics of movement (Pijpers, Oudejans, & Bakker, 2005; Nieuwenhuys, Pijpers, Oudejans, & Bakker, 2008). They found that novice climbers took longer to traverse a wall at heights in excess of four meters (high-pressure condition), and made more and longer-lasting reaching movements, compared to an identical traverse at heights of less than half a metre (low-pressure condition). Movement kinematics have also been analyzed in studies of golf putting. A one-dimensional analysis of the kinematics of the back and forth movement of the club allowed less accurate putting under conditions where demands on working memory resources were high (as occurs when one is anxious) to be attributed to increased jerkiness and

A. Cooke et al. decreased smoothness during the downswing (Maxwell, Masters, & Eves, 2003). However, no effects of pressure on the kinematics of the putting stroke were noted by Mullen and Hardy (2000) following a two-dimensional kinematic analysis that considered both club and arm movement. Taken together, these studies provide some evidence that pressure is associated with less efficient movement kinematics. Disruption to movement kinematics could provide another mechanism through which pressure exerts detrimental effects on performance. The Present Study The literature reviewed above suggests that pressure can exert effects on performance through a variety of pathways. However, previous research should be interpreted in light of some methodological limitations. First, although research has documented that pressure affects performance and produces changes in psychological (e.g., anxiety, effort), physiological (e.g., muscle activity) and kinematic (e.g., acceleration) variables, it has not established that changes in these variables cause the change in performance. This could be tested using mediation analyses (cf., Wilson, Chattington, & Marple-Horvat, 2008). Second, although several studies have assessed psychological, physiological, and movement-related effects of pressure in isolation, few studies (e.g., Williams et al., 2002; Wilson et al., 2006) have assessed them simultaneously. A multidisciplinary approach has been recommended to provide fuller insights into the mechanisms that underlie impaired performance under pressure in sport (Beilock & Gray, 2007; Nieuwenhuys et al., 2008). Third, previous golf putting studies have examined kinematics of the putting stroke in one-dimension (Maxwell et al., 2003) and two-dimensions (Mullen & Hardy, 2000), but none have characterized the effects of pressure in three-dimensions (i.e., back-andforth, lateral, and vertical axes of clubhead movement). In light of these limitations, the present study aimed to concurrently examine psychological, physiological, and three-dimensional kinematic responses to as well as performance under multiple levels (low, medium, high) of competitive pressure. The study is the first to investigate the effects of multiple levels of pressure on behavior in sport, with previous research comparing only two levels of pressure (i.e., low versus high). It was hypothesized that the experimental manipulations of pressure would elicit increases in anxiety and effort (Eysenck & Calvo, 1992). Pressure was also expected to decrease heart rate variability in the mid-frequency band (Mulder, 1992) and increase heart rate (Woodman & Davis, 2008), cardiovascular indices of increased effort, and anxiety, respectively. In addition, it was predicted that pressure would increase muscle activity (Weinberg & Hunt, 1976) and disrupt kinematics (Maxwell et al., 2003), which are indices of inefficient movements. The primary purpose was to examine whether increased pressure is associated with changes in performance and to formally evaluate possible causes using mediation analyses. Based on processing efficiency theory, we expected effort to mediate performance, should it improve with pressure, and anxiety to mediate performance, should it deteriorate with pressure. Moreover, we hypothesized that neuromuscular and kinematic changes would prove additional mediators of performance if it was impaired with increased pressure. Method Participants Male (n 5 23) and female (n 5 35) right-handed undergraduate students participated in the experiment in exchange for course credit. Participants (M age 5 19.6 years, SD 5 1.2 years) were

Pressure and performance enrolled in a sport science degree program. All were novice golfers with no formal playing experience or official handicap. Informed consent was obtained from all participants. Equipment A standard length (90 cm) steel-shafted blade style golf putter (Sedona 2, Ping, Phoenix, AZ) was used to putt regular-size golf balls (diameter 5 4.27 cm) towards a half-size hole (diameter 5 5.5 cm; depth 5 2.8 cm). A half-sized hole was adopted to increase the accuracy demands of the task. The hole was located 1.5 m from the end and 0.7 m from the side of a 7 m long ! 1.4 m wide strip of a green putting mat (Patiograss). The putting surface had a Stimpmeter reading of 4.27 m, making it faster than the greens at most American golf courses, which, according to the US Golf Association (n.d.), generally range from 2.13 m to 3.66 m. Design This experiment employed one within-subjects factor, pressure condition, with three levels: low, medium, and high. Participants completed five blocks of six putts. Blocks one and two helped acclimatization with the task demands, whereas blocks three, four, and five represented comparatively low-, medium-, and high-pressure conditions, respectively. Performance Measures Both the number of putts successfully holed in each condition and mean radial error (i.e., the mean distance of the balls from the hole) were recorded as measures of performance (Mullen & Hardy, 2000). Zero was recorded and used in the calculation of mean radial error on trials where the putt was holed (Hancock, Butler, & Fischman, 1995). Psychological Measures State anxiety. Anxiety was measured using the 5-item cognitive and 7-item somatic anxiety subscales of the Competitive State Anxiety Inventory–2 Revised (Cox, Martens, & Russell, 2003). Participants were asked to indicate how they felt ‘right now’ in relation to each item, for example, ‘‘I am concerned about performing poorly’’ and ‘‘I feel jittery.’’ Each item was scored on a 4-point Likert scale (1 5 not at all, 2 5 somewhat, 3 5 moderately so, 4 5 very much so). The item responses were averaged and multiplied by ten to provide one score for each subscale. Reliability coefficients of .83 for cognitive anxiety, and .88 for somatic anxiety were reported by Cox et al. (2003), showing a good level of internal consistency. In this experiment, alpha coefficients across the three pressure conditions were good for both cognitive (all .80) and somatic (.75 to .92) anxiety. Effort. Effort was measured using the Rating Scale for Mental Effort (Zijlstra, 1993). Participants were instructed to rate the level of mental effort that they expended using a vertical axis scale ranging from 0–150, with nine category anchors, including, at the extremes, 3 (no mental effort at all) and 114 (extreme mental effort). The scale has acceptable test–retest reliability, with a correlation coefficient of .78 (Zijlstra, 1993), and has been used previously to assess mental effort in sport (e.g., Wilson et al., 2007). In this experiment, the correlation coefficients among the three pressure conditions ranged from .73 to .85. Physiological Measures Cardiac activity. To assess heart rate and heart rate variability, an electrocardiogram was measured using three silver/silver

1111 chloride spot electrodes (Cleartrace, ConMed, Utica, NY) in a modified chest configuration. The electrocardiographic signal was amplified (ICP511, Grass), filtered (1–100 Hz), and digitized at 2500 Hz with 16-bit resolution (Power 1401, Cambridge Electronic Designs, Cambridge, UK) using Spike2 software (Cambridge Electronic Design). Heart rate and two-time domain indices of heart rate variability (SDNN, r-MSSD) for each condition were derived from the intervals between R-waves of the electrocardiogram. The SDNN (standard deviation of R-wave to R-wave [R-R] intervals) and r-MSSD (root mean square of successive R-R intervals) are correlates of the frequency domainbased mid (0.07–0.14 Hz) (SDNN, r 5 .84, po.0001) and high (0.15–0.40 Hz) (r-MSSD, r 5 .97, po.0001) spectral power bands, respectively (Carrasco, Gaitan, Gonzalez, & Yanez, 2001). The time domain correlates of the frequency bands were assessed due to the brevity of each recording epoch (block of putts, low pressure M 5 163 s, SD 5 17 s, medium pressure M 5 161 s, SD 5 19 s, high pressure M 5 234 s, SD 5 20 s), which, coupled with slow heart rates of some participants, left insufficient R-R intervals to perform spectral analyses. Muscle activity. Electromyographic activity of the extensor carpi radialis and biceps brachii muscles of the left arm was recorded; these muscles were chosen based on previous studies implicating them in the putting stroke (Smith, Malo, Laskowski, Sabick, Cooney, et al., 2000; Stinear, Coxon, Fleming, Lim, Prapavessis, & Byblow, 2006). Muscle activity was measured using single differential surface electrodes (DE 2.1, Delsys, Boston, MA) and an amplifier (Bagnoli-2, Delsys) with a ground electrode on the collar bone. Electromyographic signals were amplified, filtered (20–450 Hz), and digitized (2500 Hz). The electromyographic signal for each trial was rectified, and the mean amplitudes (microvolts) were calculated by averaging the activity over four consecutive periods: pre-initiation baseline, upswing, downswing, and post-contact follow-through. The duration of these periods was calculated from the Z-axis acceleration profile (described below). The upswing lasted from movement initiation until the top of the swing; the duration of the pre-initiation baseline was the same as the duration of the upswing. The downswing lasted from the top of the swing until ball contact; the duration of the post-contact follow-through was the same as the duration of the downswing. The trial values in each condition were averaged to provide a condition mean value for each electromyographic variable. Kinematic Measures Movement kinematics. Acceleration of the clubhead in three axes was recorded using a tri-axial accelerometer (LIS3L06AL, ST Microelectronics, Geneva, Switzerland). Acceleration on the X, Y, and Z axes corresponded to lateral, vertical, and back-andforth movement of the clubhead, and assessed clubhead orientation, clubhead height, and impact velocity, respectively. The signals were conditioned by a bespoke buffer amplifier with a frequency response of DC to 15 Hz. Both accelerometer and amplifier were mounted in a 39 mm ! 20 mm ! 15 mm plastic housing secured to the rear of the putter head. A microphone (NT1, Rode, Silverwater, Australia) connected to a mixing desk (Club 2000, Studiomaster, Leighton Buzzard, UK) was used to detect the putter-to-ball contact on each trial. These signals were digitized at 2500 Hz. A computer program determined movement kinematics for each putt from the onset of the downswing phase of the putting stroke until the point of ball contact. The

1112 average acceleration was calculated for the X, Y, and Z axes. Impact velocity, root mean square jerk, and smoothness were also calculated for the Z axis as the primary axis involved in putting (see Maxwell et al., 2003). The values from the six trials in each condition were averaged to provide a condition mean value for each kinematic variable. Manipulations Based on Baumeister and Showers (1986) research concerning the additive psychological factors that produce pressure (e.g., competition, rewards, social evaluation), we created comparatively low, medium, and high-pressure conditions, respectively. Low pressure. Participants were informed that blocks one, two, and three comprised a study to compare putting with a black-and-white ball versus a standard white ball. In block one, participants putted a standard white ball on odd-numbered trials and a black-and-white ball on even-numbered trials. In block two, they only putted a black-and-white ball. In block three, they only putted a white ball. Because we wanted to encourage participants to take the same approach to putting as golfers, they were asked to ‘‘try and get the ball ideally in the hole, but if unsuccessful, to make it finish as close to the hole as possible’’ prior to putting in each block. It was further explained that performance was to be recorded as the average distance that putts finished from the hole, with any holed putts counting as 0 cm. However, to minimize any pressure that may have been elicited by evaluation from the experimenter, participants were told that, although the distance that each putt finished from the hole would be recorded, individual performance would not be analyzed in this condition. Instead, they were informed that the data from all participants would be pooled to generate one accuracy score for the black-and-white ball and one accuracy score for the white ball. In reality, we did not compare the painted and standard balls; this dummy study was designed to minimize any pressure elicited by evaluation. Data from block three constituted the lowpressure condition. Medium pressure. The fourth block of putts constituted the medium-pressure condition. Prior to beginning this condition, participants were reminded that the next two blocks comprised the competition phase of the study. They were informed that, from now on, their accuracy would be directly compared with the performance of the other participants. To emphasize the evaluative nature of this condition, participants were told that the rank-ordered mean radial error performances of the whole sample would be displayed on a noticeboard and e-mailed to all participants. They were reminded again that, ideally, they should try and get the ball in the hole, because holed putts count as 0 cm in calculating the mean radial error performance score. If unsuccessful, they were told to try to make the ball finish as close to the hole as possible. Given that evaluative conditions with financial incentives have been used previously to induce pressure (e.g., Wilson et al., 2007), participants were informed that cash prizes of d25, d20, d15, d10, and d5 would be awarded to the top five performers at the end of the study. Further, they were made aware that they would be competing against ‘‘about fifty’’ other participants to allow them to evaluate their chances of winning (i.e., 10% would win a prize). High pressure. The fifth block of putts comprised the highpressure condition. To further increase pressure, we first introduced rewards and punishments. Participants were informed that

A. Cooke et al. they now had a chance to win an additional d12, to be awarded at the end of the session. Specifically, they were told that a d6 reward could be earned if their average distance from the hole was less than the average distance achieved by a yoked rival competitor from a previous study. In reality, this standard was their own average distance achieved in the low-pressure condition. Pilot testing indicated that this standard was perceived as both achievable and credible. Participants were verbally told this distance and visually shown it using a tape measure. It was again emphasized that they should try and get the ball ideally in the hole to give them the best chance of beating the target distance. If unsuccessful, they were to try to make the ball finish as close to the hole as possible. Following the target distance information, a stack of six d1 coins was placed in prominent view. Participants were then told that an additional d6 could also be earned in bonuses: d1 per holed putt, which was added to the stack of coins. Penalties could be incurred: d1 was removed from the stack for every putt that was worse than their yoked rival’s worst putt (in reality, this distance was the participant’s own mean plus half of their standard deviation from the low-pressure condition). Participants were also verbally told and visually shown this distance. Finally, it was emphasized that participants would lose the entire stack of coins, including any bonuses, if they failed to better their yoked rival’s mean distance. Expectations of success and self-awareness were also introduced to further increase pressure. To increase expectations of success, participants were told that one out of two participants had won money in this condition; this information was valid with 47% winning some money (M 5 d7.15, minimum 5 d4, maximum 5 d9). To increase self-awareness, a video camera (Sony, Tokyo, Japan), fitted with a wide-angled lens and light, was turned on (Buss, 1980). Full body images of the participant were prominently displayed on a color television monitor (Panasonic, Osaka, Japan) located over the mat about 50 cm above the hole. Manipulation Check To check the effectiveness of the experimental manipulations in creating multiple levels of pressure, participants completed the 5item pressure/tension subscale from the Intrinsic Motivation Inventory (Ryan, 1982). Participants were asked to rate items, including ‘‘I felt pressured’’ and ‘‘I felt anxious’’ on a 7-point Likert scale (1 5 not at all true, 4 5 somewhat true, 7 5 very true). The item responses were averaged to provide one score for the scale. McAuley, Duncan, and Tammen (1989) reported an internal constancy of .68 for this subscale. In this experiment, Cronbach’s alpha coefficients across the three pressure conditions were good to very good (.74 to .89). Moreover, participants completed single-item ratings to assess how competitive, engaging, and important they considered each condition using a 7point Likert scale (0 5 not at all, 6 5 very much so) (Veldhuijzen van Zanten, DeBoer, Harrison, Ring, Carroll, et al., 2002). This was to verify that the conditions were competitive and to ensure that participants were engaged by and valued the importance of the putting tasks, as would be expected in pressured sporting competition (cf., Hardy, Beattie, & Woodman, 2007). Procedure The protocol was approved by the local research ethics committee. First, participants were instrumented to allow the recording of physiological measures. Then, they completed a putting task comprising five blocks of six putts. Putts were made from three distances to reduce familiarity with one particular distance (e.g.,

Pressure and performance Wilson et al., 2007). The same counterbalanced order of putting distances was fixed for each block: 1.8, 1.2, 2.4, 2.4, 1.2, and 1.8 m. The first two blocks served to acclimatize participants with the task. The last three blocks represented the experimental conditions: low pressure (block three), medium pressure (block four), and high pressure (block five). The order of these conditions was fixed to minimize threats to validity (e.g., reduced motivation during low pressure) when high pressure precedes low pressure (Beilock & Carr, 2001). Prior to putting in each of the three pressure conditions, the manipulation took place and then state anxiety was measured. Putting followed, and immediately after the final putt of each condition, retrospective measures of pressure and effort were obtained. Physiological measures were recorded continuously during each block. Successful putts were recorded immediately on a data sheet. The position of any unsuccessful putt was marked by placing a sticker on the mat. The ball then was retrieved by the experimenter and placed at the next position. At the end of blocks 1–4, the distance of each marker from the hole was recorded and the marker removed from the mat. In block 5, the distance of the ball from the hole was measured after each putt to enable a d1 coin to be added to or removed from the stack. At the end of the session, participants were questioned regarding any ball preference (i.e., white versus black-and-white). Their answers indicated that the low-pressure manipulation was credible; none guessed that the dummy part of the study was false. They were then debriefed, and where appropriate, paid prize money. At the end of the study, the leaderboard was displayed, and the top five performers were paid their additional prize money. Statistical Analysis Analysis of the performance data revealed that two participants putted at least one ball off the mat during the three pressure conditions and thus these outliers were excluded. In addition, physiological data from two participants were unscorable. Thus, the effective sample size for the analyses reported comprised 22 male and 32 female participants. As our first step, we used repeated measures multivariate analysis of variance (MANOVA) to examine the effects of pressure condition (low, medium, high), our within-subject factor, on the performance, psychological, cardiac, and kinematic variables. This revealed a significant multivariate effect for pressure condition, F(36,18) 5 13.09, po.001, Z2 5 .96. A separate repeated measures MANOVA was employed to examine the effects of pressure condition and swing phase, both within-subject factors, on muscle activity. This 3 Pressure Condition ! 4 Swing Phase (pre-initiation, upswing, downswing, post-contact) MANOVA yielded significant multivariate effects for condition, F(4,50) 5 3.55, po.05, Z2 5 .22, and phase, F(6,48) 5 5.12, po.001, Z2 5 .39, but no condition ! phase interaction, F(12,42) 5 1.76, p 5 .09, Z2 5 .34. These significant multivariate effects were followed by separate ANOVAs for each variable, which are presented in the results. Significant ANOVA effects were followed by least significant difference (LSD) post-hoc comparisons. As our final step, we used ANCOVA to assess mediation. Mediation can be tested using a variety of methods including regression, structural equation modelling, and ANCOVA (MacKinnon, 2008). We chose the ANCOVA approach because it is recommended for experimental designs where the sample size is

1113 low and/or a within-subjects design is employed (Hoyle & Robinson, 2004). ANOVA and ANCOVA produce identical results to and are conceptually equivalent to regression analyses (Cohen, Cohen, West, & Aiken, 2003). We used the multivariate method for reporting the results of our analyses. The multivariate method is used to minimize the risk of violating sphericity and compound symmetry assumptions in repeated measures ANOVA designs (Vasey & Thayer, 1987); its use can be identified by the reduced degrees of freedom reported. Wilks’ Lamda, the multivariate statistic (which is not reported), equals 1 – Z2. Partial eta-squared (Z2) is reported as the effect size. In ANOVA, this equals the adjusted R2 obtained in regression analyses (Tabachnick & Fidell, 2001). Values of .02, .13, and .26 for Z2 indicate small, medium, and large effect sizes, respectively (Cohen, 1992). Results Manipulation Check Separate ANOVAs confirmed main effects for pressure condition on perceived pressure and self-reported ratings of importance, engagement, and competitiveness associated with each block of putts (see Table 1). Post-hoc comparisons showed that perceived pressure and ratings were greatest in the high-pressure condition and smallest in the low-pressure condition, confirming that our pressure manipulations, as expected, created three distinct levels of pressure. Effects of Pressure on Performance The effects of the pressure manipulation on putting performance are illustrated in Figure 1. Two separate ANOVAs revealed a pressure condition main effect on the average number of balls holed, F(2,52) 5 3.35, po.05, Z2 5 .11 (Figure 1A), but not the mean radial error, F(2,52) 5 0.00, p 5 1.00, Z2 5 .00 (Figure 1B). Significantly fewer putts were holed during medium- and highthan during the low-pressure condition. Effects of Pressure on Psychological Measures ANOVAs revealed main effects of pressure condition for cognitive anxiety, somatic anxiety, and effort (see Table 1). Post-hoc comparisons showed that anxiety and effort were greatest in the high-pressure condition and smallest in the low-pressure condition. Effects of Pressure on Physiological Measures The ANOVAs indicated a main effect of pressure condition on heart rate and SDNN but not r-MSSD (see Table 1). Post-hoc testing confirmed that heart rate increased from low- to mediumto high-pressure conditions. SDNN increased from low to high pressure. The effects of pressure on muscle activity are displayed in Figure 2. Separate 3 Condition ! 4 Phase ANOVAs were conducted for each muscle. A significant condition effect for extensor carpi radialis activity was revealed, F(2,52) 5 5.18, po.01, Z2 5 .17; muscle activity was greater under high pressure than low pressure. A significant swing phase effect for the extensor carpi radialis activity was also found, F(3,51) 5 9.96, po.001, Z2 5 .37; muscle activity was similar at baseline through the upswing but increased significantly from upswing to downswing with no further increase thereafter. There was no condition ! phase interaction, F(6,48) 5 1.98, p 5 .09, Z2 5 .20. No condition, F(2,52) 5 1.87, p 5 .17, Z2 5 .07, phase, F(3,51) 5 1.43, p 5 .24, Z2 5 .08, or condition ! phase interaction effects,

1114

A. Cooke et al.

Table 1. Mean (SD) of the Measures in Each Pressure Condition Pressure condition Low Measure (range of scores possible) Manipulation check Perceived pressure (1–7) Competitive (0–6) Important (0–6) Engaging (0–6) Psychological Cognitive anxiety (10–40) Somatic anxiety (10–40) Mental effort (0–150) Physiological Heart rate (bpm) SDNN (ms) r-MSSD (ms) Kinematic Y-axis acceleration (m.s ! 2) Z-axis acceleration (m.s ! 2) RMS Jerk (m.s ! 3) Smoothness

Medium

High

M

SD

M

SD

M

SD

F(2,52)

Z2

3.03 2.96 3.09 4.26

1.01 1.62 1.25 0.93

4.09a 4.74a 4.48a 4.41

1.20 1.09 1.18 1.11

4.82a,b 5.33a,b 4.96a,b 4.76a,b

1.23 0.80 1.15 0.89

91.80nnn 66.82nnn 58.37nnn 12.32nnn

.78 .72 .69 .32

18.82 12.78 66.89

5.17 3.39 18.20

23.30a 15.37a 81.80a

5.58 5.06 18.25

26.52a,b 18.18a,b 91.02a,b

5.86 6.32 21.05

55.75nnn 32.41nnn 87.91nnn

.69 .56 .77

79.54 83.92 53.67

12.00 33.81 32.85

80.70a 87.15 58.25

12.41 36.14 41.70

82.67a,b 91.09a 58.13

14.11 33.42 36.77

9.86nnn 4.23n 1.61

.28 .14 .06

0.36 3.59 3.59 43.16

0.12 1.05 0.99 18.08

0.37 3.51 3.52 42.53

0.11 0.79 0.77 15.37

0.35 3.61 3.61 41.58

0.12 0.98 0.95 13.66

0.26 1.25 0.97 0.86

.01 .05 .04 .03

a,b

indicate significant difference from low- and medium-pressure conditions, respectively. po.05; nnnpo.001.

n

A

F(6,48) 5 1.09, p 5 .38, Z2 5 .12, were evident for the biceps brachii. Effects of Pressure on Kinematic Measures Figure 3 presents impact velocity and X-axis acceleration as a function of the pressure manipulation. The ANOVA yielded an effect for pressure condition on X-axis acceleration, F(2,52) 5 8.74, po.001, Z2 5 .25; lateral acceleration was greater in the high-pressure condition than the low- and mediumpressure conditions. No main effects for pressure were revealed for impact velocity, F(2,52) 5 1.39, p 5 .26, Z2 5 .05, or other kinematic measures (see Table 1).

B

Figure 1. (A) Number of balls successfully holed across pressure conditions. (B) Mean radial error scores across pressure conditions. Error bars depict standard error of the means.

Mediation Analysis Mediation analyses were employed to test whether any of the psychological, physiological, or kinematic variables mediated the significant decline in putting performance, indicated by the reduction in the number of balls holed in the medium- and highpressure conditions compared to the low-pressure condition. To establish mediation, four criteria must be satisfied (Baron & Kenny, 1986). First, the independent variable must affect the dependent variable. The main effect of pressure condition on the number of balls holed satisfied this criterion. Second, the independent variable must affect the mediator. Main effects for pressure condition on all psychological variables as well as heart rate, heart rate variability (SDNN), X-axis acceleration, and extensor carpi radialis activity were found satisfying this criterion. The final two criteria are that the mediator must affect the dependent variable and that the effect of the independent variable on the dependent variable must be reduced in the presence of the mediator. Any of the psychological, physiological, and kinematic variables that satisfy these final criteria can be considered mediators of the observed breakdown in performance under pressure. The change in Z2 associated with the pressure condition when each variable is used as a covariate indicates the importance

Pressure and performance

A

B

Figure 2. (A) Extensor carpi radialis muscle activity during each phase of the swing. (B) Bicep brachii muscle activity during each phase of the swing. Error bars depict standard error of the means.

of that variable in explaining the effects of pressure condition on putting performance (see Tabachnick & Fidell, 2001). Repeated measures ANCOVA, with each potential mediator variable as the changing covariate, and pressure condition as a within-subjects factor were conducted to simultaneously test whether the last two criteria were met. The original analysis indicated that 11% of the variance in number of balls holed could be attributed to pressure condition; this corresponds to a medium-size effect for pressure on performance. This effect did not survive the covariate adjustment for effort, F(2,50) 5 1.32, p 5 .28, Z2 5 .05, extensor carpi radialis activity, F(2,50) 5 2.15, p 5 .13, Z2 5 .08, and X-axis acceleration, F(2,50) 5 1.46, p 5 .24, Z2 5 .06. These variables partially mediated the effect of pressure on performance; mediation was partial as a small-to-medium effect size remained. In sum, mediation analyses indicated that impaired putting performance under pressure was explained in part by changes in multiple processes. Discussion An understanding of the causes of pressure-induced changes in performance is needed to better inform individuals about how to perform optimally and protect against performance breakdown

1115

A

B

Figure 3. (A) X-axis acceleration across each pressure condition. (B) Impact velocity across each pressure condition. Error bars depict standard error of the means.

under pressure. This study concurrently examined psychological, physiological, and kinematic responses to multiple levels of competitive pressure. Building on previous research, the primary purpose was to formally test potential mechanisms explaining changes in performance under pressure. Manipulation checks confirmed that ratings of pressure, competitiveness, importance, and engagement were greatest in the high-pressure condition and smallest in the low-pressure condition. Accordingly, the pressure manipulations created increasingly important competitions that pressured and engaged participants, as per real-life sporting competition (cf., Hardy et al., 2007). Effects of Pressure on Performance Pressure had different effects on our two measures of performance. In terms of our outcome measure, number of putts holed, pressure had a detrimental effect on performance: Fewer putts were holed in the medium- and high-pressure conditions than in the low-pressure condition. This finding is broadly consistent with some (e.g., Wilson et al., 2006) but not all (e.g., Mullen & Hardy, 2000) studies that have used outcome measures of performance to examine the pressure–performance relationship in sport. It also resonates with anecdotal evidence. For example, Scott Hoch faced the pressure of having a simple putt to win the 1989 United States Masters Golf Championship on two occasions, but missed both times.

1116 In terms of our accuracy measure, mean radial error, pressure had no effect on performance. This finding is also consistent with some (e.g., Wilson et al., 2007) but not all (e.g., Williams et al., 2002) studies that have adopted accuracy measures of performance. It seems that adopting a reduced hole size made our outcome measure of performance more sensitive to the effects of pressure than our accuracy measure (cf. Mullen & Hardy, 2000). Specifically, the reduced-size hole minimized room for error if a successful outcome (i.e., holing a putt) was to be achieved. In contrast, our accuracy measure, mean radial error, was more likely to tolerate minor errors in aim because, although these would prevent putts from being holed, minor errors would allow the ball to finish in close proximity to the target. The mechanisms which underlie the detrimental effect of pressure on number of putts holed are discussed in the following sections. Psychological Mediators of Performance Consistent with our hypothesis, pressure increased anxiety. Processing efficiency theory attributes impaired performance under pressure to cognitive anxiety overloading attentional capacity. Contrary to this claim, increases in cognitive anxiety caused by our pressure manipulations did not mediate the reduction in putts holed under increased pressure. It is possible that cognitive anxiety is a symptom of pressure rather than a causal variable in the relationship between stress and performance (Hardy & Hutchinson, 2007). However, it is important to point out that anxiety was assessed pre-performance, in accordance with the standard use of the CSAI-2R. Perhaps a measure of anxiety that reflects feelings during performance, such as the mental readiness form (Krane, 1994), is needed to explain changes in performance. In line with the vast majority of previous research, increased pressure was also associated with greater self-reported effort (for review, see Wilson, 2008). Processing efficiency theory contends that additional effort should maintain or improve performance. Performance, in terms of mean radial error, was maintained, but not improved, with increased anxiety and pressure. Thus, mediational analyses could not evaluate the proposed beneficial role of effort on performance (Baron & Kenny, 1986). Surprisingly, when applied to our outcome measure of performance (balls holed), mediation analyses revealed that increased effort was in part responsible for the observed performance deterioration under pressure. Reinvestment and explicit monitoring theories (Beilock & Carr, 2001; Masters & Maxwell, 2008) may offer an explanation for this finding. These theories propose that additional effort under pressure can be detrimental to performance as it prompts the reinvestment of explicit knowledge to consciously guide action. This can lead to a regression from a more efficient and autonomous to a less efficient and more cognitive processing strategy. Given that our participants were novices, one might assume that they did not have explicit knowledge of the skill. However, Poolton, Masters, and Maxwell (2007) have argued that explicit knowledge is accrued when individuals make conscious attempts to identify and eradicate performance errors. Moreover, they demonstrated that novices required few trials to generate such explicit knowledge. Accordingly, by missing putts during the first three blocks of trials, our novice participants would have generated explicit knowledge that they could have reinvested in the subsequent medium- and high-pressure conditions. In support of this suggestion, evidence of improvement was noted across the first three blocks of putts. Specifically, mean radial error decreased from block one (M 5 49.74 cm) to block two (M 5 38.82

A. Cooke et al. cm) to block three (M 5 29.99 cm), the low-pressure condition. The mean number of balls holed was similar across blocks one (M 5 0.63) and two (M 5 0.43), before improving in the subsequent low-pressure condition (M 5 1.02). Thus, impaired performance observed during increased pressure could reflect reinvestment of explicit knowledge. Physiological Mediators of Performance As expected, heart rate, a physiological index of arousal and/or anxiety, increased with pressure. Although this increase was relatively small, our finding supports most previous studies (Pijpers et al., 2005; Nieuwenhuys et al., 2008). The small size of the cardiac reaction to pressure can be explained by the effect of posture on heart rate; stress-induced cardiac reactivity is blunted when participants stand (Veldhuijzen van Zanten, Thrall, Wasche, Carroll, & Ring, 2005). Increased heart rate did not mediate the effects of pressure on performance. Decreased heart rate variability in the mid-frequency spectral band is a putative psychophysiological measure of effort (Mulder, 1992). SDNN, a correlate of heart rate variability in the mid-frequency band, increased rather than decreased under conditions of increased pressure. This finding may reflect an increase in respiratory volume under pressure (Jorna, 1992). However, if this were the case then a more substantial increase in variability would be expected in the high-frequency band that is most sensitive to respiratory changes. No change in r-MSSD, a correlate of heart rate variability high-frequency, was noted. That selfreported effort increased but heart rate variability did not decrease mirrors the finding of Wilson et al. (2007). It is possible that the postural and physical demands of the golf putting task, although minimal, override the effects of mental activity on cardiovascular measures (cf. Veldhuijzen van Zanten et al., 2005). These findings argue against the use of heart rate variability as a reliable measure of mental effort during sport. Muscular activity was hypothesized to be augmented by increased pressure. Extensor carpi radialis muscle activity was elevated by pressure, concurring with Weinberg and Hunt’s (1976) observation that pressure increased muscle activity. Activity in this muscle also increased as a function of swing phase. This can be explained by a cumulative increase in the number of active motor units due to new recruitment in each successive movement phase. As the velocity at which the ball was struck remained unchanged, it seems that the increase in overall muscle activity in the high-pressure relative to the low-pressure condition was a result of tighter gripping of the club rather than changes to the dynamics of the swing. In support of this argument, Smith et al. (2000) reported higher grip force in line with increased forearm extensor muscle activity in golfers prone to impaired putting under pressure. Increased extensor carpi radialis muscle activity partially mediated the deterioration in performance under pressure, a finding consistent with anecdotal evidence, specifically, the suggestion that golfers ‘tense up’ when they miss putts under pressure. No effects were found for biceps brachii activity. The lack of an effect of pressure on bicep activity may be attributed to participants over-contracting this muscle in the low pressure condition, thus clouding the effects of increased pressure in the subsequent conditions. In support of this contention, Smith et al. (2000) and Stinear et al. (2006) showed that experienced golfers activated their extensor carpi radialis more than their biceps brachii when putting. In our study of novices, overall extensor carpi radialis activity (M 5 26.81 mV) and biceps brachii activity

Pressure and performance

1117

(M 5 30.35 mV) did not differ (p 5 .59). Thus, it seems likely that our novice participants displayed inefficient muscle activity in the bicep even before the introduction of pressure. The lack of differential activity across swing phase suggests that this muscle was not activated by the putting stroke as executed by novice golfers. Kinematic Mediators of Performance Our hypotheses that movement kinematics would be disrupted by the pressure manipulations and that changes in kinematics would mediate impaired performance were supported. Lateral acceleration of the clubhead increased with pressure and partially mediated impaired performance. This finding provides new insights concerning the kinematic basis for missed putts when performing under pressure. Specifically, additional lateral acceleration implies that the putter was swung out of line during increased pressure. Given that putter face angle and path determine 83% and 17% of ball direction, respectively (Pelz, 2000), greater lateral acceleration would have changed the putter path and should have disrupted the putter face angle. The net result would have been that pressure altered the line on which the ball was struck. This account of missed putts is consistent with what golfers refer to as ‘pulling’ and ‘pushing’ putts wide of the hole when under pressure. That misalignment of the clubhead at impact caused missed putts under pressure is further supported by the observation that impact velocity did not change across the pressure conditions. This finding rules out the possibility that more balls were understruck or overstruck. Limitations of the Study and Directions for Future Research Our results should be interpreted in light of some methodological limitations. First, performance was assessed as the average of six putts, whereas pressure-induced failure in sport is typically a oneoff event (e.g., you may only get one chance to sink a putt to win). However, single-trial performance is characterized by large variability and poor reliability (e.g., Woodman & Davis, 2008). By assessing performance over six putts, we attempted to strike a balance between ecological validity (1 putt; Woodman & Davis,

2008) and measurement reliability (20 putts; Wilson et al., 2007). Nevertheless, the effects of pressure may have been diluted by assessing performance over multiple trials, as participants had chances to redeem bad putts. Second, the number of putts holed did not differ between the medium- and high-pressure conditions. This null finding might be attributed to either an insufficient increase in pressure between conditions or to a floor effect in the number of balls holed. Specifically, performance fell from 1 out of 6 balls holed in the low-pressure condition to 0.6 in the medium-pressure, leaving little room for further deterioration with additional pressure. Future studies could design a task in which performance has more room for deterioration. Third, attentional capacity and allocation of resources were not assessed. This was to help maximize the ecological validity of the pressure manipulations. However, in employing dual-task paradigms or attentional probing tasks, future research could more directly test the predictions of attention-based theories of performance in sport (see Beilock, Carr, MacMahon, & Starkes, 2002; Murray & Janelle, 2003). Finally, we only recruited novice golfers. It is likely that different mechanisms regulate expert performance under pressure (Gray, 2004). Reinvestment theory (Masters & Maxwell, 2008) may offer one such mechanism for impaired performance under pressure in experts who have explicit knowledge of their skill. Future research should examine expert performance in this context. Conclusion By simultaneously assessing psychological, physiological, and kinematic measures, the present study adds to the mounting evidence that pressure concurrently elicits effects at multiple levels of analysis (Pijpers et al., 2005; Nieuwenhuys et al., 2008). In particular, a novel kinematic explanation of missed putts under pressure was outlined. Mediation analyses did not support the predictions of processing efficiency theory as an explanation for failure under pressure in sport. It remains for future research to continue with this promising multi-disciplinary approach and paint a richer picture of the pressure and performance relationship.

REFERENCES Baddeley, A. D. (1986). Working memory. Oxford: Clarendon Press. Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic and statistical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. Baumeister, R. F., & Showers, C. J. (1986). A review of paradoxical performance effects: Choking under pressure in sports and mental tests. European Journal of Social Psychology, 16, 361–383. Beilock, S. L., & Carr, T. H. (2001). On the fragility of skilled performance: What governs choking under pressure? Journal of Experimental Psychology: General, 130, 701–725. Beilock, S. L., Carr, T. H., MacMahon, C., & Starkes, J. L. (2002). When paying attention becomes counterproductive: Impact of divided versus skill-focused attention on novice and experienced performance of sensorimotor skills. Journal of Experimental Psychology: Applied, 8, 6–16. Beilock, S. L., & Gray, R. (2007). Why do athletes choke under pressure? In G. Tenenbaum & R. C. Eklund (Eds.), Handbook of Sport Psychology (3rd ed, pp. 425–444). New Jersey: Wiley. Buss, A. (1980). Self-consciousness and social anxiety. San Francisco: Freeman. Carrasco, S., Gaitan, M. J., Gonzalez, R., & Yanez, O. (2001). Correlation among Poincare plot indexes and time and frequency domain

measures of heart rate variability. Journal of Medical Engineering & Technology, 25, 240–248. Cohen, J. (1992). A power primer. Psychological Bulletin, 112, 155–159. Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple regression/correlation analysis for the behavioral sciences (3rd ed). London: Lawrence Erlbaum. Cox, R. H., Martens, M. P., & Russell, W. D. (2003). Measuring anxiety in athletics: The revised competitive state anxiety inventory-2. Journal of Sport & Exercise Psychology, 25, 519–533. Duffy, E. (1932). The relation between muscular tension and quality of performance. American Journal of Psychology, 44, 535–546. Eysenck, M. W., & Calvo, M. G. (1992). Anxiety and performance: The processing efficiency theory. Cognition and Emotion, 6, 409– 434. Eysenck, M. W., Derakshan, N., Santos, R., & Calvo, M. G. (2007). Anxiety and cognitive performance: Attentional control theory. Emotion, 7, 336–353. Gray, R. (2004). Attending to the execution of a complex sensorimotor skill: Expertise differences, choking, and slumps. Journal of Experimental Psychology: Applied, 10, 42–54. Gucciardi, D. F., & Dimmock, J. A. (2008). Choking under pressure in sensorimotor skills: Conscious processing or depleted attentional resources? Psychology of Sport and Exercise, 9, 45–59.

1118 Hancock, G. R., Butler, M. S., & Fischman, M. G. (1995). On the problem of two-dimensional error scores: Measures and analyses of accuracy, bias, and consistency. Journal of Motor Behavior, 27, 241– 250. Hardy, L., Beattie, S., & Woodman, T. (2007). Anxiety–induced performance catastrophes: Investigating effort required as an asymmetry factor. British Journal of Psychology, 98, 15–31. Hardy, L., & Hutchinson, A. (2007). Effects of performance anxiety on effort and performance in rock climbing: A test of processing efficiency theory. Anxiety, Stress, and Coping, 20, 147–161. Hoyle, R. H., & Robinson, J. C. (2004). Mediated and moderated effects in social psychological research: Measurement, design and analysis issues. In C. Sansone, C. C. Morf, & A. T. Painter (Eds.), The Sage handbook of methods in social psychology (pp. 213–233). Thousand Oaks, CA: Sage. Jackson, R. C., Ashford, K. J., & Norsworthy, G. (2006). Attentional focus, dispositional reinvestment and skilled motor performance under pressure. Journal of Sport & Exercise Psychology, 28, 49–68. Jorna, P. G. A. M. (1992). Spectral analysis of heart rate and psychological state: A review of its validity as a workload index. Biological Psychology, 34, 237–257. Krane, V. (1994). The mental readiness form as a measure of competitive state anxiety. The Sport Psychologist, 8, 189–202. MacKinnon, D. P. (2008). Introduction to statistical mediation analysis. New York: Lawrence Erlbaum. Martens, R. (1975). The process of competition. In R. Martens (Ed.), Social psychology & physical activity (pp. 66–86). New York: Harper & Row. Masters, R. S. W., & Maxwell, J. P. (2008). The theory of reinvestment. International Review of Sport and Exercise Psychology, 1, 160–183. Maxwell, J. P., Masters, R. S. W., & Eves, F. F. (2003). The role of working memory in motor learning and performance. Consciousness and Cognition, 12, 376–402. McAuley, E., Duncan, T., & Tammen, V. V. (1989). Psychometric properties of the Intrinsic Motivation Inventory in a competitive sport setting: A confirmatory factor analysis. Research Quarterly for Exercise and Sport, 60, 48–58. Mulder, L. J. M. (1992). Measurement and analysis methods of heart rate variability and respiration for use in applied environments. Biological Psychology, 34, 205–236. Mullen, R., & Hardy, L. (2000). State anxiety and motor performance: Testing the conscious processing hypothesis. Journal of Sports Sciences, 18, 785–799. Mullen, R., Hardy, L., & Tattersall, A. (2005). The effects of anxiety on motor performance: A test of the conscious processing hypothesis. Journal of Sport & Exercise Psychology, 27, 212–225. Murray, N. P., & Janelle, C. M. (2003). Anxiety and performance: A visual search examination of the processing efficiency theory. Journal of Sport & Exercise Psychology, 25, 171–187. Nieuwenhuys, A., Pijpers, J. R., Oudejans, R. R. D., & Bakker, F. C. (2008). The influence of anxiety on visual attention in climbing. Journal of Sport & Exercise Psychology, 30, 171–185. Pelz, D. (2000). Dave Pelz’s putting bible. New York: Doubleday. Pijpers, J. R., Oudejans, R. R. D., & Bakker, F. C. (2005). Anxietyinduced changes in movement behavior during the execution of a complex whole-body task. The Quarterly Journal of Experimental Psychology, 58A, 421–445. Poolton, J. M., Masters, R. S. W., & Maxwell, J. P. (2007). Passing thoughts on the evolutionary stability of implicit motor behavior: Performance retention under physiological fatigue. Consciousness and Cognition, 16, 456–468.

A. Cooke et al. Ryan, R. M. (1982). Control and information in the intrapersonal sphere: An extension of cognitive evaluation theory. Journal of Personality and Social Psychology, 43, 450–461. Smith, A. M., Malo, S. A., Laskowski, E. R., Sabick, M., Cooney, W. P., Finnie, S. B., et al. (2000). A multidisciplinary study of the ‘yips’ phenomenon in golf. Sports Medicine, 30, 423–437. Stinear, C. M., Coxon, J. P., Fleming, M. K., Lim, V. K., Prapavessis, H., & Byblow, W. D. (2006). The yips in golf: Multimodal evidence for two subtypes. Medicine & Science in Sport & Exercise, 38, 1980– 1989. Tabachnick, B. G., & Fidell, L. S. (2001). Computer-assisted research design and analysis. Needham Heights, MA: Allyn & Bacon. United States Golf Association (n.d.). Stimpmeter instruction booklet. Retrieved September 23, 2009, from http://www.usga.org/course_ care/articles/management/greens/Stimpmeter-Instruction-Booklet/ Vasey, M. W., & Thayer, J. F. (1987). The continuing problem of false positives in repeated measures ANOVA in psychophysiology: A multivariate solution. Psychophysiology, 24, 479–486. Veldhuijzen van Zanten, J. J. C. S., De Boer, D., Harrison, L. K., Ring, C., Carroll, D., Willemsen, G., et al. (2002). Competitiveness and hemodynamic reactions to competition. Psychophysiology, 39, 759– 766. Veldhuijzen van Zanten, J. J. C. S., Thrall, D., Wasche, D., Carroll, D., & Ring, C. (2005). The influence of hydration status on mental and postural stress-induced hemoconcentration. Psychophysiology, 42, 98–107. Weinberg, R. S., & Hunt, V. V. (1976). The interrelationships between anxiety, motor performance and electromyography. Journal of Motor Behavior, 8, 219–224. Williams, A. M., Vickers, J., & Rodrigues, S. (2002). The effects of anxiety on visual search, movement kinematics and performance in table tennis: A test of Eysenck and Calvo’s processing efficiency theory. Journal of Sport & Exercise Psychology, 24, 438–455. Wilson, M. (2008). From processing efficiency to attentional control: A mechanistic account of the anxiety-performance relationship. International Review of Sport and Exercise Psychology, 1, 184–201. Wilson, M., Smith, N. C., Chattington, M., Ford, M., & Marple-Horvat, D. E. (2006). The role of effort in moderating the anxiety-performance relationship: Testing the prediction of processing efficiency theory in simulated rally driving. Journal of Sports Sciences, 24, 1223– 1233. Wilson, M., Chattington, M., & Marple-Horvat, D. E. (2008). Eye movements drive steering: The effect of reduced eye movement distribution on steering and driving performance. Journal of Motor Behavior, 40, 190–202. Wilson, M., Smith, N. C., & Holmes, P. S. (2007). The role of effort in influencing the effect of anxiety on performance: Testing the conflicting predictions of processing efficiency theory and the conscious processing hypothesis. British Journal of Psychology, 98, 411–428. Woodman, T., & Davis, P. A. (2008). The role of repression in the incidence of ironic errors. The Sport Psychologist, 22, 183–198. Woodman, T., & Hardy, L. (2001). Stress and anxiety. In R. N. Singer, H. A. Hausenblas, & C. M. Janelle (Eds.), Handbook of sport psychology (2nd ed, pp. 290–318). New York: Wiley. Zijlstra, F. R. H. (1993). Efficiency in work behavior: A design approach for modern tools. Delft: Delft University Press.

(Received June 25, 2009; Accepted November 29, 2009)

Psychophysiology, 47 (2010), 1119–1126. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01028.x

Manipulation specific effects of mental fatigue: Evidence from novelty processing and simulated driving

STIJN A.A. MASSAR,a ANNE E. WESTER,b EDMUND R. VOLKERTS,b and J. LEON KENEMANSa,b a

Department of Experimental Psychology, Utrecht University, Utrecht, The Netherlands Department of Psychopharmacology, Utrecht University, Utrecht, The Netherlands

b

Abstract Mental fatigue has a wide range of effects on cognitive, behavioral, and motivational measures. It can be expected that specific effects in which fatigue becomes manifest is dependent on the nature of fatigue-inducing activity (e.g., level of control and working memory demands). This study examined how fatigue caused by tasks that differ on the level of working memory demands (0-Back, 2-Back) affects brain function (novelty processing, P3a) and performance (driving). Results showed that fatigue did not affect driving performance. Fatigue did reduce P3a amplitude, but only after 2-Back. P3a was also reduced during driving. The effects of fatigue and driving on P3a were additive. In summary, both driving and fatigue reduced P3a amplitude. Driving effects were always present. Fatigue effects on novelty processing were dependent on the cognitive demands of the fatigue-inducing task. Descriptors: Mental fatigue, Event-related potentials, Novelty-P3a, N-Back task

1989; Miyake, Friedman, Emerson, Witzki, Howerter, & Wager, 2000). Consistently, fatigue leads to problems in planning and coordination of action sequences, disrupting performance mostly in novel or complex tasks (Bartlett, 1943; Hockey, 1993). Mental fatigue can be caused by prolonged engagement in many different cognitive activities. People, for instance, not only report fatigue after tasks that put constant demands on working memory and updating of information (e.g., scheduling task: Van der Linden et al., 2003b), but also after performing stimulusresponse tasks that involve alertness and selective attention but minimal working memory load (e.g., flanker task: Boksem, Meijman, & Lorist, 2006). Research thus far has employed a wide variety of tasks to induce mental fatigue. However, a direct test of how fatigue effects depend on varying task demands is still lacking. Since fatigue is viewed as a depletion of cognitive resources, it seems unlikely that tasks that recruit different levels of cognitive processing will lead to uniform fatigue effects. The current study addresses this question by comparing the effects of fatigue, systematically varying the level of demands that were put on working memory and top-down control. By doing so, we expected to cause different levels of cognitive depletion. Electrocortical and behavioral fatigue effects were measured as novelty P3a event-related potential (ERP) activity (novelty processing) and simulated driving, respectively.

Mental fatigue is a multi-component phenomenon that involves changes in mood, motivation, attention, and cognition. The state of mental fatigue can derive from a variety of causes. Sleep loss, prolonged working hours, and biorhythmical fluctuations can all result in the combined state of inattention, distractibility, loss of goal-directedness, and loss of willingness to exert further effort that is well known as ‘being tired.’ Changes in performance and cognition that accompany fatigue have often been related to the occurrence of errors and accidents, with obvious risks when happening in the context of industry and transportation (Baker, Olson, & Morriseau, 1994; Brown, 1994; Petridou & Moustaki, 2000; Swaen, Van Amelsfoort, Bu¨ltmann, & Kant, 2003). In general, mental fatigue is found to cause slower responding and higher error rates in reaction time tasks. More specifically, people show less efficient preparation (Lorist, Klein, Nieuwenhuis, De Jong, Mulder, & Meijman, 2000) and more perseveration in switch tasks (Van der Linden, Frese, & Meijman, 2003a), less flexibility and a loss of systematic exploration strategies in learning tasks (Van der Linden, Frese, & Sonnentag, 2003b), and higher intrusion of irrelevant information into cognitive processing (Boksem, Lorist, & Meijman, 2005; Van der Linden & Eling, 2006). These performance decrements suggest that mental fatigue causes a depletion of top-down control capacity, leading to declined integrity of goal-directed behavior (Duncan, Emslie, Williams, Johnson, & Freer, 1996; Fuster,

Novelty Processing and Mental Fatigue Novelty processing refers to the detection of and orienting towards novel stimuli in the environment. When novel stimuli are presented, an automatic shift of attention towards those stimuli can take place. In electroencephalogram (EEG) recordings, brain

The authors would like to thank Gert Camfferman for advice and technical assistance. Address correspondence to: Stijn A.A. Massar, Department of Experimental Psychology, Utrecht University, Heidelberglaan 2, 3584 CS, Utrecht, The Netherlands. E-mail: [email protected] 1119

1120

S. A. A. Massar et al.

reactions to such novel stimuli are typically characterized by a frontocentral positivity (P3a) that peaks around 300 ms (Friedman, Cycowicz, & Gaeta, 2001). P3a generation is triggered by bottom-up deviance detection (Debener, Kranczioch, Herrmann, & Engel, 2002). Consequently, P3a is elicited irrespective of whether novelty is presented in or outside the focus of attention (Friedman, Kazmerski, & Cycowicz, 1998), even if attention is highly focused on a primary task in a different modality (MullerGass, Macdonald, Schroger, Sculthorpe, & Campbell, 2007). Despite its bottom-up origin, the amplitude of the P3a has been found sensitive to the amount of attentional resources that are available. P3a was reported to increase when attention is focused on the modality in which the novel stimuli are presented (Combs & Polich, 2006; Muller-Gass & Schro¨ger, 2007), and to decrease when attention is diverted away (Friedman et al., 1998; Holdstock & Rugg, 1995; Zhang, Chen, Yuan, Zhang, & He, 2006). For example, when attentional control and working memory are focused on an unrelated visual task (SanMiguel, Corral, & Escera, 2008), or to response categorization (Berti & Schro¨ger, 2003), P3a amplitude is decreased. On the other hand, P3a is increased when processing of the novel stimulus is included in the focus of working memory (Muller-Gass & Schro¨ger, 2007). It therefore seems that, while P3a is triggered by automatic bottomup processing, its generation can be hindered when top-down attentional resources are less available to evaluate the stimulus. During fatigue, such control resources are especially compromised. Consequently, it can be expected that P3a amplitude is attenuated during a state of mental fatigue, particularly when caused by a task that puts high demands on these control resources. Partial support for this hypothesis is provided by findings that P3a amplitude is reduced after sleep deprivation (closely associated with mental fatigue, Gosselin, De Koninck, & Campbell, 2005), and in patients with sleep-related problems (Gosselin, Mathieu, Mazza, Decary, Malo, & Montplaisir, 2006; but see Salmi, Huotilainen, Pakarinen, Siren, Alho, & Aronen, 2005). The present study explicitly tested the question whether prolonged performance on tasks that involves different levels of control and working memory causes different levels of fatiguerelated P3a reduction. The Present Study In order to examine this question, novelty processing and driving performance were measured before and after performance of a fatiguing task. Fatigue was induced by 90 min of N-Back task performance. In two separate sessions, two versions of the N-Back task were executed (0-Back and 2-Back), thereby systematically varying the demands that were put on working memory and attentional control. It was expected that putting continuous demands on control processes in the 2-Back task (in contrast to 0-Back) would lead to a depletion of cognitive control

day 1

pre-fatigue novelty oddball driving

day 2

non-driving

pre-fatigue novelty oddball driving

non-driving

resources. Consequently, fatigue-related P3a and behavioral decrements were expected to be more pronounced after 2-Back performance than after 0-Back performance, even when subjectively experienced fatigue would not differ between the two conditions. P3a is typically analyzed as the difference wave comparing novel stimulus ERP to a standard stimulus. In the present study, however, it cannot be excluded that, in addition to influencing novelty processing, fatigue also influences the processing of standard stimuli (Muller-Gass & Schro¨ger, 2007). In order to control for this possibility, additional analysis has been conducted on separate novel and standard ERPs. Methods Participants Twelve volunteers (9 males, 3 females; mean age 5 22.2, range 19–28) were recruited from a student population. All were screened to have no hearing problems, normal or corrected-tonormal vision, no sleep or fatigue-related problems, and not to work night shifts. In addition, participants did not take prescription medicine, did not use drugs, and had to restrain from drinking alcohol on the night before the experimental sessions and any caffeinated drinks on the day of the experiment. As a reward for their effort, participants received either course credits or 54 Euros. All participants signed written informed consent before starting the experiments. Procedure Fatigue was induced by 90 min of either 0-Back or 2-Back task performance. By increasing the number of memory elements that need to be remembered and updated, the different levels of the NBack task recruit different levels of working memory and control, and could thereby be expected to induce different states of fatigue. Participants were tested in both fatigue-manipulation conditions (0-Back, 2-Back), which took place in two separate sessions, before and after which driving ability and novelty processing were assessed (see Figure 1). Upon entering the lab, participants filled out informed consent, and the EEG equipment was applied. Participants performed a simulated driving task, during which novel auditory stimuli (unrelated to the driving task) were presented. Secondly, the novel stimuli were also presented when the participants did not perform the driving task. Subsequently, participants performed 90 min of N-Back task in order to induce mental fatigue. After finishing the manipulation task, post-manipulation measurement of driving ability and the novelty processing was done. A questionnaire assessing subjective fatigue was filled out three times during the experiment, before starting the fatigue manipulation task, after finishing the manipulation, and in a follow-up measurement 40 min after the manipulation task. The order in which the 0-Back and the

fatigue manipulation 1 90 min. 0-Back performance

fatigue manipulation 2 90 min. 2-Back performance

post-fatigue novelty oddball driving

non-driving

post-fatigue novelty oddball driving

non-driving

Figure 1. Schedules of the test procedure. On test day 1, participants were subjected to a novelty oddball paradigm, both during driving and non-driving, followed by a 90-min fatigue inducing N-Back task (either 0-Back or 2-Back). Subsequently, a post-fatigue novelty oddball was presented. On test day 2, participants were subjected to the same procedure, with the only difference that the other N-Back task was used as fatigue induction.

Mental fatigue and novelty processing 2-Back sessions were presented, as well as the order of driving versus no-driving oddball presentation within-sessions, were counterbalanced between subjects. Techniques and Materials Subjective Fatigue Assessment Subjective fatigue was measured using the Rating Scale Mental Effort (RSME, Zijlstra, 1993). This scale consists of seven items that index different aspects of fatigue, such as required effort for attention focusing and visual perception, tiredness, and boredom. All items were scored on an analogue scale, and load on one factor, mental fatigue. Fatigue Manipulations Fatigue was induced by 90 min of N-Back performance (0-Back, 2-Back). White letters were presented for 500 ms (with 2500 ms inter-interval trial) in the middle of a black screen using the Experimental Run Time System (ERTS, BeriSoft Corporation, Frankfurt, Germany). On presentation of a target stimulus, participants were required to respond with a right button press and with a left button press on non-target presentation. In the 2-Back condition, targets were defined as any letter that is the same as the letter that was presented two trials prior to the current trial (with non-targets defined as being not the same as two trials earlier). In the 0-Back condition, participants were instructed to react to the letter ‘X’ as a target, and to any other letter as a non-target. Stimuli could be all letters in the alphabet. In both conditions, a total of 1800 stimuli were presented, and target probability was 30%. Passive Oddball Paradigm Auditory stimuli were presented through Victory ms-28 portable loudspeakers attached to the back of the head. In each condition, a total of 520 stimuli were presented, with an inter-stimulus interval of 2.2 s, and an intensity of 75 dB at ear level. 1000 Hz standard tones (80% of the stimuli) were intermixed with deviant tones (1100 Hz, 10% of the stimuli) and novel environmental sounds (10% of the stimuli). The latter stimuli were selected from a database (Fabiani & Friedman, 1995) and had a duration between 161 ms and 403 ms. The duration of standard and deviant tones was 338 ms, which equals the mean duration of the novel stimuli. In total, two sets of novel stimuli were used, such that different stimuli were presented in pre- and post-manipulation measurement. However, within the pre- and post-manipulation measurement the same set was used for driving and non-driving conditions. Also, the sets and the order of presentation were kept constant for the two separate test sessions. Participants were instructed to ignore the sounds. In order to examine effects of attention allocation, two oddball sessions were presented during both pre- and post-manipulation. In one session a simulated driving task was performed as a background task, and in the other session no behavioral task was performed. Duration of oddball presentation was 20 min per session. Simulated Driving The simulated driving environment was presented using the Divided Attention Steering Simulator (DASS, Stowood Scientific Instruments, Oxford, UK). The DASS is a fixed-base steering simulator in which the outlines of a car and a road are presented on a black computer screen. Participants were required to keep the car on the middle of a winding road using a Thrustmaster steering wheel (Guillemot Corporation, Chantepie Cedex,

1121 France), while the car moves forward at a constant speed. The primary outcome measure is the steering error from the middle of the road (standard deviation of lateral position, SDLP). EEG Recording EEG data were recorded from 32 Biosemi active electrodes (Biosemi, Amsterdam, The Netherlands), which were positioned according to standard 10/10 EEG positions. Electrodes were placed in nine lines with outmost lateral positions Fp1/Fp2, AF3/AF4, F7/F8, FC5/FC6, T7/T8, CP5/CP6, P7/P8, PO3/ PO4, and O1/O2. Midline electrodes stretched from Fz to Oz. Electro-oculogram (EOG) electrodes were placed above and below the left eye and on the outer canthi of each eye. Reference electrodes were placed on both mastoids, for offline re-referencing. Biosemi active electrode system uses an active online referencing, through a Common Mode Sense and a Driven Right Leg electrode (CMS/DRL, MettingVanRijn, Peper & Grimbergen, 1990). All data were recorded with a 512 Hz low-pass filter at a sample rate of 2048 Hz, and stored for offline analysis. Data Reduction and Analysis Performance Driving performance was measured by calculating the standard deviation of the lateral position (SDLP) during the 20-min driving task. Data were statistically analyzed with a 2 ! 2 (time ! manipulation-type) repeated measures analysis of variance (ANOVA). ERPs EEG data were analyzed using Brian Vision Analyzer software (Brain Products GmbH, Gilching, Germany). Data were re-referenced offline to the averaged signal of both mastoids, and subsequently filtered with a 0.16 Hz high-pass filter and a slope of 24 dB/oct, a 30 Hz low-pass filter with a slope of 24 dB/oct, and a 50 Hz notch filter. Artefacts were removed semi-automatically and eye movements were corrected using the Gratton & Coles algorithm (Gratton, Coles, & Donchin, 3 1983). Average waveforms for novel and standard stimuli and a novel-standard difference wave were calculated per oddball condition (driving, non-driving), time (pre-, post-manipulation), manipulation condition (0-Back, 2-Back), and site (Fz, Cz, Pz). P3a was quantified as the average amplitude at Fz, Cz, and Pz, in a time area of 50 ms around the peak in the grand average difference wave (300– 350 ms after stimulus presentation). Amplitudes were statistically analyzed using repeated measures ANOVA. Greenhouse-Geisser corrected F values are reported where appropriate. To assess effects on novel processing independently of effects on standard processing, all analyses were repeated on the novel and standard raw waveforms (Muller-Gass & Schro¨ger, 2007). Results Fatigue Manipulation Check Self-reported fatigue was increased directly after the fatigue manipulations, and decreased during follow-up. These effects were present independent of the type of fatigue manipulation. A manipulation-type (0-Back, 2-Back) ! time-of-measurement (prefatigue, post-fatigue, follow-up) repeated measures ANOVA of the RSME data revealed a significant main effect of time-ofmeasurement F(2,10) 5 17.07, po.005. Post-hoc contrasts showed that fatigue was significantly increased after the manipulation (mean pre-fatigue 5 45.4, mean post-fatigue 5 71.9;

1122

S. A. A. Massar et al.

F(1,11) 5 37.53, po.001). In the 40 min, follow-up measurement RSME score was again decreased (mean follow-up 5 52.1; F(1,11) 5 14.08, po.005). The absence of a manipulation-type main effect and a time ! manipulation interaction (p4.5) indicated that the increase in subjective fatigue was the same after 0-Back and after 2-Back performance.

(fatigued), which was more pronounced after the 2-Back manipulation than after the 0-Back manipulation. Secondly, P3a amplitude was decreased during driving. The effects of fatigue and driving were additive, such that the reduction of P3a by driving was present both before and after the fatigue manipulation. To statistically analyze these effects, P3a amplitudes were entered into a 2 ! 2 ! 2 ! 3 repeated measures ANOVA with manipulation-type (0-Back, 2-Back), time (pre-, post-manipulation), driving (driving, non-driving), and site (Fz, Cz, Pz) as factors.

Driving Performance Driving performance was not affected by fatigue in either manipulation condition. A manipulation-type (0-Back, 2-Back) ! timeof-measurement (pre-fatigue, post-fatigue) repeated measures ANOVA for SDLP did not show any significant interactions or main effects (Ftime(1,11) 5 .02, p 5 .88; Fmanipulation(1,11) 5 .12, p 5 .74; Ftime ! manipulation(1,11) 5 1.02, p 5 .33).

Fatigue. The effect of fatigue was reflected in a time ! manipulation-type interaction (F(1,11) 5 6.01, po.05). To further analyze this interaction, post-hoc ANOVAs with data collapsed over driving condition and site were conducted for both manipulation conditions separately. These analyses showed that a reduction in P3a was highly significant after the 2-Back manipulation (F(1,11) 5 24.77, po.001), but there was no significant time effect in the 0-Back manipulation condition (F(1,11) 5 0.16, p 5 .70). Furthermore, in the omnibus analysis

P3a Difference waves The P3a (novel-standard) difference waves (depicted in Figure 2) show two important effects. First, there was a reduction in P3a amplitude in the post-manipulation measurement

A

P3a Difference Wave (novel-standard) 0-Back

Fz

Cz

Pz

10

0

0 100

300

0

500

100

300

500

100

300

500

100

300

500

−10

non-driving

pre-fatigue post-fatigue

driving

pre-fatigue post-fatigue

2-Back Fz

Cz

Pz

10

0

0 100

300

500

0 100

300

500

− 10

B

C

0-Back pre-fatigue

post-fatigue

2-Back pre-fatigue

post-fatigue

non-driving

non-driving

driving

driving

Figure 2. Difference waves calculated as novel ERP–standard ERP. (A) Difference waves at Fz, Cz, Pz for pre-fatigue (bold lines) and post-fatigue (thin lines) measurement, during non-driving (continuous lines) and driving (dashed lines). (B) Scalp distribution plots in the 0-Back session. (C) Scalp distribution plots in the 2-Back session.

Mental fatigue and novelty processing

1123

a significant time ! site interaction was found (F(1,11) 5 6.34, po.05). Post-hoc analyses, with data collapsed over manipulation-load and driving condition, revealed a significant amplitude decrease post-manipulation relative to pre-manipulation at Cz (F(1,11) 5 8.88, po.05) and Pz (F(1,11) 5 13.37, po.005), but not at Fz (F(1,11) 5 0.07, p 5 .80). Driving. The driving effect was reflected in a driving ! site interaction (F(2,10) 5 9.13, po.05). Follow-up ANOVAs on separate electrode sites (data collapsed over time and manipulation-type) showed that, compared to the non-driving condition, P3a amplitude in the driving condition was marginally significantly reduced at Fz (F(1,11) 5 4.27, p 5 .063), and significantly reduced at Cz and Pz (F(1,11) 5 35.68, po.001; (F(1,11) 5 93.18, po.001). The absence of 4-way and 3-way interactions, including driving and time or driving and manipulation-type, indicated that this P3a reduction during driving was present independent of fatigue (after both 0-Back and 2-Back manipulations). To assure that the above described effects on P3a difference wave were indeed due to differences in processing of the novel stimuli, statistical analyses were reconducted over the raw ERP waves for novel stimuli separately from standard stimuli. Novel stimuli A similar pattern of effects as found in the P3a difference waves can be seen in the novel ERPs (see Figure 3). Amplitudes were reduced during driving, and an additive reduction after the fatigue manipulation was visible after 2-Back but not after 0-Back performance. Statistical analysis on novel ERPs was conducted in the same way as the difference waves (manipulation-type ! time ! driving condition ! site repeated measures ANOVA). Similar to

the difference waves, novel wave analysis showed the presence of the time ! manipulation-type interaction (F(1,11) 5 6.16, po.05), and the driving ! site interaction (F(2,10) 5 11.86, po.005). However, the time ! site interaction that was found in the difference waves was not significant in the novel ERPs. This indicated that the effects of fatigue were present at all electrodes. Follow-up ANOVAs for the time effect (collapsed over driving and site), confirmed that a post-manipulation reduction of P3a amplitude was present in the 2-Back condition (F(1,11) 5 48.78, po.001), but not in the 0-Back condition (all ps 4.1). Followup analysis of the driving effect with separate repeated measures ANOVAs for driving and non-driving amplitudes (collapsed over time and manipulation) showed a marginally significant site effect in the non-driving (F(2,10) 5 3.52, p 5 .074; simple contrasts: FzoCz, F(2,10) 5 10.77, po.01, and Cz 5 Pz, Fo1). During the driving task, P3a amplitudes were attenuated such that no difference between sites was present (Fo1). These results confirm that the effects of driving and fatigue reflected in the P3a difference wave were present in a similar manner for the novel ERPs. The scalp distribution of the novel ERPs was more posterior, however, than the distribution of the difference waves. Standard Stimuli Standard stimuli evoked a different ERP pattern than novel stimuli (see Figure 4). In the P3a time window (300–350 ms post stimulus), no positive peak was present. Rather, a negativity occurred that peaked at Fz (site main effect F(2,10) 5 118.58, po.001; simple contrasts Fz4Cz4Pz; F(1,11) 5 9.00, po.05; F(1,11) 5 93.27, po.001). But effects of fatigue (time ! manipulation-type interaction) and driving (driving ! site inter-

Novel Stimuli

0-Back Fz

Cz

Pz

10

10

10

5

5

5

0

0 100

300

0

500

100

300

500

−5

−5

−5

− 10

− 10

− 10

non-driving

pre-fatigue post-fatigue

driving

pre-fatigue post-fatigue

2-Back Cz

Fz

10

10

5

5

5

0 100

300

500

300

500

100

300

500

Pz

10

0

100

0 100

300

500

−5

−5

−5

− 10

− 10

− 10

Figure 3. Novel stimulus locked ERPs in the 0-Back manipulation session (upper panel), and 2-Back session (lower panel) at Fz, Cz, and Pz. Waveforms are displayed separately for pre-fatigue (bold lines) and post-fatigue (thin lines) measurement, during non-driving (continuous lines) and driving (dashed lines).

1124

S. A. A. Massar et al. Standard Stimuli

0-Back Fz

Cz

Pz

10

10

10

5

5

5

0

0 100

300

0

500

100

−5

−5

−10

−10

300

500

300

500

100

300

500

−10 pre-fatigue

non-driving

post-fatigue pre-fatigue

driving

post-fatigue

2-Back Cz

Fz

Pz

10

10

10

5

5

5

0

0 100

100 −5

300

500

0 100

300

500

−5

−5

−5

−10

−10

−10

Figure 4. Standard stimulus locked ERPs in the 0-Back manipulation session (upper panel), and 2-Back session (lower panel) at Fz, Cz, and Pz. Waveforms are displayed separately for pre-fatigue (bold lines) and post-fatigue (thin lines) measurement, during non-driving (continuous lines) and driving (dashed lines).

action) were not found in the standard ERP. The frontal distribution and the polarity of this standard ERP explain the slight difference in distribution between the novel ERP and the novelstandard difference wave. Crucially, the effects of driving and fatigue that were found in P3a difference waves were not due to changes in the standard ERP. Discussion In the current study, it was investigated whether effects of mental fatigue on novelty processing and driving performance depended on working memory load of the fatigue-inducing task. The main hypothesis was that depletion of attentional control and working memory resources after performance of the high load 2-Back task would cause more pronounced fatigue effects. Novelty processing, measured as P3a amplitude, was indeed attenuated after 90 min of performance of the 2-Back task. This reduction was not significant after an equally long period of 0-Back performance. Importantly, this attenuation of P3a amplitude after 2-Back performance was specifically due to effects on ERPs following novel stimuli. ERPs locked to standard stimuli did not show these same effects. It can therefore be argued that novelty processing was diminished after a long period of continuous 2-Back execution. These findings have several implications. First, in line with our expectation, novelty processing can be compromised during a state of mental fatigue, specifically when caused by prolonged recruitment of attentional control and working memory. Depletion of these resources may obstruct processing of unexpected novel stimuli. In accordance with findings of attenuated P3a during dual task paradigms (Berti & Schro¨ger, 2003; Friedman et al., 1998; Zhang et al., 2006), this finding supports the notion that novelty processing, although it is

bottom-up triggered, is dependent on the availability of attentional control resources. Since mental fatigue is thought to often go along with depletion of attentional control (Lorist, Boksem, & Ridderinkhof, 2003; Van der Linden et al., 2003a), and since we specifically targeted this aspect (by comparing 0-Back and 2Back related fatigue effects), it seems reasonable to argue that in our study this indeed was the mechanism through which novelty processing was decreased. Different views on the functional significance of the P3a have been formulated. On the one hand, P3a can be interpreted as an index of bottom-up distraction. P3a amplitude and novelty-induced performance decrements on a primary task are found to be positively related (e.g., Berti & Schro¨ger, 2003; Escera, Yago, Corral, Corbera, & Nunez, 2003; SanMiguel et al., 2008). Some recent accounts, however, emphasize a possible role in higher cognitive functions, such as the overriding or inhibition of context (i.e., active task rules) (Luu & Tucker, 2002; Polich, 2007), possibly activating a control network that facilitates switching of action programs (Barcelo, Escera, Corral, & Perianez, 2006). The relation between depleted attentional control and attenuated P3a amplitude in the present study seems to be more in line with the latter interpretations. However, the main aim of this study was not to precisely unravel the functional significance of the P3a, but rather to examine whether effects of mental fatigue were dependent on attentional control load of the fatigue-inducing task. The present study clearly shows that the manifestation of mental fatigue effects are task related. Novelty processing decrements were significant only when fatigue was caused by the 2Back task, and not by the 0-Back task. As argued above, it can be thought that depletion of control capacity and working memory was more pronounced after the 2-Back, consequently compro-

Mental fatigue and novelty processing

1125

mising P3a generation. It must be noted that the subjective appraisal of the level of mental fatigue was similar after both tasks. Therefore, differential effects cannot easily be explained as reflecting factors relating to subjective fatigue. It therefore can be concluded that the nature of mental activity influences effects of the resulting fatigue on brain processes. An important direction for future research should be to systematically investigate the impact of different types of mental activity on cognitive, behavioral, and emotional performance. By doing so, it may increase understanding about the cognitive and biological mechanisms underlying different fatigue aspects, and may identify risk factors that are specific to certain types of activity (for instance, the manifestations of fatigue that are typical to transportation or air traffic control jobs). Considering the effects of fatigue on behavioral performance, it is remarkable that no decline in simulated driving was found. Even though subjects felt substantially more tired, both after 0Back and 2-Back performance, steering error (SDLP) did not increase. A reason for this could be that the driving task was relatively easy. The purpose of the task was to steer the car along a curved road, without the presence of other traffic or unexpected events. Generally, the effects of fatigue show mostly in complex or novel tasks (Broadbent, 1979; Hockey, 1993), which makes it plausible that fatigue effects are not (yet) showing in this task. In order to get a better understanding of the effects of fatigue (and in particular the effects of different forms of mental activity) on behavior and performance, future studies should include more complex tasks. A further limitation of this study is that no direct assessment of attentional control has been conducted. Although it is plausible that a depletion of attentional control and working memory was more pronounced after prolonged 2-Back performance than after 0-Back performance, no formal comparison of attentional control capacity was included in the current design. The outcome variable rather was an automatically triggered process that was

not related to performance on a task. Using such automatic measures in fatigue research has an important advantage over task-related performance or physiological measures. Since fatigue inherently goes along with decreased motivation (Meijman, 2000), performance-related measures are always vulnerable to loss of motivation or strategic changes in task execution. Using measures of automatic, non-effortful processing, therefore, can provide insight into the cognitive effects of fatigue that are separate from task disengagement (Van der Linden, Massar, Schellekens, Ellenbroek, & Verkes, 2006). On the other hand, to confirm that 2-Back performance indeed causes a depletion of attentional control resources, future studies should include direct measurements of cognitive control. Furthermore, no non-fatigued control condition was included in this study. Therefore, effects of habituation of P3a cannot be ruled out (Friedman & Simpson, 1994; Kazmerski & Friedman, 1995). However, it is not very likely that it was the primary cause of P3a reduction after the fatigue manipulation. The set of novel stimuli that was used post-manipulation was different from the stimuli in the pre-manipulation measurement. Therefore, any differences between pre-and post-manipulation P3a amplitude cannot be explained by habituation due to repetition of stimuli. Furthermore, there is no reason to suspect that habituation would be different after different fatigue manipulations. The finding that P3a attenuation is most prominent after 2-Back performance thus cannot be accounted for by habituation. In summary, fatigue effects on novelty processing are dependent on the cognitive demands of the fatigue-inducing mental activity. Strong depletion of attentional control and working memory capacity after 2-Back task performance was associated with a marked reduction of P3a. 0-Back task performance did not show such effects. Furthermore, depletion-related decrease of P3a was present both when attention was focused on a driving task and when no simultaneous task was performed, while driving performance was not affected by fatigue.

REFERENCES Baker, K., Olson, J., & Morriseau, D. (1994). Work practice fatigue, and nuclear power plant safety performance. Human Factors, 36, 244– 257. Barcelo, F., Escera, C., Corral, M. J., & Perianez, J. A. (2006). Task switching and novelty processing activate a common neural network for cognitive control. Journal of Cognitive Neuroscience, 18, 1734– 1748. Bartlett, F. C. (1943). Fatigue following highly skilled work. Proceedings of the Royal Society B, 131, 247–257. Berti, S., & Schro¨ger, E. (2003). Working memory controls involuntary attention switching: Evidence from an auditory distraction paradigm. European Journal of Neuroscience, 17, 1119–1122. Boksem, M. A. S., Lorist, M. M., & Meijman, T. F. (2005). Effects of mental fatigue on attenion: An ERP study. Cognitive Brain Research, 25, 106–117. Boksem, M. A. S., Meijman, T. F., & Lorist, M. M. (2006). Mental fatigue, motivation and action monitoring. Biological Psychology, 72, 123–132. Broadbent, D. E. (1979). Is a fatigue test possible now? Ergonomics, 22, 1277–1290. Brown, I. D. (1994). Driver fatigue. Human Factors, 36, 298–317. Combs, L. A., & Polich, J. (2006). P3a from auditory white noise stimuli. Clinical Neurophysiology, 117, 1106–1112. Debener, S., Kranczioch, C., Herrmann, C. S., & Engel, A. K. (2002). Auditory novelty oddball allows reliable distinction of top-down and bottom-up processes of attention. International Journal of Psychophysiology, 46, 77–84.

Duncan, J., Emslie, H., Williams, P., Johnson, R., & Freer, C. (1996). Intelligence and the frontal lobe: The organization of goal-directed behavior. Cognitive Psychology, 30, 257–303. Escera, C., Yago, E., Corral, M. J., Corbera, S., & Nunez, M. I. (2003). Attention capture by auditory significant stimuli: Semantic analysis follows attention switching. European Journal of Neuroscience, 18, 2408–2412. Fabiani, M., & Friedman, D. (1995). Changes in brain activity patterns in aging: The novelty oddball. Psychophysiology, 32(6), 579–594. Friedman, D., Cycowicz, Y. M., & Gaeta, H. (2001). The novelty p3: An event-related brain potential (ERP) sign of the brain’s evaluation of novelty. Neuroscience and Biobehavioral Reviews, 25, 355–373. Friedman, D., Kazmerski, V. A., & Cycowicz, Y. M. (1998). Effects of aging on the novelty p3 during attend and ignore oddball tasks. Psychophysiology, 35(5), 508–520. Friedman, D., & Simpson, G. V. (1994). ERP amplitude and scalp distribution to target and novel events: Effects of temporal order in young, middle-aged and older adults. Brain Research: Cognitive Brain Research, 2, 49–63. Fuster, J. M. (1989). The prefrontal cortex: Anatomy, physiology, and neuropsychology of the frontal lobe (2nd ed). New York: Raven Press. Gosselin, A., De Koninck, J., & Campbell, K. B. (2005). Total sleep deprivation and novelty processing: Implications for frontal lobe functioning. Clinical Neurophysiology, 116, 211–222. Gosselin, A., Mathieu, A., Mazza, S., Decary, A., Malo, J., & Montplaisir, J. (2006). Deficits in involuntary attention switching in obstructive sleep apnea syndrome. Neuroscience Letters, 408, 73–78.

1126 Gratton, G., Coles, M. G., & Donchin, E. (1983). A new method for offline removal of ocular artifact. Electroencephalography and Clinical Neurophysiology, 55(4), 468–484. Hockey, G. R. J. (1993). Cognitive-energetical control mechanisms in the management of work demands and psychological health. In A. D. Baddeley & L. Weiskrantz (Eds.), Attention, selection, awareness and control. A tribute to Donald Broadbent (pp. 328–345). Oxford: Oxford University Press. Holdstock, J., & Rugg, M. D. (1995). The effect of attention on the p300 deflection elicited by novel sounds. Journal of Psychophysiology, 9, 18–31. Kazmerski, V. A., & Friedman, D. (1995). Repetition of novel stimuli in an ERP oddball paradigm: Aging effects. Journal of Psychophysiology, 9, 298–311. Lorist, M. M., Boksem, M. A. S., & Ridderinkhof, K. R. (2003). Impaired cognitive control and reduced cingulate activity during mental fatigue. Cognitive Brain Research, 24, 199–205. Lorist, M. M., Klein, M., Nieuwenhuis, S., De Jong, R., Mulder, G., & Meijman, T. F. (2000). Mental fatigue and task control: Planning and preparation. Psychophysiology, 37, 1–12. Luu, P., & Tucker, D. M. (2002). Self-regulation and the executive functions: Electrophysiological clues. In A. Zani & A. M. Proverbio (Eds.), The cognitive electrophysiology of mind and brain (pp. 199– 219). New York: Academic Press. Meijman, T. F. (2000). The theory of the stop emotion: On the functionality of fatigue. In D. Pogorski & W. Karwowski (Eds.), Ergonomics and safety for global business quality and productivity: Proceedings of the 2nd International Conference, Ergon-Axia. Warsaw, Poland: Central Institute for Labour Protection. MettingVanRijn, A. C., Peper, A., & Grimbergen, C. A. (1990). High quality recording of bioelectric events. I: Interference reduction, theory and practice. Medical & Biological Engineering & Computing, 28, 389–397. Miyake, A., Friedman, N. P., Emerson, M. J., Witzki, A. H., Howerter, A., & Wager, T. D. (2000). The unity and diversity of executive functions and their contributions to complex ‘‘frontal lobe’’ tasks: A latent variable analysis. Cognitive Psychology, 41, 49–100. Muller-Gass, A., Macdonald, M., Schroger, E., Sculthorpe, L., & Campbell, K. (2007). Evidence for the auditory p3a reflecting an automatic process: Elicitation during highly-focused continuous visual attention. Brain Research, 1170, 71–78.

S. A. A. Massar et al. Muller-Gass, A., & Schro¨ger, E. (2007). Perceptual and cognitive task difficulty has differential effects on auditory distraction. Brain Research, 1136, 169–177. Petridou, E., & Moustaki, M. (2000). Human factors in the causation of road traffic crashes. European Journal of Epidemiology, 16, 819–826. Polich, J. (2007). Updating p300: An integrative theory of p3a and p3b. Clinical Neurophysiology, 118, 2128–2148. Salmi, J., Huotilainen, M., Pakarinen, S., Siren, T., Alho, K., & Aronen, E. T. (2005). Does sleep quality affect involuntary attention switching system? Neuroscience Letters, 390(3), 150–155. SanMiguel, I., Corral, M. J., & Escera, C. (2008). When loading working memory reduces distraction: Behavioral and electrophysiological evidence from an auditory-visual distraction paradigm. Journal of Cognitive Neuroscience, 20, 1131–1145. Swaen, G. M. H., Van Amelsfoort, L. G. P. M., Bu¨ltmann, U., & Kant, I. (2003). Fatigue as a risk factor for being injured in an occupational accident: Results from the Maastricht Cohort Study. Occupational Environmental Medicine, 60(Suppl.1), i88–i92. Van der Linden, D., & Eling, P. (2006). Mental fatigue disturbs local processing more than global processing. Psychological Research, 70, 395–402. Van der Linden, D., Frese, M., & Meijman, T. F. (2003a). Mental fatigue and the control of cognitive processes: Effects on perseveration and planning. Acta Psychologica, 113, 45–65. Van der Linden, D., Frese, M., & Sonnentag, S. (2003b). The impact of mental fatigue on exploration in a complex computer task: Rigidity and loss of systematic strategies. Human Factors, 45(3), 483–494. Van der Linden, D., Massar, S. A., Schellekens, A. F., Ellenbroek, B. A., & Verkes, R. J. (2006). Disrupted sensorimotor gating due to mental fatigue: Preliminary evidence. International Journal of Psychophysiology, 62, 168–174. Zhang, P., Chen, X., Yuan, P., Zhang, D., & He, S. (2006). The effect of visuospatial attentional load on the processing of irrelevant acoustic distractors. NeuroImage, 33, 715–724. Zijlstra, F. R. H. (1993). Efficiency in work behavior: A design approach for modern tools. Delft: Delft University Press.

(Received July 15, 2009; Accepted December 18, 2009)

Psychophysiology, 47 (2010), 1127–1133. Wiley Periodicals, Inc. Printed in the USA. Journal compilation r 2010 Society for Psychophysiological Research No claim to original US government works DOI: 10.1111/j.1469-8986.2010.01030.x

Adaptation effects to sleep studies in participants with and without chronic posttraumatic stress disorder

ELLEN HERBST,a,b THOMAS J. METZLER,a,b MARYANN LENOCI,b SHANNON E. MCCASLIN,a,b SABRA INSLICHT,a,b CHARLES R. MARMAR,a,b and THOMAS C. NEYLANa,b a

Department of Psychiatry, University of California, San Francisco, San Francisco, California Psychiatry Service, San Francisco Veterans Affairs Medical Center, San Francisco, California

b

Abstract The ‘‘first night effect’’ (FNE) is the alteration of sleep architecture observed on the first night of polysomnographic (PSG) studies. It is unclear whether the FNE reflects adaptation to the equipment, sleeping environment, or both. Moreover, it is possible that certain patient populations, such as those with posttraumatic stress disorder (PTSD), demonstrate greater adaptation effects that are highly context dependent. We assessed FNE in participants with PTSD and healthy controls in a cross-sectional study consisting of PSG testing at home and in the hospital. Contrary to our expectations, the PTSD group showed no adaptation effects in either setting. Only the control group assigned to the ‘‘hospital first’’ condition showed significant decreases in total sleep time on night 1 versus night 2 of the study. The results suggest that the FNE is related to adaptation to the combination of the hospital environment and the recording equipment. Descriptors: Posttraumatic stress disorder, Adaptation effects, First night effect, Ambulatory polysomnography

eralized anxiety disorder (Saletu, Klosch, Gruber, Anderer, Udomratn, & Frey, 1996) conclude that adaptation effects occur in certain subgroups regardless of setting. Others have postulated that adaptation to PSG recording equipment plays a significant role in FNE. Lorenzo and Barbanoj (2002) studied the FNE in healthy volunteers during three nonconsecutive sets of laboratory recordings one month apart. They found FNE only in the ‘‘very first night’’ of the first series of recordings. These results suggest that familiarity with PSG equipment may eliminate FNE in subsequent PSG studies. Individuals with posttraumatic stress disorder (PTSD) are an important test population for PSG studies that examine FNE. Most patients with PTSD report nightmares and insomnia, which are listed separately in the re-experiencing and hyperarousal clusters in the DSM-IV criteria for the disorder (First, Spitzer, Williams, & Gibbon, 1996). Subjective sleep disturbances are frequent among patients with PTSD, both in treatment-seeking (Roszell, McFall, & Malas, 1991) and epidemiologic samples (Neylan, Marmar, Metzler, Weiss, Zatzick, et al., 1998), while laboratory-based PSG studies have produced mixed results. A recent meta-analysis of 20 studies found that patients with PTSD had more stage 1 sleep, less slow wave sleep, and greater REM density (REM activity/minutes REM sleep) compared to people without PTSD (Kobayashi, Boarts, & Delahanty, 2007). Given the high frequency of reported sleep disturbances and the hypothesized state of nighttime hypervigilance in subjects with PTSD, it has been proposed that FNE would be prominent in these subjects, particularly in an unfamiliar sleep environment.

The ‘‘first night effect’’ (FNE) is a well-known phenomenon in polysomnographic (PSG) recordings characterized by decreased total sleep time, lower sleep efficiencies, reduction in rapid-eyemovement (REM) sleep, and longer REM latencies on the first night of testing (Agnew, Webb, & Williams, 1966). First night data are often excluded in analyses of polysomnographic (PSG) recordings because they are considered to reflect a period of adaptation that is unrepresentative of usual sleep patterns. Although the FNE has been widely studied in healthy subjects and clinical populations, few studies have systematically examined the causes of FNE. Some ambulatory PSG studies suggest that providing a comfortable sleeping environment or conducting home recording eliminates or reduces FNE (Coates, George, Killen, Marchini, Hamilton, & Thorensen, 1981; Edinger, Fins, Sullivan, Marsh, Dailey, et al., 1997; Sharpley, Solomon, & Cowen, 1988). Other home PSG studies of healthy participants (Le Bon, Staner, Hoffmann, Dramaix, San Sebastian, et al., 2001), elderly individuals (Wauquier, van Sweden, Kerkhof, & Kamphuisen, 1991; Edinger et al., 1991) and patients with genThis work was supported by the National Institutes of Health (TCN: MH057157 & MH73978), the Sierra Pacific Mental Illness and Education Clinical Center (MIRECC), and from the NIH/NCRR UCSF-CTSI Grant Number UL1 RR024131. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIH. Address correspondence to: Dr. T. Neylan, PTSD Program, Psychiatry Service 116P, VA Medical Center, 4150 Clement Street, San Francisco, CA 94121. E-mail: [email protected] 1127

1128 Two laboratory-based PSG studies comparing first night adaptation effects in PTSD subjects and controls have reported mixed findings (Ross, Ball, Sanford, Morrison, Dinges, et al., 1999; Woodward, Bliwise, Friedman, & Gusman, 1996). Ross and colleagues found no differences in adaptation effects in a mixed sample of outpatient and residential treatment PTSD subjects compared to outpatient controls in a laboratory study (Ross et al., 1999). However, increased REM activity and density was observed in PTSD subjects on the first versus the second night. In contrast, Woodward et al. (1996) found that FNEs in PTSD subjects were dependent on whether the subjects were currently in a residential treatment program versus outpatient treatment. In this laboratory-based study, PTSD inpatients showed decreased FNEs compared to outpatient controls, whereas PTSD outpatients had enhanced FNE compared to outpatient controls (Woodward et al., 1996). These results suggest that adaptation effects observed in PTSD may reflect enhanced sensitivity to a novel sleeping environment. As most previous PSG studies in PTSD have been conducted only in the sleep laboratory, it is difficult to discern whether FNEs observed in PTSD represent adaptation to recording equipment, novel sleeping environment, or both. A direct comparison of PSG testing in the two settings would clarify whether the recording context affects the results, allowing for more accurate study and enhanced understanding of PTSD-related sleep disruption. To date, this is the first study to examine FNE in medically healthy medication-free subjects with PTSD and age-matched controls with two pairs of PSG studies conducted in both home and hospital settings. We hypothesized that both the PTSD group and the control group would have greater FNE in the hospital than at home, and that the PTSD group would have greater FNE compared to controls in night 1 versus night 2 of the study in both settings. Finally, we hypothesized that adaptation effects in both groups would be attenuated in the second pair of PSG studies.

Methods Participants and Procedures Medically healthy male and female subjects were recruited from internet and newspaper advertisements and from the San Francisco Veterans Affairs Medical Center (SFVAMC) PTSD Outpatient Program. Subjects were recruited as part of a larger study examining the relationship between sleep architecture and hypothalamic-pituitary-adrenal axis activity in chronic PTSD. All subjects were given details of the study and asked to sign a written informed consent form if they wished to participate. The study protocol and consent form were approved by the Committee on Human Research at the University of California, San Francisco (UCSF). Overall, 151 subjects were initially consented, 89 enrolled, 19 refused (due to scheduling conflicts), and 43 did not meet study criteria. The final data analysis included 60 subjects; 29 subjects had to be excluded due to equipment failure. Subjects were paid up to $400 for completion of the study, which included 5 nights of PSG testing. The PTSD group with complete night 1 and 2 data in both settings used for these analyses consisted of 34 male and female subjects with current chronic PTSD as assessed by the Clinician Administered PTSD Scale (CAPS) (Blake, Weathers, Nagy,

E. Herbst et al. Kaloupek, Gusman et al., 1995). The CAPS measures frequency and intensity of PTSD-related symptoms, with possible scores ranging from 0 to 136. In a recent review of studies utilizing the CAPS, Weathers, Keane, and Davidson (2001) proposed the following severity score ranges: 0–19 5 Asymptomatic/few symptoms; 20–39 5 Mild PTSD/subthreshold; 40–59 5 Moderate PTSD/threshold; 60–79 5 Severe PTSD symptomatology; 480 5 Extreme PTSD. The PTSD group was mixed with respect to trauma type and severity, and civilian and military trauma; however, all PTSD subjects had symptoms for at least 1 year in duration and none had been exposed to traumatic events within the past year. The control group was composed of 26 healthy participants, negative for lifetime PTSD or major depression. Subjects were age balanced within gender. In both groups, subjects were excluded if they met criteria for alcohol or substance abuse within the past 2 years, lifetime criteria for psychotic disorder, bipolar disorder, or obsessive compulsive disorder as assessed by the Structured Clinical Interview for DSM-IV (SCIDP) (Spitzer, Williams, Gibbon, & First, 1996). Medical exclusion criteria included any history of neurologic disease (traumatic brain injury, seizure, hemorrhage, stroke, or other brain injury), current systemic illness affecting central nervous system function, or use of any medication affecting the brain. All subjects were required to be free of any psychiatric medications for at least 2 months prior to participation. Subjects were asked to abstain from any alcohol the week before the study. An oximeter (Respironics Cricket, Monroeville, PA) was used to screen for obstructive sleep apnea (OSA). The cutoff criterion for apnea was 10 desaturation events per hour in bed, which has been shown to have a sensitivity of 98% and specificity of 48% in detecting OSA (Se´rie`s, Marc, Cormier, & La Forge, 1993). Subjects who screened positive for OSA were excluded. All subjects were alcohol-free and allowed up to one caffeinated beverage each morning during the sleep recordings. Subjects completed five nights of ambulatory polysomnography, two in participants’ homes and three in a hospital research unit, separated by three nights of sleep in the participant’s home without wearing the sleep equipment. The study was counterbalanced such that subjects were randomly assigned to begin the study either at home or in the hospital. Subjects received metyrapone prior to the third night of sleep in the hospital research unit in an investigation of hypothalamic-pituitary adrenal axis activity and sleep. (Metyrapone was not administered during the nights of the study reported here.) The results of the metyrapone study compared nights 2 and 3 in the hospital only and have been previously published (Neylan, Lenoci, Maglione, Rosenlicht, Metzler, et al., 2003; Otte, Lenoci, Metzler, Yehuda, Marmar, & Neylan, 2005, 2007). The current study examines the two nights of home sleep recordings, never previously published, the hospital adaptation night, never previously published, versus the second night of sleep recordings in the hospital.

Measures At baseline, subjects completed a set of subjective measures, including Symptom Checklist-90-Revised (SCL-90-R) (Derogatis, 1994), a standard self-report measure of general psychopathology, and the Pittsburgh Sleep Quality Index (PSQI) (Buysse, Reynolds, Monk, Berman, & Kupfer, 1989), a self-report measure that provides a subjective assessment of sleep quality, sleep latency, sleep duration, sleep maintenance, sleep disturbances (including nightmares), use of sedative-hypnotics, and daytime energy.

Adaptation effects to sleep studies and PTSD Subjects adhered to a stable sleep-wake schedule at their habitual times during the entire study period. Self-report ratings of subjective sleep quality were obtained for each night of sleep using a 100 mm visual analog scale, which ranged from ‘‘very bad’’ to ‘‘very good’’ sleep quality. Objective sleep quality was measured with ambulatory polysomnography, using an Oxford MR95 digital recorder (Oxford Instruments, Oxford, UK). The parameters recorded included an electroencephalogram (EEG) at leads C3 and C4, left and right electrooculograms (EOG), submental electromyogram (EMG), and electrocardiogram (EKG) in accordance with standardized guidelines (Rechtschaffen & Kales, 1968). The EEG and EOG leads were referenced to linked mastoids. All sleep was imported into Pass Plus (Delta Software, St. Louis, MO) analytic software and visually scored in 30-sec epochs in accordance to Rechtschaffen and Kales (1968). Sleep architecture was delineated as the percentage of time spent asleep in non-REM (NREM) stages one through four and stage REM. Sleep continuity was measured by calculating sleep maintenance, defined as the ratio of total time spent asleep divided by the total recording period between sleep onset and offset. An awakening was defined by EEG arousals lasting 30 s or longer. REM periods were defined by at least 3 min of consecutive REM sleep with no less than 30 min of NREM sleep separating two REM periods. PSG measures included total sleep time in minutes, sleep maintenance (percent of actual sleep time between sleep onset and the final awakening), and total time awake after sleep onset in minutes (WASO). REM measures included REM percent (minutes of REM sleep/minutes of total sleep time), REM latency (minutes from sleep onset to first REM period), REM activity (number of rapid eye movements), and REM density (REM activity/minutes REM sleep). Statistical Analyses Differences in demographic characteristics between PTSD subjects and controls were compared using t-tests for continuous variables and chi-square tests for dichotomous variables. Measures of sleep architecture and subjective sleep ratings on the four nights of PSG testing (hospital nights 1 and 2 and home nights 1 and 2) were analyzed by a linear mixed model fitting PTSD status as a between-groups fixed effect and time as a within-subjects repeated effect. We also included location (home versus hospital) and order (first versus second series of recordings) in the model. Group differences were then examined using t-tests for continuous variables. The relationships between PSG variables and subjective sleep ratings were analyzed by two-tailed Pearson correlations. All values are expressed as mean and standard deviation. A nominal level of significance a 5 .05 was accepted. Results Subject Characteristics The mean age for the PTSD subjects was 42.2 (SD 5 10.5) and control subjects was 39.8 (SD 5 11.3). The mean CAPS score for the PTSD subjects was 62.7 (SD 5 18.2), reflecting moderate PTSD symptom severity, compared to .8 (SD 5 1.5) in controls. Demographic characteristics of the PTSD group and the control group are summarized in Table 1. PSG data from nights 1 and 2 in each location are presented in Table 2. Main effects and significant interactions of group, night, location, and order on sleep architecture and subjective sleep ratings are summarized in Tables 3a and 3b. As noted above, the nights 2 and 3

1129 Table 1. Characteristics of Healthy Medication-Free PTSD Subjects and Age-Matched Controls

Measures Age (years) Gender Male Female CAPS PSQI SCL-90 Depression Yes No Past SUD Yes No

PTSD group (Mean " SD)

Control group (Mean " SD)

Contrasts

42.2 " 10.5

39.8 " 11.3

t(60) 5 .84, p 5 .40

61.8% (n 5 21) 50.0% (n 5 13) w2(1) 5 .83, p 5 .36 38.2% (n 5 13) 50.0% (n 5 13) 62.7 " 18.2 .8 " 1.5 t(60) 5 17.3, po.0001 11.2 " 4.4 4.1 " 2.0 t(60) 5 7.60, po.0001 1.26 " .69 .26 " .25 t(60) 5 7.00, po.0001 70.59% (n 5 24) 0 29.41% (n 5 10) 100% (n 5 26)

w2(1) 5 9.18, p 5 .002

64.71% (n 5 22) 88.46% (n 5 23) w2(1) 5 17.14, po.0001 35.29% (n 5 12) 11.54% (n 5 3)

Note: CAPS 5 Clinician-Administered PTSD Scale; PSQI 5 Pittsburgh Sleep Quality Index (scores 45 associated with significant insomnia); SCL-90 5 Symptom Checklist-90; Depression 5 Current major depressive disorder; Past SUD 5 Past substance use disorder.

PSG data acquired in the hospital has been reported previously in the male (Neylan et al., 2003) and female (Otte et al., 2007) subsamples. Polysomnographic Data First night effects. In the overall sample, total sleep time was reduced on night 1 compared with night 2 as evidenced by a main effect of night (F[1,55.66] 5 5.00, po.029). As expected, a night ! location effect was seen with respect to total sleep time (F[1,55.54,] 5 13.92, po.001), reflecting larger first night effects in the hospital versus the home. Night ! order effects were observed on total sleep time (F[1,55.54] 5 5.81, po.019), demonstrating that the first night effect was more pronounced in the initial nights of the study. Note that the ‘‘order’’ variable refers to the first versus second series of recordings and does not reflect first night changes (night 1 versus night 2 within each series). Group differences. We observed a group ! location ! order effect on total sleep time (F[1,55.82] 5 7.10, po.010), indicating that differences in total sleep time between PTSD and control groups were modified both by location and order. The PTSD group demonstrated no first night changes in sleep duration or architecture in either location. The control group assigned to the ‘‘hospital first’’ condition showed adaptation changes in total sleep time its first night in the hospital (F[1,30.00] 5 5.77, po.023), but not in the subsequent nights of the study at home. The control group assigned to the ‘‘home first’’ condition did not have any FNE, at home or later in the hospital (see Figure 1). Three-way group ! location ! order interactions were seen with respect to sleep maintenance (F[1,55.88] 5 5.16, po.027] and REM density (F[1, 56.11] 5 4.83, po.032). Several PSG findings emerged in the PTSD group, whose first recordings took place at home. In that setting, the PTSD group assigned to the ‘‘home first’’ condition had reduced sleep maintenance relative to the ‘‘home first’’ control group (F[1,28.93] 5 5.17, po.031). The ‘‘home first’’ PTSD group also demonstrated higher REM densities on its home nights than the ‘‘hospital first’’ PTSD group (F[1,32.21] 5 7.60, po.010). Location effects. In the overall sample, participants slept longer at home than in the hospital as evidenced by a main effect

1130

E. Herbst et al.

Table 2. Night 1 versus Night 2 Sleep in Home and Hospital Recordings PTSD Group (n 5 34)

Measures Total sleep time (min) Sleep maintenance Subjective rating WASO (min) REM density REM percent REM latency

Home PSG

Hospital PSG

Mean (SD)

Mean (SD)

Night 1

Night 2

Night 1

Night 2

387.84 (73.93) .84 (.13) 55.42 (26.76) 77.54 (69.56) 4.08 (2.27) 23.14 (6.90) 95.17 (64.13)

370.94 (91.57) .85 (.11) 59.61 (27.38) 75.31 (72.75) 3.85 (1.99) 20.39 (7.55) 83.42 (66.63)

332.21 (87.01) .77 (.18) 48.21 (27.03) 98.38 (81.15) 4.04 (1.74) 19.99 (6.17) 93.34 (45.30)

367.87 (61.02) .83 (.12) 56.09 (26.94) 79.10 (57.95) 4.42 (2.03) 21.79 (6.63) 83.37 (58.44)

Control Group (n 5 26)

Measures Total sleep time (min) Sleep maintenance Subjective rating WASO (min) REM density REM percent REM latency

Home PSG

Hospital PSG

Mean (SD)

Mean (SD)

Night 1

Night 2

Night 1

Night 2

386.56 (83.98) .87 (.10) 73.12 (24.80) 62.00 (80.75) 4.55 (1.96) 21.56 (5.41) 95.04 (61.53)

395.27 (68.49) .88 (.10) 83.96 (23.09) 51.50 (35.66) 4.30 (1.98) 22.44 (6.68) 99.32 (47.63)

352.48 (109.86) .81 (.13) 54.60 (28.42) 80.29 (56.83) 4.17 (1.99) 20.17 (6.87) 89.12 (54.12)

404.04 (83.06) .84 (.11) 72.64 (27.58) 80.12 (53.91) 3.88 (2.08) 2.25 (6.24) 97.15 (57.35)

Note: WASO 5 Wake After Sleep Onset; REM percent 5 REM time (minutes)/Total sleep time (minutes).

of location on total sleep time (F[1,56.06] 5 5.18, po.027) and sleep maintenance (F[1,56.23] 5 13.80, po.001). Three-way group ! location ! order interactions were seen with respect to total sleep time (F[1,55.82] 5 7.10, po.010), sleep maintenance (F[1,55.88] 5 5.16, po.027) and REM density (F[1,56.11] 5 4.83, po.032). Subjective sleep ratings. Mixed model analyses of subjective sleep ratings showed significant main effects for group (F[1,55.13] 5 11.85, po.001), night (F[1,53.72] 5 11.97, po.001), location (F[1,55.11] 5 14.30, po.000) and order (F[1,55.11] 5 7.01, po.011). Overall, mean subjective sleep ratings in the PTSD group were lower than those seen in the control group. Two-way interactions were seen between night and order (F[1,55.10] 5 4.29, p 5 .043) and between group and order (F[1,55.11] 5 9.00, po.004). Effect of comorbid depression and substance use disorders. About 29% of the PTSD group (n 5 10) had current major depressive disorder, and 65% had a history of substance use disorder (SUD) (n 5 24). This contrasts with the control group, in which no subjects met criteria for depression, and about 12% (n 5 3) had a history of SUD. We calculated the effect sizes of depression and SUD on the first night effect (total sleep time on night 1 versus night 2) in the PTSD and control groups. The effect sizes (Cohen’s d) for the first night effect on total sleep time for the PTSD/depression (1), PTSD/depression ( " ), and control groups were .48, .47, and .49, respectively. Thus, comorbid depression did not affect total sleep time in our sample. A history of SUD also did not appear to affect total sleep time in our study. The observed effect size in the PTSD/SUD (1) group on total sleep time was .47, compared with .50 in the PTSD/SUD ( " ) group, and .48 in the control/SUD ( " ) group. In keeping with

our results, the ‘‘hospital first’’ control group demonstrated the largest effect size (d 5 .87). Discussion This is the first study to examine the presence of FNE in PTSD outpatients and controls studied in both the hospital and home settings. We expected that the PTSD group would have an exaggerated adaptation response to the novel stressor of hospital PSG testing. Contrary to our prediction, we found that the PTSD group showed no first night changes in sleep architecture, at home or in the hospital. Only the control group assigned to the ‘‘hospital first’’ condition showed first night changes in total sleep time in the hospital. The lack of a significant FNE in our outpatient sample of PTSD subjects (e.g., no group ! night effect) differs from the results of other studies examining FNE in PTSD. Ross and colleagues (1999) observed increases in REM density in the first REM period and mean REM activity in a mixed sample of inpatients and outpatients with PTSD compared to control subjects. Woodward and colleagues (1996) noted more pronounced FNE among PTSD outpatients, and attenuated FNE among PTSD inpatients, compared with outpatient controls. Some investigators have postulated that individuals with PTSD perceive the hospital as ‘‘safe’’ because of the presence of a sleep laboratory technician (Sheikh, Woodward, & Leskin, 2003). In our study, the PTSD group’s lack of adaptation response in the hospital, compared with the marked FNE seen in the control group, may reflect the PTSD group’s subjective experience of safety in a monitored environment. It has been hypothesized that individuals with PTSD have ‘‘sleep state misperception,’’ often reporting worse sleep than is objectively seen on PSG testing (Klein, Koren, Arnon, & Lavie,

Adaptation effects to sleep studies and PTSD

1131

Table 3a. Main Effects of Group, Night, Location and Order on Sleep Parameters Measures Total sleep time (min) Sleep maintenance Subjective rating WASO (min) REM density REM percent REM latency

Group (F, p)

Night (F, p)

Location (F, p)

Order (F, p)

1.40, .242 .98, .327 11.85, .001n .83, .367 1.71, .196 .01, .915 .121, .729

5.00, .029n 3.10, .084 11.97, .001n 1.20, .277 .05, .828 .50, .484 .272, .604

5.18, .027n 13.80, .001n 14.30, .001n 6.65, .013n .04, .852 1.56, .217 .006, .939

.06, .812 .16, .691 7.01, .011n .11, .741 1.15, .287 1.39, .244 3.20, .081

Note: WASO 5 Wake After Sleep Onset; REM percent 5 REM time (minutes)/Total sleep time (minutes). n Indicates po.05.

Table 3b. Significant Interactions Between Group, Night, Location, and Order on Sleep Parameters

Measures Total sleep time (min) Sleep maintenance WASO (min) REM density Subjective rating

Night ! Loc (F, p)

Night ! Ord (F, p)

13.92, .000

5.81, .019

Loc ! Ord (F, p)

Grp ! Ord (F, p)

7.10, .010 5.16, .027 4.51, .038 4.83, .032

4.00, .050 4.29, .043

Grp ! Loc ! Ord (F, p)

9.00, .004

Note: Loc 5 Location; Ord 5 Order; Grp 5 Group.

2003; Lavie, 2001). One potential explanation for this discrepancy is that those with PTSD maintain a state of intact vigilance during lighter stages of sleep, which may not be detected on PSG. If PTSD subjects sustain a state of hypervigilance during sleep, they may not demonstrate the adaptation effects to PSG testing seen in controls. PTSD has also been described as a disorder of faulty adaptation of neurobiological systems after exposure to traumatic stressors (Germain, Buysse, & Nofzinger, 2008). If the FNE represents a normal adaptation process to the conditions of PSG testing (Schmidt & Kaelbling, 1971), the lack of FNE in the PTSD may be understood as an inability to adapt to the ‘‘stressor’’ of PSG testing. The inability to adapt to the novel stressor of PSG testing has also been proposed in major depression, which has been associated with an attenuated FNE (Mendels & Hawkins, 1967; Toussaint, Luthringer, Schaltenbrand, Carelli, Lainey, et al., 1995). In keeping with our hypothesis, neither group showed adaptation effects at home. Previous ambulatory PSG studies have demonstrated attenuated or absent FNE (Coates et al., 1981;

Figure 1. Total sleep time in PTSD and control groups in hospital and home settings in the first series of recordings.

Edinger, Fins, Sullivan, Marsh, Dailey, et al., 1997; Sharpley et al., 1988). Our findings indicate that the practice of excluding first night data from analyses of home PSG studies may not be necessary in groups similar to those studied here. The study’s counterbalanced design allowed us to evaluate whether first night effects are due to environment, recording equipment, or both. Although other studies investigating the FNE have directly compared PSG data from home and hospital recordings, ours is the first study to test for differences in sleep duration and architecture between the first and second series of recordings. In the overall sample, a night ! order effect was observed, indicating that previous experience with PSG mitigates adaptation effects. In both groups, subjects demonstrated no readaptation FNE in second series of PSG testing in either setting, confirming our hypothesis that the FNE would be reduced in the second pair of PSG recordings. These results are consistent with those of Lorenzo and Barbanoj (2002), who observed in an extended series of laboratory PSG sessions of healthy subjects that adaptation changes only occurred on the ‘‘very first night’’ of testing. In longitudinal sleep studies of groups similar to those studied here, only one adaptation night may be required, as participants habituate to laboratory PSG testing on the initial night of testing. We observed that ‘‘home first’’ PTSD group had higher REM densities on its home nights than the ‘‘hospital first’’ PTSD group (F[1,32.21] 5 7.60, po.010). It has been hypothesized that disruption of REM sleep mechanisms may underlie the insomnia and nightmares that characterize PTSD (Ross et al., 1999). Elevated REM density in the home PSG recording may reflect conditioned arousal in PTSD subjects when monitored during their habitual sleep environment. However, the finding of elevated REM densities in the ‘‘home first’’ PTSD group must be interpreted with caution, as this group demonstrated elevated REM densities in its second series of hospital-based recordings. It is possible that the group’s high REM densities were due to sampling error rather than the setting.

1132

E. Herbst et al.

Interestingly, participants with PTSD reported lower ratings of sleep quality than controls, even though there were no overall group differences in any measures of sleep architecture across the four nights. This discrepancy between objective and subjective measures of sleep disturbance is consistent with the findings of other investigators (Klein et al., 2003; Lavie, 2001) and supports the hypothesis that sleep misperception rather than objective sleep alteration may underlie sleep disturbances in PTSD. Individuals with PTSD have been observed to have a range of negative cognitive biases in many symptom domains (Hembree & Foa, 2000), and thus may report more subjective distress with respect to sleep as well. Another potential explanation is that traditional PSG scoring only provides an overview of sleep architecture and that more subtle abnormalities, like hypervigilance, may be undetected with these measures. It is possible that, as a consequence of traumatic exposure, PTSD subjects maintain alertness in light stages of sleep and perceive worse sleep quality than is demonstrated objectively. A third possibility is that our control subjects, who were specifically recruited into a sleep study, represent a biased sample. It is notable that the overall sleep maintenance of the control group was below than expected for a sample free from sleep disorders. Our results must be interpreted in light of several limitations. As described above, it is possible that controls recruited into our multi-night sleep study may have mild insomnia, although there were no clinical or psychometric manifestations of sleep disorder. The control group exhibited surprisingly poor sleep maintenance and long wake times. Another potential explanation for the lack of group differences is that our medically healthy, medicationfree outpatient PTSD group may reflect a biased sample not reflective of the broad population of PTSD patients. It has been noted previously that PTSD patients who volunteer for sleep studies may have less sleep disruption than those who decline (Woodward, Stegman, Pavao, Arsenault, Hartl, et al., 2007).

The PTSD group’s mean CAPS score indicates moderate symptom severity, but it is possible that ours is a less acutely disturbed sample than those typically seen in clinical settings. We considered the possibility that comorbid depression and/ or SUD may have affected our findings. As noted in the Results section, current major depression or history of SUD did not affect total sleep time in either the PTSD group or controls. Psychiatric outpatients (Mendels & Hawkins, 1967) and inpatients (Toussaint et al., 1995) with depression have been shown to have attenuated first night effects relative to healthy controls. No studies have investigated the presence of adaptation effects in individuals with substance use disorders. Further research with larger sample sizes would further elucidate the relationship between depression, SUD, and adaptation effects. Finally, PSG studies have demonstrated that adaptation varies significantly with age (Edinger et al., 1991; Wauquier et al., 1991), gender (Goel, Kim, & Lao, 2005), psychiatric comorbidities (Saletu et al., 1996), and medical conditions (Le Bon, Minner, Van Moorsel, Hoffmann, Gallego, et al., 2003). Future investigations with different study populations may allow us to clarify the relationship between demographic factors, comorbid diagnoses, and adaptation effects. In summary, our finding that outpatients with chronic PTSD demonstrated no FNE supports the retention of first night data in sleep studies of individuals with PTSD. The absence of FNE in either group at home highlights the utility and cost effectiveness of ambulatory sleep studies. The absence of a re-adaptation phenomenon in the second series of PSG recordings has important implications for longitudinal sleep studies, as it suggests that only the ‘‘very first’’ night of laboratory PSG data may need to be discarded. Finally, home-based PSG studies and treatment trials with interventions targeted at negative cognitive appraisals of sleep may be useful in defining and addressing the sleep disruption seen in PTSD.

REFERENCES Agnew, H. W., Webb, W. B., & Williams, R. L. (1966). The first night effect, an EEG study of sleep. Psychophysiology, 2, 263–266. Blake, D. D., Weathers, F. W., Nagy, L. M., Kaloupek, D. G., Gusman, F. D., Charney, D. S., & Keane, T. M. (1995). The development of a clinician-administered PTSD scale. Journal of Traumatic Stress, 8, 75–90. Buysse, D. J., Reynolds, C. F., Monk, T. H., Berman, S. R., & Kupfer, D. J. (1989). The Pittsburgh Sleep Quality Index, a new instrument for psychiatric practice and research. Psychiatry Research, 28, 193–213. Coates, T. J., George, J. M., Killen, J. D., Marchini, E., Hamilton, S., & Thorensen, C. E. (1981). First night effects in good sleepers and sleepmaintenance insomniacs when recorded at home. Sleep, 4, 293–298. Derogatis, L. (1994). Symptom checklist-90-revised (SCL-90 R) administration, scoring, and procedures manual (3rd ed). Minneapolis: National Computer Systems, Inc. Edinger, J. D., Marsh, G. R., McCall, W. V., Erwin, C. W., & Lininger, A. W. (1991). Sleep variability across consecutive nights of home monitoring in older mixed DIMS patients. Sleep, 14, 13–17. Edinger, J., Fins, A. I., Sullivan, R. J., Marsh, G. R., Dailey, D. S., Hope, T. V., et al. (1997). Sleep in the laboratory and sleep at home, comparisons of older insomniacs and normal sleepers. Sleep, 20, 1119–1126. First, M. B., Spitzer, R. L., Williams, J. B. W., & Gibbon, M. (1996). Structured clinical interview for DSM-IV (SCID-I). New York: New York State Psychiatric Institute, Biometrics Research. Germain, A., Buysse, D. J., & Nofzinger, E. (2008). Sleep-specific mechanisms underlying posttraumatic stress disorder: Integrative review and neurobiological hypotheses. Sleep Medicine Reviews, 12, 185–195.

Goel, N., Kim, H., & Lao, R. P. (2005). Gender differences in polysomnographic sleep in young healthy sleepers. Chronobiology International, 22, 905–915. Hembree, E. A., & Foa, E. B. (2000). Posttraumatic stress disorder, psychological factors and psychosocial interventions. Journal of Clinical Psychiatry, 61, 33–39. Klein, E., Koren, D., Arnon, I., & Lavie, P. (2003). Sleep complaints are not corroborated by objective sleep measures in post-traumatic stress disorder: A 1-year prospective study in survivors of motor vehicle crashes. Journal of Sleep Research, 12, 35–41. Kobayashi, I., Boarts, J., & Delahanty, D. (2007). Polysomnographically measured sleep abnormalities in PTSD: A meta-analytic review. Psychophysiology, 44, 660–669. Lavie, P. (2001). Sleep disturbances in the wake of traumatic events. New England Journal of Medicine, 345, 1825–1832. Le Bon, O., Staner, L., Hoffmann, G., Dramaix, M., San Sebastian, I., Murphy, J. R., et al. (2001). The first-night effect may last more than one night. Journal of Psychiatric Research, 35, 165–172. Le Bon, O., Minner, P., Van Moorsel, C., Hoffmann, G., Gallego, S., Lambrecht, L., et al. (2003). First-night effect in the chronic fatigue syndrome. Journal of Psychiatric Research, 120, 191–199. Lorenzo, J. L., & Barbanoj, M. J. (2002). Variability of sleep parameters across multiple laboratory sessions in healthy young subjects: The ‘‘very first night effect’’. Psychophysiology, 39, 409–413. Mendels, J., & Hawkins, D. R. (1967). Sleep laboratory adaptation in normal subjects and depressed patients (‘‘first night effect’’). Electroencephalography and Clinical Neurophysiology, 22, 556–558. Neylan, T. C., Marmar, C. R., Metzler, T. J., Weiss, D. S., Zatzick, D. F., Delucchi, K. L., et al. (1998). Sleep disturbances in the Vietnam

Adaptation effects to sleep studies and PTSD generation: Findings from a nationally representative sample of male Vietnam veterans. American Journal of Psychiatry, 155, 929–933. Neylan, T. C., Lenoci, M., Maglione, M. L., Rosenlicht, N. Z., Metzler, T. J., Otte, C., et al. (2003). Delta sleep response to metyrapone in post-traumatic stress disorder. Neuropsychopharmacology, 28, 1666–1676. Otte, C., Lenoci, M., Metzler, T., Yehuda, R., Marmar, C. R., & Neylan, T. C. (2005). Hypothalamic-pituitary-adrenal axis activity and sleep in posttraumatic stress disorder. Neuropsychopharmacology, 30, 1173–1180. Otte, C., Lenoci, M., Metzler, T., Yehuda, R., Marmar, C. R., & Neylan, T. C. (2007). Effects of metyrapone on hypothalamic-pituitary-adrenal axis and sleep in women with post-traumatic stress disorder. Biological Psychiatry, 61, 952–956. Rechtschaffen, A., & Kales, A. (1968). A manual of standardized terminology, techniques and scoring system for sleep stages of human subjects. Bethesda: US Dept. of Health Education and Welfare. Ross, R. J., Ball, W. A., Sanford, L. D., Morrison, A. R., Dinges, D. F., Silver, S. M., et al. (1999). Rapid eye movement sleep changes during the adaptation night in combat veterans with posttraumatic stress disorder. Biological Psychiatry, 45, 938–941. Roszell, D. K., McFall, M. E., & Malas, K. L. (1991). Frequency of symptoms and concurrent psychiatric disorder in Vietnam veterans with chronic PTSD. Hospital and Community Psychiatry, 42, 293–296. Saletu, B., Klosch, G., Gruber, G., Anderer, P., Udomratn, P., & Frey, R. (1996). First-night-effects on generalized anxiety disorder (GAD)-based insomnia, laboratory versus home sleep recordings. Sleep, 19, 691–697. Schmidt, H. S., & Kaelbling, R. (1971). The differential laboratory adaptation of sleep parameters. Biological Psychiatry, 3, 33–45. Se´rie`s, F., Marc, I., Cormier, Y., & La Forge, J. (1993). Utility of nocturnal home oximetry for case finding in patients with suspected sleep

1133 apnea hypopnea syndrome. Annals of Internal Medicine, 119, 449– 453. Sharpley, A. L., Solomon, R. A., & Cowen, P. J. (1988). Evaluation of first night effect using ambulatory monitoring and automatic sleep stage analysis. Sleep, 11, 273–276. Sheikh, J. I., Woodward, S.H, & Leskin, G. A. (2003). Sleep in posttraumatic stress disorder and panic: Convergence and divergence. Depression and Anxiety, 18, 187–197. Spitzer, R., Williams, J., Gibbon, M., & First, M. (1996). The Structured clinical interview for DSM-IV SCID: I. (SCD-1NP). New York: Biometrics Research Department, New York State Psychiatric Institute. Toussaint, M., Luthringer, R., Schaltenbrand, N., Carelli, G., Lainey, E., Jacqmin, A., et al. (1995). First-night effect in normal subjects and psychiatric inpatients. Sleep, 18, 463–469. Wauquier, A., van Sweden, B., Kerkhof, G. A., & Kamphuisen, H. A. (1991). Ambulatory first night sleep effect recording in the elderly. Behavioural Brain Research, 42, 7–11. Weathers, F. W., Keane, T. M., & Davidson, J. R. (2001). Clinicianadministered PTSD Scale: A review of the first ten years of research. Depression and Anxiety, 13, 132–156. Woodward, S. H., Bliwise, D. L., Friedman, M. J., & Gusman, F. D. (1996). First night effects in post-traumatic stress disorder inpatients. Sleep, 19, 312–317. Woodward, S. H., Stegman, W. K., Pavao, J. R., Arsenault, N. J., Hartl, T. L., Drescher, K. D., & Weaver, C. (2007). Self-selection bias in sleep and psychophysiological studies of posttraumatic stress disorder. Journal of Traumatic Stress, 20, 619–623. (Received June 19, 2009; Accepted October 12, 2009)

Psychophysiology, 47 (2010), 1134–1141. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01033.x

The temporal locus of the interaction between working memory consolidation and the attentional blink

ELKAN G. AKYU¨REK, MARCIN LESZCZYN´SKI, and ANNA SCHUBO¨ Department of Psychology, Ludwig Maximilian University, Munich, Germany

Abstract An increase in concurrent working memory load has been shown to amplify the attentional blink. The present study investigated the temporal locus of this phenomenon, by using a dual rapid serial visual presentation paradigm that enabled the measurement of lateralized event-related potentials. The P3 component was shown to be affected by both working memory load and the lag between the target stimuli, consistent with current models of temporal attention and a functional explanation of the P3 in terms of memory consolidation. P3 amplitude was reduced for short target lags and high memory loads. The P2 component was affected by lag only, and not memory load. Importantly, the N2pc component was modulated also by both lag and memory load. The results showed that early attentional processing (as marked by the N2pc) was suppressed by increased involvement of working memory, a phenomenon not well predicted by many current theories of temporal attention. Descriptors: Attentional blink, Working memory, Event-related potential, N2pc, P3

assumption that the current study sought to test. Before discussing the logic of this test in detail, it is appropriate to briefly characterize these existing models of the blink. Starting with the class of capacity-limited models, these typically assume that the cognitive system is limited at a late stage of processing, responsible for memory consolidation and response selection, in which only one item can be processed at a time. Early stage processing is thought to proceed relatively freely, but representations of stimuli that cannot proceed to the next stage are prone to decay and interference. By contrast, models that attempt to avoid capacity limitations as a causal factor assume that it is not a lack of a given cognitive capacity that causes the blink, but rather that it is a consequence of the configuration of the attentional filter, which selects information for further processing. In the model of Di Lollo and colleagues, an executive control function called the ‘‘central processor’’ is required to maintain an appropriate attentional set, and when this processor is occupied by the processing of the first target stimulus, the system temporarily loses control and becomes vulnerable to exogenous, distractor-generated triggers, which causes the blink (Di Lollo et al., 2005). The model of Olivers and Meeter (2008) takes a slightly different approach, in which it is assumed that the attentional filter is too slow to ‘‘boost’’ the representation of a target stimulus upon its detection, and ends up boosting the next (irrelevant) item instead. This situation is then imperfectly corrected by the issue of an inhibitory ‘‘bounce’’ signal, which causes the blink by suppressing subsequent incoming stimuli for some time. The discrepancy between these two types of models regarding the role of working memory in the emergence of the blink should be acknowledged at this point. Whereas the first, classic type of theory posits a major bottleneck at the consolidation of infor-

The study of temporal attention has enjoyed a surge of interest over the past two decades due to the discovery of the attentional blink phenomenon. The attentional blink (AB) is the marked difficulty of identifying the second of two briefly flashed, successive target stimuli if these follow each other within approximately 500 ms (Broadbent & Broadbent, 1987; Raymond, Shapiro, & Arnell, 1992). A host of theories have attempted to explain the emergence of the attentional blink. The majority of these accounts focused on cognitive capacity limitations (e.g., Bowman & Wyble, 2007; Chun & Potter, 1995; Jolicœur & Dell’Acqua, 1998; Shapiro, Raymond, & Arnell, 1994), but more recently alternative models that are (relatively) capacityfree have also been proposed (Di Lollo, Kawahara, Ghorashi, & Enns, 2005; Olivers & Meeter, 2008). Working memory is represented in both capacity-limited and capacity-agnostic models of the attentional blink, despite apparent discrepancies regarding its role. Without immediately going into more detail, it is important to stress that virtually all available models of either type hold that working memory operations are a relatively late process, in other words, taking place only after considerable attentional processing. As a consequence, any interference due to memory involvement during an attentional task is thus necessarily thought to have a late locus in the cognitive system. It is this A. S. and E. G. A were supported by the German Research Foundation (DFG) as part of the excellence cluster ‘‘Cognition for Technical Systems’’ (CoTeSys), project #148/433. M. L. was supported by the German Academic Exchange Service (DAAD), grant A/07/72330. Address correspondence to: Elkan Akyu¨rek, Department of Psychology, Experimental Psychology, University of Groningen, Grote Kruisstraat 2/1, 9712 TS Groningen, The Netherlands. Email: e.g. [email protected] 1134

Interaction between working memory and attentional blink mation into working memory, the second type of theory assumes that, although working memory limits may indeed exist, they are normally not the main cause of the blink. In a direct test of this issue, the impact of loading working memory on the deployment of temporal attention was investigated by Akyu¨rek, Hommel, and Jolicœur (2007). In this study, a classic attentional blink task was embedded within a memory-probe task. The memory task was contingent on the first target stimulus (T1), so that T1 served as the memory probe. Thus, participants were asked to detect T1 and to match its identity to a memory set presented at the beginning of each trial. The logic of this implementation was that this T1 contingency would lead observers to access their memory at the critical time; that is, they were using working memory during the attentional task, rather than having it as a relatively passive store. Increasing the (numerical) load on working memory load had a clear effect: A higher load impaired the identification of the second target stimulus (T2), and did so in particular at short T1-T2 lag, which resulted in a deeper attentional blink. The interaction between working memory and the attentional blink found here should be attributed to active use of memory, and not so much the mere maintenance of information, as previous studies indicated that the latter type of task does not interact with the blink (Akyu¨rek & Hommel, 2005, 2006). It would thus seem that theories attributing a critical role to working memory are vindicated. Yet, the behavioral evidence leaves some leeway in the theoretical interpretation. Even the nearly completely capacity-agnostic model of Olivers and Meeter (2008) assumes that processing delays could occur when working memory is filled, even if that does not normally cause the blink. As far as the study by Akyu¨rek et al. (2007) is concerned, the interaction between AB magnitude and working memory load could be attributed to a temporary and perhaps exceptional overload of memory. Note that this assumption is only different from the one made by classic capacity-limited models in that the latter assume a working memory bottleneck underlies the AB even without additional memory load, whereas the former does not. Either way, all of these current theories of the blink would place a working memory problem of any kind at a late phase of processing (but see also Wyble, Nieuwenstein, & Bowman, 2009), even though this has not been tested decisively. The available models would predict an initial phase of normal deployment of attention, with the memory consolidation problem arising only after feature identification. Early attentional processing is expected to be intact and fully completed. In classic models such as the one by Chun and Potter (1995), lack of memory (consolidation) resources would be a part of limited late stage (‘‘Stage 2’’) processing. In the model of Olivers and Meeter (2008), it would similarly be a problem originating at or beyond the gating function that controls entry to memory, after stimulus identification (but not response mapping) has taken place. To find the locus of interference in time, electrophysiology is a promising method, due to its ability to reveal the time course of processing in the brain with high precision. A number of studies on the attentional blink have employed electrophysiological measurements to date. The evidence they have brought forward is unfortunately mixed with regard to finding a single locus in time at which the blink occurs. Pioneering studies conducted by Luck, Vogel, and Shapiro (1996; Vogel, Luck, & Shapiro, 1998) showed that the P3 component of the event-related potential (ERP) is suppressed during the attentional blink, whereas the N400 component is not (see also Kranczioch, Debener, & Engel,

1135 2003). Since the P3 is associated with consolidation in working memory, these results fit existing theory quite well (Donchin & Coles, 1988; Kok, 2001; Polich, 2007). The N400 component is associated with semantic processing, which presumably is a relatively advanced type of stimulus processing (Kutas & Hillyard, 1980). Its survival during the blink is somewhat paradoxical. However, a recent study by Giesbrecht, Sy, and Elliott (2007) has shown that the N400 does not survive if perceptual load is high (i.e., the target is hard to discern). They argued that the survival of the N400 in previous work was indicative of a postperceptual locus of selection, whereas high perceptual load such as used in their study necessitates selection at an earlier, perceptual level. The evidence produced by these studies concerning earlier components is less clear. The P2 has been reported to decrease during the blink, as estimated by T1-T2 lag, by Vogel et al. (1998), but a slightly different approach based on the analysis of correct and incorrect trials by Kranczioch et al. (2003) has lead to the idea that blink-related processing starts only after the P2 time-window (beyond 300 ms post-T2). However, subsequent work has revealed earlier effects of the blink on the ERP, casting doubt on the hypothesis that the blink is exclusively the result of a late cognitive bottleneck. Sergent, Baillet, and Dehaene (2005) presented a thorough analysis of ERP components elicited in a dual target-mask paradigm, and found evidence for a divergence of the ERP between blink and no-blink trials at around 270 ms post-T2, around the time of the N2 component. Still earlier components were observed to be affected by the blink in a series of studies by Jolicœur, Dell’Acqua, and colleagues (Dell’Acqua, Sessa, Jolicœur, & Robitaille, 2006; Jolicœur, Sessa, Dell’Acqua, & Robitaille, 2006). In their studies, a dual stream rapid serial visual presentation (RSVP) paradigm was employed. This design carries the benefit that it allows for the measurement of lateralized attention, as indexed by the early N2pc component (Eimer, 1996; Kiss, van Velzen, & Eimer, 2008; Luck & Hillyard, 1994). In all of their experiments, the N2pc (peaking around 220 ms) was shown to be suppressed during the attentional blink. Since the N2pc component is thought to reflect the attentional processing of stimulus features at a particular location, a suppression of this component would seem to indicate a processing bottleneck at an earlier locus than what might be expected from a working memory depletion scenario. However, the electrophysiological evidence to date leaves the possibility open that the attentional blink is a multi-faceted phenomenon (a plausible claim made previously by, e.g., Kawahara, Enns, & Di Lollo, 2006). Namely, it might be that the late effects on the P3 component are indeed caused by the ‘real’ blink-related bottleneck, but that the earlier N2pc time range effects are due to other factors that perhaps contribute to the overall performance deficit, but that have their origin in stimulus-driven contingencies such as target or distractor salience. In a similar vein, for example, task and location switching have been shown to augment the blink, even though these are separable effects (Visser, Bischof, & Di Lollo, 1999). Thus, in order to establish the locus of the memory bottleneck in time, it has to be studied in isolation of other factors that may contribute to the blink. The electrophysiological correlate of the interaction between working memory and the attentional blink offers just this possibility. Therefore, to test whether working memory has a late locus of interference with the attentional blink, as available theories would predict, the present study set out to investigate this issue by using the interaction between working memory and the blink as observed in the paradigm by Akyu¨rek

1136 et al. (2007) as a starting point for an electrophysiological investigation. In the present experiment, participants were asked to memorize a set of letters, and to match the identity of T1 in a subsequent RSVP stream to this memory set. The behavioral effect of both T1-T2 lag and that of memory set size (load) on T2 identification accuracy as well as the T2-locked ERP was examined.

Method Participants Twenty-two students (11 female, 11 male) at the Ludwig Maximilian University Munich participated in the experiment for course credit or monetary compensation. The electroencephalogram (EEG) data from the first participant were discarded because of a programming error that (only) affected the timing signals sent from the stimulus computer to the EEG acquisition computer. The behavioral data were retained. Participants were unaware of the purpose of the experiment and reported normal or corrected-to-normal vision. Mean age was 23.5 years (range 18–26 years). Apparatus and Stimuli Participants were individually seated in a comfortable chair in an acoustically and electrically shielded testing chamber that was dimly lit, at a distance of approximately 100 cm from the screen. The 20’’ CRT screen was driven by a Core 2 Duo computer with a discrete graphics board, and refreshed at a frequency of 100 Hz with a resolution of 800 ! 600 pixels in 16 bit color. The experiment was programmed in E-Prime 1.2. Responses were logged on a standard USB keyboard polling at 125 Hz. A light gray background was maintained throughout the experimental trials. A plus sign (‘‘1’’) in 18 pt. Courier New font served as the fixation cross. The stimuli presented as the memory set were 36 pt. size red letters. The stimuli in the dual RSVP streams consisted mostly of 36 pt. size black letters. Amidst these black letters, T1 consisted of two identical red letters presented simultaneously in each stream, and T2 was a single red digit number. The letters for the memory set, and for each of the RSVP streams separately, were randomly chosen from all the letters of the alphabet, without replacement. The digit number for T2 was chosen from 2–9. The RSVP streams were aligned to the center of the screen and separated on the horizontal plane by 128 pixels. Procedure and Design The experiment started with 24 practice trials, which were excluded from all analyses, after which a total of 1200 experimental trials followed. There were two experimental variables (factors): Lag (0, 3, and 8) and memory Load (1 and 4). Lag 0 trials consisted of ‘catch’ trials, in which no T2 was shown and a normal distractor took its place. As explained below, catch trials were used to compute difference waves. T1 could appear at the 5th or 7th position in the stream, and it could either match or not match the memory set. T2 could appear in the left or the right stream. All of these trial types were randomly mixed and evenly distributed. Trials were self-paced and divided into 5 blocks of 240 trials each. Between blocks of trials, participants were encouraged to take a break. Participants initiated each trial by pressing ‘‘Enter.’’ Each trial started with a blank screen of 100

E. G. Akyu¨rek et al. ms, which was followed by the working memory set consisting of either one or four letters, which remained on screen for 1000 ms and was followed by an 800-ms blank screen. A fixation cross of 200 ms preceded the RSVP streams, and remained visible throughout the streams. The streams consisted of 20 stimuli each presented for 70 ms and followed by a 30-ms blank. After a delay of 100 ms, two successive response screens for T1 and T2 were presented for 1500 ms each, or until a response was given. For T1, participants were asked to judge whether T1 was part of the memory set or not. For T2, the task was to identify the digit number. At the end of the trial, a 200-ms blank pause allowed for the transmission of response signals to the EEG acquisition computer. Figure 1 shows a schematic representation of the main structure of the experimental trials. Mean accuracy was analyzed using repeated measures analyses of variance (ANOVAs), with the factor T2 Lag (Lag 3 or Lag 8), and the factor Memory Load (1 or 4 items). Comparisons were made between short and long lags to estimate the effect of the blink. Note that this approach is conservative by design, as it includes contributions from both correct and incorrect trials at each lag.1 In case of a significant test of sphericity, the degrees of freedom were adjusted using the Greenhouse-Geisser epsilon correction.

Electrophysiological Recording and Data Analysis EEG was recorded with Ag-AgCl electrodes from 64 electrodes according to the extended International 10–20 System. The electrodes were referenced to Cz and re-referenced offline to the average of both mastoids. Horizontal electro-oculogram (EOG) was recorded from the outer canthi of the eyes and the vertical EOG from above and below the left eye. Electrode impedance was kept below 5 kO. The amplifier used a 125-Hz cut-off and a 0.1-Hz high-pass filter. EEG was recorded at a frequency of 500 Hz. The data were filtered off-line with a 40-Hz low-pass filter at " 12 dB (48 dB/oct roll-off), and a 0.1-Hz high-pass at " 6 dB (24 dB/oct roll-off). EEG was averaged into 1000 ms segments, starting 200 ms prior to the onset of T2 and ending 800 ms afterwards. Ocular artifacts (blinks and eye movements) were corrected using the Gratton-Coles procedure (Gratton, Coles, & Donchin, 1983). For each electrode, trials with amplitudes exceeding # 80 mV, voltage steps exceeding # 50 mV between two sampling points, and trials with voltages lower than 0.10 mV for a 100-ms interval were excluded from analysis. A 200-ms pre-T1 interval (included in the artifact rejection procedure) was used for baseline correction. This interval was chosen rather than one 1 Although it is common to report conditional T2 performance (i.e., T2|T1), this approach was not taken in the present paradigm. The logic of the conditional approach is that only when T1 is reported correctly can one be sure that it has an effect on T2. Even if this is true, there is another method to ascertain the same: The difficulty of perceiving T2 depends critically on the temporal distance between T1 and T2. If T1 is not perceived, lag is meaningless. Thus, taking lag as an indicator of the blink is, in fact, a more conservative approach, as it includes trials on which T1 was perhaps perceived to some degree, but not sufficiently so for correct report, in which case the blink effect would be reduced. In the present study, there was, furthermore, a specific reason for not taking the conditional approach: The working memory task caused differences in T1 performance by design. When taking only T1-correct trials, low and high memory load trials would be composed of a systematically different number of trials, which might distort the averages. By taking the cautious approach of using all trials instead, this potential confound was avoided.

Interaction between working memory and attentional blink

-response...

V + L X + N ___ T2 4 + H ... T + W C + D ___ T1 F + F ... P + S ___ 70 + 30 ms B + K ___ 200 ms

+

___ 800 ms ___ 1000 ms

A F G K or:

F

Figure 1. Schematic representation of the experimental trial structure. First, a memory set of 1 or 4 letters was shown. Second, after a delay and a fixation cross, two rapid serial visual presentation streams were presented (20 frames). Three dots (. . .) indicate the variable presence of several more intermediate RSVP frames. In the actual experiment, targets were red. Finally, two response screens queried the observers for their input.

preceding T2 to avoid Lag-dependent contamination of the baseline caused by the different temporal onsets of T1. Separate ANOVAs were performed for mean amplitude values obtained in the time windows corresponding to the N2pc (150–260 ms after the onset of T2), the central-anterior P2 (170– 260 ms), and the parietal P3 (320–500 ms). These time windows were chosen to be compatible with existing studies of these components, while fitting the presently observed peaks. The following electrodes were chosen for analysis: The PO7/PO8 pair for the N2pc, FCz for the P2, and Pz for the P3. The electrodes were chosen a priori to match those commonly used for these components. To calculate the N2pc, ipsilateral waveforms (i.e., recorded from the left hemisphere electrode site while T2 was in the left visual field, and the same for right hemisphere site and right visual field) were subtracted from contralateral waveforms (i.e., left hemisphere electrode site and right visual field, etc.). The resulting N2pc difference waveform reflects lateralized attention. The non-lateralized P2 and P3 components were calculated by means of another difference wave procedure. Since the dual RSVP paradigm elicits individual ERPs for each stimulus in the stream, the ERP of the target of interest is drowned in this activity. To get rid of the ERP components elicited by irrelevant stimuli, the ERP of the catch trials in which no T2 was shown was subtracted from the ERP of T2 trials (a very similar approach

1137 was used by Vogel et al., 1998). The logic of this subtraction is that, because linear summation applies to EEG data, the ERPs to the irrelevant items cancel out against each other while the ERP to T2 is preserved. For reference, a set of raw ERP traces has been included as an Appendix.

Results and Discussion Behavioral Results Because of the high number of trials (1200) that was used to boost the signal-to-noise ratio for the ERP analyses, the behavioral analyses discerned even relatively small differences reliably. Figure 2 shows the performance on T1 (left panel) and T2 (right panel) as a function of lag. T1 accuracy showed significant effects of both Lag, F(1,21) 5 10.88, MSE 5 .001, po.005, and Load, F(1,1) 5 164.34, MSE 5 .003, po.001. Performance was higher at Lag 8 (75.7%) than at Lag 3 (73.9%), and higher with low memory load (81.8%) than with high load (67.8%). Overall, the T1 task was relatively difficult for the observers. The lag effect on T1 reflected a degree of competition between targets, as has been found in RSVP studies before (Hommel & Akyu¨rek, 2005; Potter, Staub, & O’Connor, 2002). The factors Lag and Load also interacted reliably, F(1,21) 5 5.47, MSE 5 .001, po.05. The interaction reflected a larger difference between high and low memory load at Lag 3 (15.2%) compared to Lag 8 (12.8%). T2 identification accuracy was likewise affected by Lag, F(1,21) 5 29.83, MSE 5 .003, po.001, and Load, F(1,21) 5 18.80, MSE 5 .004, po.001. T2 identification performance was relatively high overall, probably due to the salience of T2 (red among black). Nonetheless, a blink was observed at Lag 3 where performance averaged 81.5%, compared to 88.3% at Lag 8. Mean performance was affected to a similar degree by Load; a high memory load resulted in 81.9% accuracy, and a low load in 87.9%. The interaction between Lag and Load was reliable as well, F(1,21) 5 19.67, MSE 5 .001, po.001. The blink was increased when memory load was high (76.6% vs. 87.1%, at Lag 3 and 8, respectively), compared to when it was low (86.3% vs. 89.5%). Comparison of the individual means showed that there was a significant blink even for the low memory load, t(21) 5 3.77, po.001. This pattern of performance replicated the results of Akyu¨rek et al. (2007). Electrophysiological Results The N2pc showed significant main effects of both factors: Lag, F(1,20) 5 38.44, MSE 5 .518, po.001, and Load, F(1,20) 5 17.33, MSE 5 .302, po.001. A short lag between T1 and T2 reduced N2pc amplitude (! .87 mVat Lag 3 compared to ! 1.84 mV at Lag 8). Similarly, a high memory load reduced N2pc amplitude as well (! 1.10 mV at Load 4 compared to ! 1.60 mV at Load 1). The interaction between Lag and Load was also significant, F(1,20) 5 4.67, MSE 5 .301, po.05. There was a stronger suppression of the N2pc by memory load at Lag 3.2 There was a ! .76 mV difference at Lag 3 compared to ! .24 mV at Lag 8. To make sure the interaction effect was not driven by the outer end of the waveform (where visual inspection of the 2 An overall analysis including the two neighboring electrode pairs of PO3/PO4 and O1/O2 also showed a reliable interaction effect, F(2,40) 5 9.814, MSE 5 .067, po.001, indicating that, although the PO7/PO8 paid might have driven the effect, the pattern was the same on the other pairs.

E. G. Akyu¨rek et al. 100

100

90

90 T2 Accuracy (%)

T1 Accuracy (%)

1138

80 70 60 50

3 Load 1 Load 4

8

80 70 60 50

3

Lag

8 Lag

Load 1 Load 4

Figure 2. Behavioral performance on the experimental task. Memory match performance on T1 (left panel) and identification accuracy on T2 (right panel) is plotted separately for high and low memory loads (solid and dashed lines, respectively), as a function of T1-T2 lag. Error bars represent one standard error of the mean.

waveform might give rise to this suspicion), we conducted a second analysis in which we considered only the rising flank of the N2pc, by cutting the analysis time-window in half (i.e., 150–200 ms). This analysis revealed that the interaction between Lag and Load was reliable there also, F(1,20) 5 7.22, MSE 5 .397, po.05. The suppression of the N2pc by Load at Lag 3 was ! .92 mV, compared to ! .18 mV at Lag 8. The N2pc results thus showed clear effects of working memory load and lag, indicating that relatively early attentional processing was affected by both. Figure 3 (upper left panel) shows lateralized differential amplitude as a function of time for the PO7/PO8 electrode pair. For the P2 component, Lag had a reliable main effect, F(1,20) 5 13.53, MSE 5 1.934, po.001. At Lag 3, P2 amplitude was close to zero (.11 mV), compared to the amplitude at Lag 8 (1.23 mV). Neither Load (Fo1) nor its interaction with Lag was reliable (Fo2.5). The results indicated that memory load did not have a detectable effect on P2 amplitude, even though the attentional blink (i.e., Lag) did. Figure 3 (upper right panel) shows ERP amplitude as a function of time for the FCz electrode. The analysis of the P3 component showed significant effects of Lag, F(1,20) 5 20.06, MSE 5 2.236, po.001, and of Load, F(1,20) 5 4.26, MSE 5 3.184, po.05, as was expected. P3 amplitude averaged 2.87 mVat Lag 3, compared to 4.34 mVat Lag 8. A high memory load decreased the P3 amplitude (3.20 mV), compared to a low load (4.04 mV). The lag effect replicated established findings that the P3 is suppressed during the blink (e.g., Vogel et al., 1998). The effect of memory load was new, yet in the line of expectations given the functional attribution of the P3 to memory consolidation (e.g., Kok, 2001). The fact that T2 was rather salient in the presentation stream in the present paradigm likely prevented a complete suppression of the P3, as the component remained visible even when suppression was maximal. Figure 3 (lower panel) shows ERP amplitude as a function of time for the Pz electrode. There was a hint of residual RSVP noise in the waveforms, which may reflect a less-than-perfect outcome of the subtraction procedure. Since the number of catch trials amounted to only 1/3 of all trials, there may have been slight amplitude differences between these trials and the T2present trials related to trial frequency. However, other than causing a small visual distortion, the residual activity did not influence the signal in a systematic way.

Current source density maps falling within the considered time windows for all four conditions are also shown in Figure 3. Since the underlying data for these maps consisted of difference waves, the maps should be considered with some caution. There was evidence for amplitude changes between lags as well as memory loads, which was consistent with the ERP analyses.

General Discussion The present experiment intended to investigate the temporal locus of the interaction between working memory and the attentional blink. The results clearly showed that attentional processing is disrupted by working memory load at an early stage of processing. The primary evidence concerned the suppression of the N2pc component at high memory load, in addition to the suppression effect of short temporal lag. The lag effect has been demonstrated previously (Dell’Acqua et al., 2006; Jolicœur et al., 2006), but the memory load effect is the first evidence of an early effect of memory load on attentional processing in a dual-task situation, starting at 190 ms post-T2. In the present paradigm, the N2pc (like the P3) was attenuated rather than eliminated, which was presumably due to the salience of T2. Note that the salience of T2 would work against the hypothesis that memory processing should affect the N2pc, not for it. In any case, the results indicated that stimulus-driven salience is able to elicit an N2pc even when attention is taxed. The early effect of working memory load on attentional processing is not well predicted by the vast majority of available theories of temporal attention. Even though N2pc suppression during the blink has been observed previously (for T1-T2 lag), this effect could have been driven by a different, early component of the blink, such as backward masking (e.g., Kawahara et al., 2006), and thus does not necessarily connect a working memory bottleneck to the N2pc, while the present results do in two ways: First by the existence of a memory load main effect on the N2pc, and second by the interaction effect of memory load with lag, showing increased N2pc suppression at short lag. This connection is indeed notable, given the functional ties of the N2pc component to spatially specific processing of stimulus features (Kiss et al., 2008), rather than to memory operations. Indeed, it has been claimed that the N2pc is distinctly insensitive to memory pro-

Interaction between working memory and attentional blink

1139

178 – 198 ms

204 – 224 ms

2 1

µV

µV

0 –1 –2 –3 –4

N2pc PO7 / PO8 0

200

400

600

8 7 6 5 4 3 2 1 0 –1 –2 –200

P2

FCz 0

200

ms Lag 3 Load 1 Lag 3 Load 4 Lag 8 Load 1 Lag 8 Load 4

ms

400

600

800

Lag 3 Load 1 Lag 3 Load 4 Lag 8 Load 1 Lag 8 Load 4

µV

412 – 432ms 8 7 6 5 4 3 2 1 0 –1 –2 –200

P3

Pz 0

200

ms

400

600

800

Lag 3 Load 1 Lag 3 Load 4 Lag 8 Load 1 Lag 8 Load 4 Figure 3. N2pc contra- minus ipsilateral difference waveforms (DmV) as a function of time in ms, recorded at the PO7/PO8 electrode pair (upper left). ERP amplitude in mV as a function of time in ms, derived from a difference wave procedure (see main text), recorded at FCz (upper right), and at Pz (lower panel). Time zero denotes T2 onset, and the box outline represents the analysis window. Topographical map insets were constructed using spherical spline interpolation and represent a 20-ms average centered on the observed waveform peaks of the difference waves (see main text). For each graph, from left to right, the maps represent Lag 3 & Load 1, Lag 3 & Load 4, Lag 8 & Load 1, and Lag 8 & Load 4.

cesses (Jolicœur, Brisson, & Robitaille, 2008). It is not claimed here that the N2pc reflects working memory consolidation, however. Recall that the experimental paradigm required access to working memory at the time of T1: The deployment of this process was then measured on the ERP of T2. Thus, the conclusion must be that working memory access impaired the deployment of spatially selective feature processing on subsequent stimuliFin other words, the results showed an effect of memory load on temporal attention. The effects on the presently observed P2 component were somewhat of a puzzle. The amplitude of the P2 was affected by lag, but not by memory load. The functional significance of the P2 is unfortunately unclear, and the component is not always

observed in attentional blink tasks (Kranczioch et al., 2003; Sergent et al., 2005). Yet, the current results do suggest that the cognitive processes underlying the P2 component are qualitatively different from those associated with the P3 and N2pc. More research is obviously needed to investigate the role of the P2 in temporal attention. Finally, replicating previous work, the present study showed the suppression of the P3 due to short temporal lag (Luck et al., 1996; Kranczioch et al., 2003; Vogel et al., 1998). The presence of the lag-induced P3 suppression confirmed the existence of a modulation at the stage of working memory consolidation. Such modulation is consistent with all available models of the blink, as the P3 can be thought of as the final result of attentional

1140

E. G. Akyu¨rek et al.

processing. In other words, if interference is present at an earlier stage (e.g., at the attentional filter), this would affect downstream operations as well; filtered distractors do not need to be consolidated. The presently observed effect of memory load on the P3 component supported the behavioral results that pointed to an interaction between memory and attentional performance. Although an interaction effect, such as observed at the N2pc, was not apparent for the P3, this need not be surprising even if one were to expect a strong correlation between all ERP components and behavioral performance. As a consequence of attentional deployment, the P3 effect may be understood as another additive factor giving rise to the observed behavioral difficulty. Indeed, given the presence of earlier modulations of the ERP (i.e., of the N2pc), doubt is cast on the idea that increased working memory demands of T1 interfere exclusively with consolidation of T2. Taken together, the present results demonstrated an early locus of interference between working memory and temporal attention. This finding is not well predicted by several models of the attentional blink (Bowman & Wyble, 2007; Chun & Potter, 1995; Di Lollo et al., 2005; Jolicœur & Dell’Acqua, 1998; Olivers & Meeter, 2008; Shapiro et al., 1994). Regardless of whether working memory consolidation plays the causal role in the development of the blink or not, none of these models would predict that

relatively early feature processing would be impaired by working memory load. In classic capacity-limited models, the interference suffered during the blink is located in late stage processing, in which consolidation, response selection, and full stimulus binding and identification take place. Feature processing is thought to be unimpaired (e.g., Chun & Potter, 1995). A more recent model by Wyble and colleagues (2009) does offer a way to incorporate the present findings. In this model, a direct link exists between working memory encoding and attention, such that early attentional deployment could be stalled by working memory load. Thus, this newly developed model shows an adaptation that other models would need to make as well. In capacity-agnostic models, working memory interference would be just a limit on the number of items that can be stored in memory (Olivers & Meeter, 2008). Running into such a limit is not normally supposed to give rise to the blink, and it would not be expected to interfere with feature processing. The present results show that there are links between late processing stages involved in accessing working memory and earlier ones in dual-tasks, and during the attentional blink. Theories of temporal attention should thus accommodate the idea that spatially specific attention not only freezes as a consequence of the blink, but also as a consequence of working memory operations.

REFERENCES Akyu¨rek, E. G., & Hommel, B. (2005). Short-term memory and the attentional blink: Capacity versus content. Memory & Cognition, 33, 654–663. Akyu¨rek, E. G., & Hommel, B. (2006). Memory operations in rapid serial visual presentation. European Journal of Cognitive Psychology, 18, 520–536. Akyu¨rek, E. G., Hommel, B., & Jolicœur, P. (2007). Direct evidence for a role of working memory in the attentional blink. Memory & Cognition, 35, 621–627. Bowman, H., & Wyble, B. (2007). The simultaneous type, serial token model of temporal attention and working memory. Psychological Review, 114, 38–70. Broadbent, D. E., & Broadbent, M. H. P. (1987). From detection to identification: Response to multiple targets in rapid serial visual presentation. Perception & Psychophysics, 42, 105–113. Chun, M. M., & Potter, M. C. (1995). A two-stage model for multiple target detection in rapid serial visual presentation. Journal of Experimental Psychology: Human Perception and Performance, 21, 109– 127. Dell’Acqua, R., Sessa, P., Jolicœur, P., & Robitaille, N. (2006). Spatial attention freezes during the attention blink. Psychophysiology, 43, 394–400. Di Lollo, V., Kawahara, J.-I., Ghorashi, S. M. S., & Enns, J. T. (2005). The attentional blink: Resource depletion or temporary loss of control? Psychological Research, 69, 191–200. Donchin, E., & Coles, M. G. (1988). Is the P300 component a manifestation of context updating? Behavioral & Brain Sciences, 11, 357– 374. Eimer, M. (1996). The N2pc component as an indicator of attentional selectivity. Electroencephalography and Clinical Neurophysiology, 99, 225–234. Giesbrecht, B., Sy, J. L., & Elliott, J. C. (2007). Electrophysiological evidence for both perceptual and postperceptual selection during the attentional blink. Journal of Cognitive Neuroscience, 19, 2005–2018. Gratton, G., Coles, M. G., & Donchin, E. (1983). A new method for offline removal of ocular artifact. Electroencephalography and Clinical Neurophysiology, 55, 468–484. Hommel, B., & Akyu¨rek, E. G. (2005). Lag 1 sparing in the attentional blink: Benefits and costs of integrating two events into a single episode. Quarterly Journal of Experimental Psychology, 58A, 1415– 1433.

Jolicœur, P., Brisson, B., & Robitaille, N. (2008). Dissociation of the N2pc and sustained posterior contralateral negativity in a choice response task. Brain Research, 1215, 160–172. Jolicœur, P., & Dell’Acqua, R. (1998). The demonstration of short-term consolidation. Cognitive Psychology, 36, 138–202. Jolicœur, P., Sessa, P., Dell’Acqua, R., & Robitaille, N. (2006). On the control of visual spatial attention: Evidence from human electrophysiology. Psychological Research, 70, 414–424. Kawahara, J.-I., Enns, J. T., & Di Lollo, V. (2006). The attentional blink is not a unitary phenomenon. Psychological Research, 70, 405–413. Kiss, M., van Velzen, J., & Eimer, M. (2008). The N2pc component and its links to attention shifts and spatially selective visual processing. Psychophysiology, 45, 240–249. Kok, A. (2001). On the utility of P3 amplitude as a measure of processing capacity. Psychophysiology, 38, 557–577. Kranczioch, C., Debener, S., & Engel, E. K. (2003). Event-related potential correlates of the attentional blink phenomenon. Cognitive Brain Research, 17, 177–187. Kutas, M., & Hillyard, S. A. (1980). Reading senseless sentences: Brain potentials reflect semantic incongruity. Science, 207, 203–205. Luck, S. J., & Hillyard, S. A. (1994). Spatial filtering during visual search: Evidence from human electrophysiology. Journal of Experimental Psychology: Human Perception and Performance, 20, 1000–1014. Luck, S. J., Vogel, E. K., & Shapiro, K. L. (1996). Word meanings can be accessed but not reported during the attentional blink. Nature, 383, 616–618. Olivers, C. N. L., & Meeter, M. (2008). A boost and bounce theory of temporal attention. Psychological Review, 115, 836–863. Polich, J. (2007). Updating P300: An integrative theory of P3a and P3b. Clinical Neurophysiology, 118, 2128–2148. Potter, M. C., Staub, A., & O’Connor, D. H. (2002). The time course of competition for attention: Attention is initially labile. Journal of Experimental Psychology: Human Perception and Performance, 28, 1149–1162. Raymond, J. E., Shapiro, K. L., & Arnell, K. M. (1992). Temporary suppression of visual processing in an RSVP task: An attentional blink? Journal of Experimental Psychology: Human Perception and Performance, 18, 849–860. Sergent, C., Baillet, S., & Dehaene, S. (2005). Timing of the brain events underlying consciousness during the attentional blink. Nature Neuroscience, 8, 1391–1400.

Interaction between working memory and attentional blink

1141

Shapiro, K. L., Raymond, J. E., & Arnell, K. M. (1994). Attention to visual pattern information produces the attentional blink in rapid serial visual presentation. Journal of Experimental Psychology: Human Perception and Performance, 20, 357–371. Visser, T. A. W., Bischof, W. F., & Di Lollo, V. (1999). Attentional switching in spatial and non-spatial domains: Evidence from the attentional blink. Psychological Bulletin, 125, 458–469. Vogel, E. K., Luck, S. J., & Shapiro, K. L. (1998). Electrophysiological evidence for a postperceptual locus of suppression during the attent-

ional blink. Journal of Experimental Psychology: Human Perception and Performance, 24, 1656–1674. Wyble, B., Nieuwenstein, M. R., & Bowman, H. (2009). The attentional blink provides episodic distinctiveness: Sparing at a cost. Journal of Experimental Psychology: Human Perception and Performance, 35, 787–807. (Received October 7, 2009; Accepted January 7, 2010)

APPENDIX

8 7 6 5 4 3 2 1 0 –1 –2 –200

µV

µV

Figure A1

Pz 0

200

400

600

800

8 7 6 5 4 3 2 1 0 –1 –2 –200

Pz 0

200

600

800

Raw Lag 3 Load 1 Raw Lag 3 Load 4 Raw Lag 8 Load 1 Raw Lag 8 Load 4

Raw Lag 3 Load 1 Raw Lag 3 Load 4 Raw Lag 8 Load 1 Raw Lag 8 Load 4

µV

400 ms

ms

8 7 6 5 4 3 2 1 0 –1 –2 –200

Pz 0

200

400

600

800

ms Raw Lag 3 Load 1 Raw Lag 3 Load 4 Raw Lag 8 Load 1 Raw Lag 8 Load 4 Figure A1. Raw ERP amplitude relative to T2 onset in mVas a function of time in ms recorded at the Pz electrode (upper left panel). Raw ERP amplitude used for difference wave subtraction (see main text) in mVas a function of time in ms recorded at Pz (upper right panel). Raw ERP amplitude relative to T1 onset in mV as a function of time in ms recorded at Pz (lower panel). Identical to the processing of all the presently reported ERP data, the baseline used here was the 200 ms interval before T1 onset.

Psychophysiology, 47 (2010), 1142–1150. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01015.x

Validation of ICA as a tool to remove eye movement artifacts from EEG/ERP

MAARTEN MENNES,a HEIDI WOUTERS,a BART VANRUMSTE,b,c LIEVEN LAGAE,a and PETER STIERSa,d a

Department of Woman & Child, Section Paediatric Neurology, K.U. Leuven, Leuven, Belgium SCD/ESAT, K.U. Leuven, Leuven, Belgium MOBILAB, Katholieke Hogeschool Kempen, Geel, Belgium d Department of Neuropsychology and Psychopharmacology, Maastricht University, Maastricht, The Netherlands b c

Abstract Eye movement artifacts in electroencephalogram (EEG) recordings can greatly distort grand mean event-related potential (ERP) waveforms. Different techniques have been suggested to remove these artifacts prior to ERP analysis. Independent component analysis (ICA) is suggested as an alternative method to ‘‘filter’’ eye movement artifacts out of the EEG, preserving the brain activity of interest and preserving all trials. However, the identification of artifact components is not always straightforward. Here, we compared eye movement artifact removal by ICA compiled on 10 s of EEG, on eye movement epochs, or on the complete EEG recording to the removal of eye movement artifacts by rejecting trials or by the Gratton and Coles method. ICA performed as well as the Gratton and Coles method. By selecting only eye movement epochs for ICA compilation, we were able to facilitate the identification of components representing eye movement artifacts. Descriptors: EEG/ERP, Artifact removal, Validation, Blink artifact, Independent component analysis

with clinical populations or children, for whom it is difficult to refrain from blinking. A method that allows the cleaning up of the contaminated trials by filtering or removing the eye movement artifacts from the data would, therefore, be highly beneficial in terms of research effort efficiency. Consequently, several methods that try to remove as much of the eye movement artifacts without reducing data quality have been introduced (Berg & Scherg, 1994; Croft & Barry, 2000; Gratton, Coles, & Donchin, 1983). More recently, independent component analysis (ICA) was introduced as a method for separating artifactual data (blinks, eye movements, and muscle and line noise) from useful brain activity in EEG recordings (Flexer, Bauer, Pripfl, & Dorffner, 2005; Ford, Sands, & Lew, 2004; Frank & Frishkoff, 2007; Jung et al., 2000a, 2000b; Vigario, 1997). ICA is a data-driven blind source separation method that is applied on biomedical data such as EEG, ERP, magnetoencephalography, and functional magnetic resonance imaging (fMRI; Bell & Sejnowski, 1995; Stone, 2002; Vigario, Sarela, Jousmaki, Hamalainen, & Oja, 2000). It is used to reduce large sets of research data to a small number of independent components in order to ease interpretation or to locate possible brain sources for a measured signal (Esposito et al., 2003; Iidaka, Matsumoto, Nogawa, Yamamoto, & Sadato, 2006; Kansaku et al., 2005; Makeig et al., 2004; Zeki, Perry, & Bartels, 2003). It appears however, that only few research groups effectively apply ICA or other blind source separation algorithms (e.g., SOBI, JADE) to remove eye movement artifacts from their EEG

The increasing popularity of the electroencephalogram (EEG) and event-related potentials (ERP) for research and clinical purposes has resulted in better acquisition devices minimizing the amount of noise picked up during an EEG recording. However, an important source of noise that cannot be avoided by better amplifiers or electromagnetically shielded rooms is the electrical activity that is associated with eye movements. Among eye movements, blinks cause the largest distortions, mainly because of the movement of the eyelids across the surface of the eyes, but also saccades can cause large distortions in EEG signals and ERP waveforms, particularly at frontal electrodes (Iwasaki et al., 2005). Because these artifacts can hamper correct interpretation, it is best to remove them from the data. In most ERP studies this is done by rejecting trials containing these artifacts. Such trials can be detected by setting an absolute amplitude threshold. However, the amount of data lost by rejecting trials containing eye movements can be unacceptably high, especially when one is working The authors thank Maarten De Vos, Nikolai Novitski, Jennifer Ramautar, Stefan Sunaert, Bea Van den Bergh, Katrien Vanderperren, Sabine Van Huffel, Anneleen Vergult, and Johan Wagemans for helpful discussions. This work is supported by grants from Fonds voor Wetenschappelijk Onderzoek, Vlaanderen (# G.0211.03 and #7.0008.03) and K.U. Leuven (IDO/05/010 EEG-fMRI). Address reprint request to: Maarten Mennes, Phyllis Green and Randolph Cowen Institute for Pediatric Neuroscience, NYU Child Study Center, 215 Lexington Ave, 14th Floor, New York, NY 10016, USA. E-mail: [email protected] 1142

Validation of ICA to remove eye movement artifacts data (Debener, Hine, Bleeck, & Eyles, 2008; Huang, Jung, Delorme, & Makeig, 2008; Jung et al., 2001; Joyce, Gorodnitsky, & Kutas, 2004; Slagter et al., 2007). We believe that the reason for this is, in some respect, inherent to ICA. Because it is a blind source decomposition method, the resulting independent components are statistically valid, but their physiological meaning is not always clear. It is up to the user to decide which components represent artifactual data and which brain activity of interest. However, ICA can be applied in several ways, which differ in the selection of data from the continuous data set entered into the decomposition analysis (Ille, Berg, & Scherg, 2002). Because this affects the salience of components representing the targeted artifact, the appropriate selection of data used for ICA provides a way to improve component selection and interpretation and, hence, the quality of artifact removal. In addition, entering only a part of the data into the ICA analysis can considerably reduce the computation time, thus compensating one of the drawbacks when one uses ICA to remove artifacts from continuous EEG data sets. Especially when one uses a large number of electrodes, the computation time and computer memory required to run ICA on a complete EEG data set could exceed commonly available resources. Short computation times would also allow ICA to be used for near real-time correction of EEG data. Here, we introduce a procedure for easily identifying eye movement artifacts among independent components. By applying ICA only on sections of data that were contaminated with eye movement artifacts (i.e., epochs selected around the eye movements; Verleger, Gasser, & Mo¨cks, 1982), we anticipated capturing the eye movement artifacts in easily identifiable components, leading to straightforward criteria to select components to remove, making quantitative selection of components viable. We compared the results of this procedure to the application of ICA on 10 s of EEG and on the complete continuous EEG data. In all these procedures, the resulting components were subsequently used to remove the eye movement artifacts from the continuous data. To underline the validity of our component selection procedure and the usefulness of ICA in general for the removal of eye movement artifacts, we compared the results of the eye artifact removal by ICA to (a) data uncorrected for eye movements, (b) data corrected for eye movements by rejecting EEG epochs with an amplitude exceeding ! 100 mV, and (c) to the results of eye artifact removal by the well-established Gratton and Coles method (Gratton et al., 1983). Various aspects of the performance of these procedures were evaluated. (1) We visually evaluated the cleaned continuous data. (2) Heavily contaminated trials were compared to trials with only little contamination by eye movements. (3) The impact of the artifact removal on trials with little contamination was assessed. (4) The variation across trials after removal of the artifacts was measured. (5) The statistical sensitivity of the data to experimentally induced effects was evaluated.

Methods Participants Data are presented for 6 right-handed adults (2 female; age: 24–34 years). None had a history of substance abuse or psychiatric or neurological disorder. All participants had normal or correctedto-normal vision. Data were collected in the course of ongoing studies within the laboratory. The study was approved by the local ethics committee, and all participants gave their informed consent.

1143 EEG Recordings Nineteen electrodes were applied using the standard 10–20 system of electrode placement: Fp1, Fp2, F3, F4, F7, F8, C3, C4, T3, T4, P3, P4, T5, T6, O1, O2, Fz, Cz, and Pz. A ground electrode was placed on the forehead above the nose. Additionally two electrodes were placed on the outer canthi (horizontal electrooculogram [HEOG]) and two above and below the right eye (vertical electrooculogram [VEOG] 5 below eye-above eye) to detect horizontal and vertical eye movements. All electrodes were referenced to linked left and right mastoids, and all electrode impedances were below 5 kO at the start of the recording session. Sampling rate was 1000 Hz, with an analog bandpass of 0.095–70 Hz. Prior to removal of the eye movement artifacts, data were filtered off-line using a 30-Hz digital low-pass filter. Off-line analysis of the data, including removal of the eye artifacts, was performed using the EEGLAB v4.5 toolbox (Delorme & Makeig, 2004) under Matlab v7.0 (Mathworks, Natick, MA), except for the removal of the eye movement artifacts using the Gratton and Coles method, which was done with Brain Vision Analyzer 2 (Brain Products, Germany). Recently, a plug-in implementing the Gratton and Coles method also became available for EEGLAB (http://pinguin.uni-psych.gwdg.de/" ihrke/wiki/ index.php/Ocular_Correction_EEGlab_Plugin). Cued Attention Paradigm EEGs were recorded during a cued attention paradigm based on paradigms of Corbetta, Kincade, and Shulman (2002) and Posner (1980). Participants were asked to respond to a total of 150 stimuli appearing in the left or right visual field and were either unaware (no cue, 20%) or informed (cue, 80%) about the location at which the stimulus would appear. In 20% of all cued trials the cue was invalid, that is, the stimulus appeared on the side opposite to the one indicated by the cue. For ERP analysis, task-relevant epochs were extracted from 150 ms before to 800 ms after target stimulus onset. During the task, central fixation had to be maintained, and participants were encouraged to make as few eye movements as possible, including blinks. Procedures for Eye Movement Artifact Removal The correction for eye movement artifacts was done for each subject separately. An individual approach to eye movement artifacts is needed, as, for instance, differences in blink artifacts exist between participants because of differences in eyelid size and electrode position (Iwasaki et al., 2005). Figure 1 gives an overview of the six procedures to remove or correct for eye movement artifacts used in the current study. First, data were used in their ‘‘raw’’ format, with only the 30-Hz low-pass filter applied (RAW). Second, task-specific epochs were extracted from the continuous raw data. Subsequently epochs in which the maximum amplitude exceeded ! 100 mV were rejected for further analysis (REJECT). Third, the Gratton and Coles ocular correction algorithm as implemented in Brain Vision Analyzer 2 was applied to the raw continuous data (GC). In this algorithm event-related potentials of interest are first subtracted from the raw EEG and EOG signals. Afterward, the proportion of EOG that is represented in each of the EEG channels is calculated first for blinks and subsequently for saccades and subtracted from the EEG recording. Finally, the subtracted event-related potentials are again added to the recording. Details can be found in Gratton et al. (1983). In procedures four, five, and six, we used independent component analysis to remove eye movement artifacts. We used the

1144

M. Mennes et al. raw EEG recording

30Hz low-pass filter

task relevant epochs

10 seconds around 10th blink

1-RAW

eye movement epochs perform ICA select artifact components

reject pochs with max. amplitude exceeding (±) 100 µV

Gratton & Coles method

2 REJECT

3 GC

transfer ICA weights and sphere to continuous recording remove selected components from continuous recording

4 10sICA

5 eyeICA

6 cICA

Figure 1. Diagram of the procedures used to remove eye movement artifacts from the EEG recording.

standard runICA command integrated in EEGLAB, which uses the logistic infomax ICA algorithm of Bell and Sejnowski (1995). Before applying ICA for each individual subject, we first marked all eye movements in the EEG using a simple detection algorithm. We used the fast and extreme rise in amplitude typical for the representation of the different eye movements in the VEOG and HEOG channels compared to normal EEG signals to mark the occurrence of all individual eye movements (left, right, blink, down). Vertical eye movements were marked by searching VEOG. These included blinks, as blink artifacts had similar, although more extreme, characteristics as an upward directed eye movement. Left and right directed eye movements were marked by searching HEOG. Other EEG artifacts might also exhibit a fast and extreme rise in amplitude. Such artifacts were also marked as eye movements by our algorithm, but were easily identifiable as non-eye movements and removed from the final selection of eye movements upon visual inspection of the markers placed by our algorithm. For procedure four, we selected a 10 s long EEG epoch (Jung et al., 1998a, 1998b, 2000a) around the 10th blink present in the raw data. After performing ICA on this 10 s epoch and selecting components representing eye movement artifacts (see below), we transferred the calculated ICA weights and sphere to the raw data and removed the selected components by subtracting the projection of the artifactual components from the original data of the subject. The fifth procedure applied ICA on all eye movement artifacts. For each individual subject, the continuous EEG recording was epoched based on the marked occurrences of eye movements. Epochs were extracted from 500 ms before onset of an eye movement to 600 ms thereafter. The length of these epochs was chosen in order to include the whole eye movement plus a small portion of non-eye-movement data. This portion was included to provide ICA with data from sources other than the highly prominent eye movement artifact sources. Preliminary analyses indicated that including more then 500 ms of non-eye-movement artifact data did not result in an improved separation of the artifactual data. Using less than 500 ms resulted

in a less optimal separation of artifacts and cognitive sources in separate components. Subsequently, ICA was applied to these eye movement epochs. The mean total length of EEG data provided for ICA was 2.5 min, with a minimum of 1.1 min. The maximum of 4.2 min was still a lot shorter than the mean complete raw EEG (7 min). After the selection of components to remove, the ICA weights and sphere were transferred to the complete raw data and the selected components were removed (eyeICA). In the sixth and final procedure, we applied ICA to the complete raw data, subsequently subtracting components that were identified as representing eye movement artifacts (cICA). The mean length of the complete raw data was 7 min (minimum 5 6.9 min, maximum 5 7.2 min). Selection of Eye Artifact Components from the ICA For each ICA method we used the same procedure to identify components that represented eye movement artifacts. First, we selected components that each explained at least 15% of the variance of the EEG signal in VEOG and whose summed explained variance in that channel exceeded 90%. Usually one or two components were selected. In addition, we selected all other components that individually explained more than 15% of the variance of the EEG signal in HEOG and together with the components already selected from VEOG explained over 90% of the variance in HEOG. In most cases one additional component was selected, resulting in a total of two or three components that were removed from the data. We first looked at VEOG because components explaining most variance in this channel were easy to recognize, as they included the easily identifiable eyeblinks. However, eyeblinks are also registered in HEOG and, although registered to a much lesser extent, they sometimes cause crosstalk between components. The example below will clarify this issue and the selection procedure. For one of the participants the variance explained by the first four components was for the VEOG channel 82.4%, 16.5%, 0.2%, and 0.1% and for the HEOG channel 29.6%, 14.6%, 0.2%, and 53.9%. The first two

Validation of ICA to remove eye movement artifacts

1145

components reflected eyeblinks picked up by both VEOG and HEOG. Following the above selection procedure, we first selected Components 1 and 2, which together explained 82.4%116.5% 5 98.9% of the variance in VEOG. Second, Component 4 was selected based on the explained variance in HEOG. Together with Components 1 and 2, Component 4 explained 98.03% of the variance in HEOG (29.59%114.57% 153.87%). Thus, in this example, Components 1, 2, and 4 were removed from the continuous EEG data. The identification of eye movements, the application of ICA to the data, and the selection of components to remove were all integrated in a Matlab script. The only user input needed consisted of the evaluation and confirmation of the ICA components that were identified for removal. User confirmation of the artifactual ICA components was preserved to increase control over the correctness of the proposed solution. Scripts are available on request. Validation Procedures The quality of the removal of the eye movement artifacts was evaluated in five procedures as indicated in Table 1. First (Comparison 1, Table 1), grand average event-related potentials of the eye movement artifacts were calculated and compared across removal methods. The inspection of the raw average is suggested as a valid, initial form of evaluation (Croft & Barry, 2000; Verleger et al., 1982). The rejection method could not be included in this comparison. For the next two comparisons, we used a method previously described by Jung et al. (2000b). Using each trial’s absolute maximum potential values at the two EOG channels as an indication of the amount of contamination with eye movement artifacts, all validly cued trials of the administered cued attention paradigm were categorized as least, moderately, and most contaminated with eye movement artifacts. The categorization was based on the local maxima in a histogram of the maximal amplitude in all trials. The threshold separating the least and moderately contaminated trials ranged between 20 and 35 mV, and the threshold for separating the moderately and most contaminated trials ranged between 55 and 100 mV. Based on this method, two comparisons were made. In Comparison 2 we tested how efficient the removal procedures were by statistically comparing the least and most contaminated trials after removal of the eye movement artifacts according to the different removal procedures. Upon adequate removal of the eye movement artifacts, Table 1. Procedures Used to Evaluate the Performance of the Different Eye Artifact Removal Methods Evaluation procedure

Inspection

1. Grand average of eye movements 2. Least contaminated trials after removal of eye movement artifacts versus most contaminated trials after removal of eye movement artifacts 3. Least contaminated trials before removal of eye movement artifacts versus least contaminated trials after removal of eye movement artifacts 4. Standard deviation across 0–1000 ms after stimulus presentation 5. Valid cue trials versus invalid cue trials

Visual Statistical (SPM) Statistical (SPM) Statistical (ANOVA) Statistical (SPM)

Note: In Steps 1–4, only valid trials from the cued attention paradigm were included. In Steps 1, 4, and 5, all removal procedures were compared. The REJECT procedure could not be included in Steps 2 and 3.

the grand average of the trials labeled as least and most contaminated should not differ statistically. In Comparison 3, we tested whether the removal procedures affected data without eye movements, because the correction methods were applied to the complete raw EEG recording, including the least contaminated trials. We statistically compared the least contaminated trials before removal of the eye movement artifacts with the least contaminated trials after removal of these artifacts. It is assumed that the least contaminated trials contain almost no eye movements. Therefore, they should be barely affected by the different artifact removal methods. Differences in this comparison could indicate overcorrection of the data. For these two comparisons, all removal procedures were included except the rejection method, as this method resulted in the rejection of all most contaminated trials. The last two evaluations concerned the ERP characteristics of the data before and after removal of the eye movement artifacts. The first assessed the reduction of eye movement artifact-related variance in the data. By removing the eye movement artifacts, the standard deviation of the task-relevant epochs should decrease, because a large portion of noise is removed. This evaluation does not consider whether the different removal procedures might have distorted the underlying neural potentials (this is evaluated in Comparison 3). Here, we compared the standard deviation for each removal procedure calculated across all validly cued trials and all time points within the first second (1000 ms) after stimulus presentation (Comparison 4). A decrease in standard deviation while the number of trials remains constant should increase the statistical power of an analysis. This consideration was assessed in the final evaluation (Comparison 5) by statistically comparing the validly and invalidly cued trials of the cued attention paradigm after eye movement artifact removal. Comparisons 4 and 5 allowed including the REJECT procedure next to all other procedures for removing the eye movement artifacts.

Statistical Analysis All statistical comparisons for the ERPs were made using statistical parametrical mapping (SPM), which is widely used in fMRI research. The benefit of this procedure compared to traditional statistics used in ERP analysis is that it allows for a simultaneous statistical comparison across all time points included in the EEG epochs and across all channels included in the EEG recording. This avoids the subjective selection of time points and electrodes for statistical analysis. Instead, the SPM analysis objectively indicates where and when significant effects occur. SPM is a mass univariate approach in which spatiotemporal neuroimaging data are modeled within the statistical framework of the general linear model. The software package used in this study was SPM5 (Wellcome Department of Cognitive Neurology, University College, London, UK). EEG epochs were entered into the analysis as space–time volumes with the anterior–posterior and left–right dimensions of each electrode’s position as a two-dimensional spatial array. Time was entered as a third dimension. This implied that each data point (i.e., the EEG signal measured at a particular millisecond at a particular electrode) was entered into the analysis while preserving its spatiotemporal relationship to other data points. Taking into account the three-dimensional structure of the data justifies calling these data points ‘‘voxels.’’ As such, voxels were defined containing the amplitude information at one electrode (with X and Y dimensions) at 1 ms (Z dimension).

1146 Because the ERP data were low-pass filtered during preprocessing, no additional filtering was performed in the SPM analyses to control for possible serial autocorrelations in the data. Likewise, no global scaling was performed, because baseline corrections were applied in the preprocessing of the ERP epochs. A fixed effects design was used in which conditions were modeled separately for each subject. Sessions per subject, on the other hand, were not modeled separately. For the least versus most contaminated comparison (#2) a separate analysis was done for each removal method (RAW, GC, 10sICA, eyeICA, cICA). Statistical differences between the least and most contaminated trials were assessed with an F contrast. For the comparison between the least contaminated trials without correction for eye movement artifacts and the least contaminated trials after removal of the eye artifacts (Comparison 3) all removal methods were modeled in the same SPM analysis. Four F contrasts were defined to assess the difference between the least contaminated uncorrected and least contaminated trials corrected according to method X (i.e., GC, 10sICA, eyeICA, or cICA). Finally, the cued-attention paradigm valid versus invalid comparison (#5) was assessed using a t contrast for validoinvalid in a separate fixed effects SPM analysis for each removal method (RAW, REJECT, GC, 10sICA, eyeICA, cICA). Differences between conditions were evaluated at a significant alpha level of .05, corrected for multiple comparisons based on the random field theory (Friston, Firth, Liddle, & Frackowiak, 1991; Worsley, Evans, Marrett, & Neelin, 1992). Because we expected no differences in the comparisons involving the contaminated conditions, the significance level was in a second step lowered to po.001 uncorrected for multiple comparisons to show that even at this liberal threshold no significant voxels were found. Clusters should contain at least 20 voxels to be regarded as significant. Results The amount of eye movements registered in each subject during the total duration of the cued attention paradigm ranged from 59 to 229 (mean 5 137). Only 10% of all eye movements were horizontal. The number of trials categorized as least contaminated ranged between 14 and 45 (mean 5 30), compared to 19 and 69 (mean 5 35) for the most contaminated trials.

M. Mennes et al. Grand Average of Eye Movements Figure 2a shows the ERP of the blinks using the eye movements as marked by our algorithm (see above). It is clear that the different removal procedures had a different impact on the blinkrelated ERP measured at the frontal channels. The smaller blinkrelated ERP measured for the eyeICA and cICA methods suggests better correction compared to the GC and 10sICA methods. The impact of correction on horizontal eye movements was less clear (Figure 2b). In addition, the impact of horizontal eye movements on the EEG channel was limited, as evident from the RAW horizontal eye movement ERP, especially when compared to the impact of blinks on the frontal channels. Least Contaminated versus Most Contaminated Trials after Eye Artifact Removal ERPs synchronized on the presentation of validly cued trials and subsequently categorized as least and most contaminated by eye movements are presented in Figure 3, before (Figure 3a,b) and after removal of the eye movement artifacts with the various procedures (Figure 3c–f). It is evident from Figure 3 that obvious differences that existed at the frontal electrodes before removal of the eye movement artifacts between the trials categorized as least and most contaminated have disappeared after applying the removal methods. The average ERP waveforms of the least and most contaminated trials groups have become almost identical. This is true for each of the removal methods and is confirmed by the SPM analyses. No significant differences were found when comparing the least with the most contaminated trial groups after removing the eye movement artifacts. Even at the uncorrected, voxel-wise significance level of po.001, this comparison yielded no significant differences for any of the removal procedures. Least Contaminated RAW Trials versus Least Contaminated Trials after Eye Artifact Removal The SPM analysis and subsequent F contrasts comparing the least contaminated valid trials before removal of the eye movement artifacts to the least contaminated valid trials after removal of the eye movement artifacts yielded no significant differences for any of the removal methods. This was also true when we lowered the threshold for significance to uncorrected po.001. As shown in Figure 4, none of the removal methods induced con-

Figure 2. (a) Grand-average waveforms of the blink artifacts at electrodes Fp (calculated as the mean of Fp1 and Fp2) and VEOG. The signal at VEOG was multiplied by 0.2 to improve comparison with the blink ERP as observed on Fp in the continuous data after removal of the eye artifacts with the different procedures. The RAW signal was multiplied by 0.2. (b) Grand-average waveforms of horizontal eye movement artifacts at electrodes F7/8 and HEOG. HEOG shows the mean of left and right (multiplied by !1) horizontal eye movements at the HEOG channel multiplied by 0.66. All other signals show the mean signal across electrodes F7 and F8 (multiplied by !1).

Validation of ICA to remove eye movement artifacts

1147

Figure 3. Grand-average ERP waveforms at electrode Fp (calculated as the mean of Fp1 and Fp2) for the valid trials of the cued attention paradigm classified as least (black line) or most (gray line) contaminated by eye movement artifacts. (a) Uncorrected signal measured at the VEOG channel. (b) Trials not corrected for eye movement artifacts (RAW). (c–f ) Trials after the removal of the eye movement artifacts with the respective procedures (GC, 10sICA, eyeICA, cICA).

siderable changes in epochs where no eye movement artifacts were detected. Standard Deviation A repeated measures analysis of variance (ANOVA) with removal method (RAW, REJECT, GC, 10sICA, eyeICA and cICA) as the within-participants factor on the standard deviation measured across trials and time points in the 1000-ms interval after stimulus presentation yielded a significant effect of methods (Greenhouse–Geisser corrected po.005, e 5 .29) at Fp (Figure 5). Post hoc comparisons using the Tukey procedure indicated

Figure 4. Grand-average waveforms at electrode Fp (calculated as the mean of Fp1 and Fp2) for the valid trials of the cued attention paradigm classified as least contaminated by eye movements. ERPs are shown uncorrected for eye movement artifacts (RAW) and corrected for these artifacts according to the different removal procedures (GC, 10sICA, eyeICA, cICA).

that there was no difference in standard deviation between the REJECT, GC, 10sICA, eyeICA, and cICA methods, and these methods all had a smaller standard deviation compared to the RAW data (po.001 for each comparison). This effect gradually decreased toward the posterior electrodes (see Figure 5; Fz: po.003; Cz: p 5 .07; Pz: p 5 .2). This result indicates that all eye artifact removal methods equivalently reduced the variance associated with eye movement artifacts.

Figure 5. Mean standard deviation across participants for all time points in the 0–1000-ms interval following stimulus presentation and all validly cued trials of the cued attention paradigm. Values are shown for electrodes Fp (calculated as the mean of Fp1 and Fp2), Fz, Cz, and Pz. At electrode Fp, RAW was significantly higher than each of the other procedures (po.001 for each procedure). This effect gradually decreased toward the posterior electrodes (Fz: po.003; Cz: p 5 .07; Pz: p 5 .2).

1148

M. Mennes et al.

Figure 6. Grand-average ERP waveforms at electrode Fp (calculated as the mean of Fp1 and Fp2) of the valid (black lines) and invalid (gray lines) trials of the cued attention paradigm. ERPs for the six different removal procedures are shown. (a) RAW (lines) and the uncorrected signal as measured at the VEOG channel (multiplied by 0.3; dotted lines). (b) REJECT. (c) GC. (d) 10sICA. (e) eyeICA. (f) cICA method. Gray bars in the c, e, and f graphs indicate significant intervals as found in the SPM validoinvalid t contrast. Lighter colors indicate more significant t values at po.05, corrected for multiple comparisons.

Valid versus Invalid Cue Trials Figure 6 shows the grand average ERP waveforms for the valid and invalid trials of the cued attention paradigm at electrode Fp (calculated as the mean of Fp1 and Fp2) for the different artifact removal methods. The largest condition-related difference could be observed in the P300 peak, around 300 ms after stimulus onset. Invalid trials were associated with a larger P300 peak amplitude compared to the valid trials, on both the frontal and posterior electrodes. This effect was already evident in the RAW trials. However, in these trials, the P300 effect at the frontal electrodes was partially obscured by the eye movement artifact that could be observed in the ERP from the P300 onward (see Figure 6a). Accordingly, before the removal of the eye movement artifacts, the SPM analysis revealed significant differences between valid and invalid trials for the P300 peak but only at posterior electrodes (not shown in Figure 6). Over the posterior scalp, an extensive significant difference was observed, with a local maximum at electrode P3 (t 5 5.57, corrected po.001). Over the frontal scalp, a small activation area comprising only electrode F3 was found (t 5 4.49, corrected po.001). Figure 6b–f indicates that the different rejection and removal methods each seemed to adequately eliminate the eye movement artifacts from the ERP. However, the effects of the artifact removal for the statistical analysis differed as a function of the removal method. Table 2 gives an overview of the number of significant voxels and clusters in each analysis. Remarkably, the numbers in Table 2 suggest that rejecting trials based on extreme values yielded less significant differences compared to using the RAW data. The same local maxima as for the RAW data were found, but the number of significant voxels was greatly reduced (see Table 2). Here, the gain in statistical power because of the

smaller standard deviation is undone by the loss of power because of the smaller number of trials. The other four methods (GC, 10sICA, eyeICA, and cICA) all yielded more significant voxels compared to the analysis for the raw data, immediately showing the usefulness of these methods. Within these four methods the analysis for the 10sICA method yielded the least significant voxels, whereas the number of significant voxels for the GC, eyeICA, and cICA was similar.

Discussion The results presented here indicate that removal of eye movement artifacts by ICA-based methods achieved similar performance compared to the more frequently used Gratton and Coles method. Removing the eye movement artifacts by rejecting trials based on extreme values was the least favorable option. Except Table 2. Results of the SPM Analyses Comparing the Valid and Invalid Cue Trials of the Cued Attention Paradigm REJECT RAW 10sICA eyeICA N significant voxels Nclusters Clusterso20 voxels N local maxima Highest t value Critical t value

128 2 1 2 4.84 3.83

GC

cICA

342 434 736 844 945 1 2 1 1 2 0 1 0 0 1 2 4 3 3 4 5.57 5.47 5.72 6.08 6.00 3.86 3.87 3.84 3.83 3.83

Note: For each removal method a validoinvalid t contrast was defined. Results were assessed at a po.05 alpha level, corrected for multiple comparisons.

Validation of ICA to remove eye movement artifacts for the REJECT method, all eye movement artifact removal methods improved statistical sensitivity at the frontal channels without affecting data at channels that were not as strongly influenced by the eye movements. Selecting only eye movement epochs for ICA compilation facilitated the component identification process, thus limiting subjective user input to a minimum. In the current study, rejecting epochs based on extreme values as a means of removing eye movement artifacts seemed worse than applying no correction at all. The loss of statistical power due to the reduction in the number of trials outweighed the gain in power achieved by reducing the standard deviation and easily outweighed the ease of use of this technique. Here the rejectionbased method for removing the eye movement artifacts resulted in a loss of 30% of trials. It should be noted that the participants in the current study were adults instructed to maintain central fixation and to blink at a minimum. It is likely that testing clinical populations or children would result in even higher numbers of lost trials. In contrast, if only few eye movements are present in a data set, removing epochs containing an eye movement artifact might be the method of choice, thus excluding ‘‘data manipulation’’ as present in the subtraction-based artifact removal methods. It should be noted that other factors, such as the specific task used in an experiment, could also influence the frequency of eye movements. This should be taken into account when opting for the REJECT method. Selecting only eye movement epochs for ICA compilation facilitated the component identification process. The selection of data to use for ICA is important, as the algorithm will calculate sources from this data. The better a source is represented in the raw data, the higher the likelihood that its activity will be captured in one single component and the easier it becomes to identify this source or component as being of interest. Here we compared three data selections for ICA. First we performed ICA on only 10 s of EEG recording, as suggested by others (Jung et al., 1998a, 1998b, 2000a). This eliminates the need to compile large amounts of EEG data with ICA, which is time-consuming. The current results suggest that, although the 10sICA method is better than using raw data, it seems not fully satisfactory. Although this method provided the fastest computation times, the selection of components to remove was less straightforward, with components apparently representing eye movement artifacts not reaching our cutoff of 15% explained variance on the EOG channels. In contrast, component selection was more straightforward when performing ICA on the complete EEG recording (cICA), indicating a good representation of eye movement artifact in the data entered. The drawback to this procedure is that it may take a considerable amount of time for the ICA to compile, especially with large data sets. This drawback largely disappears when performing ICA only on epochs containing the eye movement artifacts (eyeICA). Indeed, fewer data will need to be compiled while the straightforward identification of components is retained. Our current results indicate that the performances of the cICA and eyeICA were almost similar. The main difference was that the eyeICA yielded a smaller number of significant ‘‘voxels’’ in the task-relevant ERP comparison. In light of these findings, eyeICA provided the best opportunity for making eye movement artifact removal based on ICA more straightforward. Quantitative identification of components to remove provides an advantage over user-driven, subjective selection of components (Joyce et al., 2004; Romero, Mananas, & Barbanoj, 2008). It increases objectivity of data preprocessing,

1149 making results more comparable among research groups. Of course, evaluation of the components selected by the algorithm is warranted at all times. In the current study, selecting data for ICA computation based on the amount of explained variance in the EOG channels facilitated the identification of components representing artifacts. Further research is needed to assess whether the selection criteria used in the current study are usable across studies. Here, we applied ICA at an early stage in data processing and subsequently ‘‘filtered’’ the data by subtracting components representing eye movement artifacts, similar to the GC method. Another approach would be to calculate ERPs of interest without first correcting the data for eye movement artifacts and subsequently applying ICA to remove components in the ERP waveforms that are identified as eye movement artifacts. Further research is needed to assess the difference in results between these two approaches, as it might be more difficult to disentangle eye movement artifacts from cognition in ERPs that primarily represent cognition as compared to a complete EEG data set that contains all sorts of variance (Ille et al., 2002). A second consideration entails the number of components that are used by ICA and the GC method to describe the eye movement artifacts. The GC method first estimates a correction factor for vertical eye movements, which is dominated by blinks. Subsequently, after subtraction of this ‘‘blink factor,’’ a second factor is calculated for the remaining saccades (Gratton et al., 1983). ICA is unconstrained in the number of components identified as eye movement artifacts. In general and evident in the example described in the Methods section, ICA will identify separate components for vertical and horizontal eye movement artifacts. In our example, a second component representing a part of the vertical eye movement artifacts was selected. Together with the first vertical component they possibly account for more of the eye movement artifact-related variance than the vertical eye movement artifact component resulting from the GC method. However, the component solution provided by ICA entirely depends on the data, possibly resulting in a different number of eye artifact components for different participants. The number of components to remove from the data is an important issue when using blind source separation methods such as ICA (Ille et al., 2002; Jung et al., 2000a). Special caution is needed when interpreting the raw eye movement artifact average (Figure 2). Intuitively one would interpret a smaller eye movement-related ERP with better correction of the eye movement artifact, implying a flat ERP for optimal correction. However, this disregards that eye movements could interfere with perception and are related to cognition (e.g., fewer blinks in a difficult task), which might cause the brain to produce an eye movement-related ERP (Berg & Davies, 1988). More research is needed to fully address the effect of eye movements on cognition and to assess whether the current methods and ICA in particular are sufficiently successful in separating eye movements artifacts from cognition associated with eye movements. The number of participants and trials in the current study were limited. However removal of eye movement artifacts should be completed for each subject separately because artifacts will differ among subjects (Iwasaki et al., 2005). Second, it remains difficult to evaluate the different eye artifact removal methods because we do not know what a corrected waveform should look like (Croft, Chandler, Barry, Cooper, & Clarke, 2005). Here, we tried to counteract this limitation by evaluating the removal

1150

M. Mennes et al.

methods on different aspects of the data. Across these evaluations we showed that ICA is a valid technique for the removal of eye movement artifacts. Current software allows for an easy implementation of blind source separation methods, which can also

achieve good results in the absence of EOG channels (Romero et al., 2008). Selecting only eye movement epochs for ICA compilation facilitated identification of components representing eye movement artifacts.

REFERENCES Bell, A. J., & Sejnowski, T. J. (1995). An information-maximization approach to blind separation and blind deconvolution. Neural Computation, 7, 1129–1159. Berg, P., & Davies, M. B. (1988). Eyeblink-related potentials. Electroencephalography and Clinical Neurophysiology, 69, 1–5. Berg, P., & Scherg, M. (1994). A multiple source approach to the correction of eye artifacts. Electroencephalography and Clinical Neurophysiology, 90, 229–241. Corbetta, M., Kincade, J. M., & Shulman, G. L. (2002). Neural systems for visual orienting and their relationships to spatial working memory. Journal of Cognitive Neuroscience, 14, 508–523. Croft, R. J., & Barry, R. J. (2000). EOG correction of blinks with saccade coefficients: A test and revision of the aligned-artifact average solution. Clinical Neurophysiology, 3, 444–455. Croft, R. J., Chandler, J. S., Barry, R. J., Cooper, N. R., & Clarke, A. R. (2005). EOG correction: A comparison of four methods. Psychophysiology, 42, 16–24. Debener, S., Hine, J., Bleeck, S., & Eyles, J. (2008). Source localization of auditory evoked potentials after cochlear implantation. Psychophysiology, 45, 20–24. Delorme, A., & Makeig, S. (2004). EEGLAB: An open source toolbox for analysis of single-trial EEG dynamics. Journal of Neuroscience Methods, 134, 9–21. Esposito, F., Seifritz, E., Formisano, E., Morrone, R., Scarabino, T., Tedeschi, G., et al. (2003). Real-time independent component analysis of fMRI time-series. NeuroImage, 20, 2209–2224. Flexer, A., Bauer, H., Pripfl, J., & Dorffner, G. (2005). Using ICA for removal of ocular artifacts in EEG recorded from blind participants. Neural Networks, 18, 998–1005. Ford, M. R., Sands, S., & Lew, H. L. (2004). Overview of artifact reduction and removal in evoked potential and event-related potential recordings. Physical Medicine and Rehabilitation Clinics of North America, 15, 1–17. Frank, R. M., & Frishkoff, G. A. (2007). Automated protocol for evaluation of electromagnetic component separation (APECS): Application of a framework for evaluating statistical methods of blink extraction from multichannel EEG. Clinical Neurophysiology, 118, 80–97. Friston, K. J., Frith, C. D., Liddle, P. F., & Frackowiak, R. S. J. (1991). Comparing functional (PET) images: The assessment of significant change. Journal of Cerebal Blood Flow and Metabolsim, 11, 690–699. Gratton, G., Coles, M. G. H., & Donchin, E. (1983). A new method for off-line removal of ocular artifact. Electroencephalography and Clinical Neurophysiology, 55, 468–484. Huang, R. S., Jung, T.-P., Delorme, A., & Makeig, S. (2008). Tonic and phasic electroencephalographic dynamics during continuous compensatory tracking. NeuroImage, 39, 1896–1909. Iidaka, T., Matsumoto, A., Nogawa, J., Yamamoto, Y., & Sadato, N. (2006). Frontoparietal network involved in successful retrieval from episodic memory. Spatial and temporal analyses using fMRI and ERP. Cerebal Cortex, 16, 1349–1360. Ille, N., Berg, P., & Scherg, M. (2002). Artifact correction of the ongoing EEG using spatial filters based on artifact and brain signal topographies. Journal of Clinical Neurophysiology, 19, 113–124. Iwasaki, M., Kellinghaus, C., Alexopoulos, A. V., Burgess, R. C., Kumar, A. N., Han, Y. H., et al. (2005). Effects of eyelid closure, blinks, and eye movements on the electroencephalogram. Clinical Neurophysiology, 116, 878–885.

Joyce, C. A., Gorodnitsky, I. F., & Kutas, M. (2004). Automatic removal of eye movement and blink artifacts from EEG data using blind component separation. Psychophysiology, 41, 313–325. Jung, T.-P., Humphries, C., Lee, T.-W., Makeig, S., McKeown, M. J., Iragui, V., et al. (1998a). Extended ICA removes artifacts from electroencephalographic data. Advances in Neural Information Processing Systems, 10, 894–900. Jung, T.-P., Humphries, C., Lee, T.-W., Makeig, S., McKeown, M. J., Iragui, V., et al. (1998b). Removing electroencephalographic artifacts: Comparison between ICA and PCA. Neural Networks Signal Processing, VIII, 63–72. Jung, T.-P., Makeig, S., Humphries, C., Lee, T.-W., McKeown, M. J., Iragui, V., et al. (2000a). Removing electroencephalographic artifacts by blind source separation. Psychophysiology, 37, 163–178. Jung, T.-P., Makeig, S., Westerfield, M., Townsend, J., Courchesne, E., & Sejnowski, T. J. (2000b). Removal of eye activity artifacts from visual event-related potentials in normal and clinical participants. Clinical Neurophysiology, 111, 1745–1758. Jung, T.-P., Makeig, S., Westerfield, W., Townsend, J., Courchesne, E., & Sejnowski, T. J. (2001). Analysis and visualization of single-trial event-related potentials. Human Brain Mapping, 14, 166–185. Kansaku, K., Muraki, S., Umeyama, S., Nishimori, Y., Kochiyama, T., Yamane, S., et al. (2005). Cortical activity in multiple motor areas during sequential finger movements: An application of independent component analysis. NeuroImage, 28, 669–681. Makeig, S., Delorme, A., Westerfield, M., Jung, T.-P., Townsend, J., Courchesne, E., et al. (2004). Electroencephalographic brain dynamics following manually responded visual targets. PLOS Biology, 2, 747–762. Posner, M. I. (1980). Orienting of attention. Quarterly Journal of Experimental Psychology, 32, 3–25. Romero, S., Mananas, M. A., & Barbanoj, M. J. (2008). A comparative study of automatic techniques for ocular artifact reduction in spontaneous EEG signals based on clinical target variables: A simulation case. Computers in Biology and Medicine, 38, 348–360. Slagter, H. A., Lutz, A., Greischar, L. L., Francis, A. D., Nieuwenhuis, S., Davis, J. M., et al. (2007). Mental training affects distribution of limited brain resources. PLOS Biology, 5(6), e138. Stone, J. V. (2002). Independent component analysis: An introduction. Trends in Cognitive Sciences, 6, 59–64. Verleger, R., Gasser, T., & Mo¨cks, J. (1982). Correction of EOG artifacts in event-related potentials of the EEG: Aspects of reliability and validity. Psychophysiology, 19, 472–480. Vigario, R., Sarela, J., Jousmaki, V., Hamalainen, M., & Oja, E. (2000). Independent component approach to the analysis of EEG and MEG recordings. IEEE Transactions on Biomedical Engineering, 47, 589– 593. Vigario, R. N. (1997). Extraction of ocular artefacts from EEG using independent component analysis. Electroencephalography and Clinical Neurophysiology, 103, 395–404. Worsley, K. J., Evans, A. C., Marrett, S., & Neelin, P. (1992). A threedimensional statistical analysis for rCBF activation studies in human brain. Journal of Cerebral Blood Flow and Metabolism, 12, 900–918. Zeki, S., Perry, R. J., & Bartels, A. (2003). The processing of kinetic contours in the brain. Cerebal Cortex, 13, 189–202. (Received October 9, 2008; Accepted November 12, 2009)

Psychophysiology, 47 (2010), 1151–1158. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01019.x

Never mind the spider: Late positive potentials to phobic threat at fixation are unaffected by perceptual load

JOAKIM NORBERG,a NATHALIE PEIRA,a,b and STEFAN WIENSa,b a

Department of Psychology, Stockholm University, Stockholm, Sweden Stockholm Brain Institute, Stockholm, Sweden

b

Abstract Research suggests that processing of emotional stimuli may be eliminated if a concurrent task places sufficient demands on attentional resources. To test whether this holds for stimuli with strong emotional significance, pictures of spiders as well as mushrooms were presented at fixation to spider-fearful and non-fearful participants. Concurrently, perceptual load was manipulated in two levels with a peripheral letter discrimination task. Results of event-related potentials showed that, compared with non-fearful participants, spider-fearful participants showed greater late positive potentials (LPP) to spiders than mushrooms, which provides a manipulation check that spiders were emotionally meaningful to spider-fearful participants. Critically, this effect was not affected by level of perceptual load. These findings suggest that strong emotional stimuli at fixation may resist manipulations of perceptual load. Descriptors: Emotion, Attention, EEG/ERP

ERP response was eliminated (Holmes, Vuilleumier, & Eimer, 2003). Similar effects were obtained for emotional expressions other than fear. That is, the ERP effects to attended emotional expressions in the periphery were eliminated when participants performed a line discrimination task at fixation (Eimer, Holmes, & McGlone, 2003). Last, when faces were shown at fixation rather than in the periphery, an enhanced positivity after 220 ms to attended fearful than neutral faces was eliminated when faces were unattended (Holmes, Kiss, & Eimer, 2006). Notably, in these ERP studies, a stronger positivity after about 250 ms over central electrodes was elicited by attended emotional faces than attended neutral faces (Eimer & Holmes, 2007). This ERP response was interpreted as a late positive potential (LPP). The LPP refers to a greater central-parietal positivity to emotional than neutral pictures and is commonly obtained for both positive and negative emotional pictures such as erotic scenes and mutilations (Cuthbert, Schupp, Bradley, Birbaumer, & Lang, 2000; Olofsson, Nordin, Sequeira, & Polich, 2008). It has been identified as an index of subjective significance and denotes allocation of attentional resources to motivationally relevant stimuli (Bradley, 2009). Because in the ERP studies with emotional faces the LPP to emotional versus neutral faces was eliminated when faces were unattended (Eimer & Holmes, 2007), these findings suggest that processing of task-irrelevant emotional stimuli is reduced if attentional resources are depleted. The role of attentional resources on processing of task-irrelevant stimuli has been investigated by Lavie and colleagues in a number of studies, resulting in a proposal called Load theory (for reviews, see Lavie, 1995, 2005). This theory states that attentional capacity is limited and is affected by the number of perceptual processes and the number of items that need to be

People need to process objects and events in the environment that are relevant to their goals (Bradley, 2009). Because emotional stimuli are highly relevant to people, they may be processed irrespective of whether or not they are attended (for review, see Vuilleumier, 2005). For example, when participants were shown picture pairs of houses, fearful faces, and neutral faces in horizontal and vertical orientations, results showed similar amygdala activations to fearful faces irrespective of whether or not they were attended (Vuilleumier, Armony, Driver, & Dolan, 2001). However, results from other studies have challenged the idea that emotional stimuli are processed independent of attention (for review, see Pessoa, 2005). For example, when pictures of emotional faces at fixation were flanked by bars on both sides, and participants judged whether the bars were of the same orientation, results showed amygdala activation to fearful faces at fixation only when the bar task was simple but not when it was difficult (Pessoa, Padmala, & Morland, 2005). Also, an eventrelated potential (ERP) study used the same task as that in the original fMRI study by Vuilleumier et al. (2001) and thus showed pairs of fearful faces, neutral faces, and houses in the periphery. When faces were attended, ERP findings showed a greater frontal positivity from 100 ms onwards to pairs of fearful than pairs of neutral faces. In contrast, when houses were attended, this

This research was financed in part by a grant of the Swedish Research Council (Vetenskapsra˚det) to Stefan Wiens. We thank Jonas Olofsson and Karsten Hoechstetter for helpful discussions. Address reprint requests to: Stefan Wiens, Frescati Hagva¨g 14, Psykologiska institutionen, Stockholms universitet, 106 91 Stockholm, Sweden. E-mail: Stefan Wiens at [email protected] or Joakim Norberg at [email protected] 1151

1152 processed, that is, the level of perceptual load. Under low perceptual load, processing of task-relevant information does not consume all attentional capacity. As a consequence, there is residual attentional capacity to process task-irrelevant information. In contrast, under high perceptual load, processing of taskrelevant information may consume all attentional capacity, thus leaving little or no residual attentional capacity to process taskirrelevant information. There is extensive evidence supporting the Load theory in that perceptual load attenuates processing of task-irrelevant information when shown both at fixation and in the periphery (Lavie, 2005). Load theory predicts that, if perceptual load is sufficiently high, all processing of task-irrelevant information will seize. In support, research on processing of emotional faces suggests that task-irrelevant emotional faces are not processed under high perceptual load (Pessoa, 2005). However, because emotional faces produce only weak emotional reactions (Whalen, 1998), it is unresolved whether responses can be eliminated to strong emotional stimuli as easily as responses to emotional faces. Research indicates that spider pictures produce strong emotional reactions in spider-fearful individuals. That is, spider-fearful individuals respond with flight-orfight reactions in the form of increased heart rate and blood pressure and greater amygdala activity to pictures of feared objects (spiders or snakes) than to fear-irrelevant pictures (A˚hs, Pissiota, Michelga˚rd, Frans, Furmark, et al., 2009; Globisch, Hamm, Esteves, & O¨hman, 1999; Wendt, Lotze, Weike, Hosten, & Hamm, 2008). To study strong emotional stimuli, we showed pictures of spiders to people who were afraid of spiders, with pictures of mushrooms as control pictures and people who were unafraid of spiders as the control group. To index emotional responses, we measured ERPs with a focus on the LPP because it has been identified as an index of motivated attention (Bradley, 2009). In support, spider-fearful individuals show elevated LPP amplitudes to photographs of spiders and schematic spiders (Flykt & Caldara, 2006; Kolassa, Musial, Mohr, Trippe, & Miltner, 2005; Kolassa, Musial, Kolassa, & Miltner, 2006; Miltner, Trippe, Krieschel, Gutberlet, Hecht, & Weiss, 2005; Trippe, Hewig, Heydel, Hecht, & Miltner, 2007). Because the LPP to emotional pictures correlates with enhanced activation in lateral occipital, inferotemporal, and parietal cortex (Sabatinelli, Lang, Keil, & Bradley, 2007), these findings support the idea that the LPP to spiders in spider-fearful participants reflects motivated attention and increased visual processing. The present study avoided two common methodological problems in manipulating perceptual load. First, in many previous studies, the task-irrelevant emotional pictures were shown in the periphery. But, task-irrelevant pictures in the periphery may not be processed under high perceptual load because they are task irrelevant, not spatially attended, or both (Okon-Singer, Tzelgov, & Henik, 2007). To isolate effects of spatially attended but task-irrelevant emotional pictures under perceptual load, task-irrelevant spiders and mushrooms were shown at fixation while participants performed a letter discrimination task on either three or six letters on an imaginary circle surrounding the pictures. The letter discrimination task with six letters matched high perceptual load in other studies (e.g., Bishop, Jenkins, & Lawrence, 2007; Okon-Singer et al., 2007). Second, in many previous studies, task relevance was manipulated over blocks so that the emotional pictures were task relevant during some parts of the experiment. For example, participants may rate in different blocks either the gender of the faces or their emotional expression. Unfortunately, this procedure may bias participants to

J. Norberg et al. attend to the emotional aspects of the pictures even when they are task irrelevant. If so, this procedure has the negative consequence that it actually reduces and, thus, underestimates effects of perceptual load on emotional processing. To avoid this potential confound, the emotional pictures in the present study were always task irrelevant. In sum, the goal of this study was to investigate whether processing of strong emotional but task-irrelevant pictures at fixation is affected by perceptual load. Pictures of spiders and mushrooms were presented to spider-fearful and non-fearful participants while they performed a letter discrimination task with two levels of perceptual load. Greater LPP amplitudes to spiders than mushrooms for spider-fearful than non-fearful participants would indicate that the manipulation of emotion was successful. In analysis of variance (ANOVA) terms, this manipulation check corresponds to a two-way interaction between fear group and picture type. Critically, Load theory (Lavie, 2005) would further predict that this two-way interaction is moderated by perceptual load: Greater LPP amplitudes to spiders versus mushrooms for the spider-fearful group than the non-fearful group would be apparent under low perceptual load, but this effect would be reduced, if not eliminated, under high perceptual load. In ANOVA terms, Load theory would predict a three-way interaction between fear group, picture type, and load. In contrast, if perceptual load does not moderate strong emotional reactions, the greater LPP amplitudes to spiders versus mushrooms for the spider-fearful than non-fearful group would be similar during high and low perceptual load. Therefore, an absence of evidence for a three-way interaction (i.e., no significance and tiny effect size) would provide a limit for Load theory in that strong emotional stimuli at fixation may resist manipulations of perceptual load. Method Participants All undergraduate psychology students at Stockholm University, Sweden, were contacted via e-mail and invited to participate in this study. The e-mail contained a link to an internet site that allowed students to read about the study (i.e., spider pictures and EEG). A total of 185 students filled in electronic questionnaires about their spider fear as well as other variables. Spider fear as well as snake fear was assessed with modified versions of the German Spider Anxiety Screening (Rinck, Bundschuh, Engler, Muller, Wissmann, et al., 2002), which consists of four items with possible responses between 0 and 6 and, thus, a maximum sum of 24 points. Participants with the highest scores (418) and the lowest scores (o3) on spider fear were recruited for the present study (n 5 17 in each group). Their participation was based on informed consent, and the study was approved by the regional ethics review board. After the experiment, participants received either class credits or two movie vouchers. In each group, two participants had to be excluded because of excessive electroencephalogram (EEG) artifacts (see below). Table 1 shows the demographics for the final sample (n 5 15 in each group) as well as results for other questionnaires that participants filled in online or in the lab. Participants were right-handed (all except one) and had normal or corrected-to-normal vision. Apparatus The experiment was conducted in a darkened room. Pictures were shown at a distance of about 80 cm on a 21-inch View Sonic

Never mind the spider

1153

Table 1. The Means (SD) of Demographics, Questionnaire Data, and Task Ratings for Spider-fearful (n 5 15) and Non-fearful (n 5 15) Participants Variable Demographics Age (years) Genderb Questionnaires Spider fearcn Affect (PANAS)d Negative before task Negative after task Fear ! time (after-before) Positive before task Positive after task Fear ! time (after-before) Snake fearcn Disgust sensitivityen Trait anxietyfn Task ratings Difficulty with letter detection Difficulty ignoring picturesn Difficulty ignoring spidersn Difficulty ignoring mushrooms Fear ! difficulty (spi-mush)n Feelings toward spidersn Feelings toward mushrooms Fear ! feelings (spi-mush)n

Spider fear

No fear

t

ra

24.1 (3.8) 13 f/2 m

27.3 (7.1) 10 f/5 m

1.56

" .28

22.1 (1.8)

0.4 (0.7)

43.13

.99

.001n

10.5 (2.3) 11.4 (5.2) 0.9 (5.1) 25.4 (3.7) 20.4 (5.8) " 5.0 (3.4) 10.2 (5.1) 18.2 (3.9) 47.0 (9.7)

9.2 (1.4) 9.6 (2.1) 0.4 (1.7) 25.5 (2.8) 21.9 (3.6) " 3.7 (3.2) 4.9 (7.0) 12.0 (3.9) 35.3 (7.8)

1.85 1.18 0.36 0.11 0.85 1.04 2.35 4.41 3.62

.33 .22 .07 " .02 " .16 " .20 .41 .64 .57

.077 .24 .73 .91 .41 .31 .026n .001n .001n

6.9 (1.5) 6.3 (3.0) 7.0 (3.3) 4.3 (2.7) 2.7 (2.5) " 2.0 (1.7) 0.2 (1.0) " 2.2 (2.3)

6.7 (1.9) 3.3 (3.1) 3.4 (2.9) 2.3 (2.5) 1.1 (1.8) 0.1 (1.5) 0.3 (1.1) " 0.1 (1.1)

0.42 2.67 3.20 2.03 2.10 3.60 0.17 3.16

.08 .45 .52 .36 .37 " .56 " .03 " .51

.68 .013n .003n .052 .045n .001n .87 .004n

p .13 .20

a

In the correlation analyses, spider fear was coded as 1 and no fear as 0. The ratios of female (f) to male (m) participants in the two groups were compared in a w2 analysis, which yielded a value of 1.68, p4.20. c Spider and snake fear scores ranged between 0 and 24 (Rinck et al., 2002). d The positive and negative affect scales (PANAS) comprised eight items each (Watson et al., 1988). e Disgust sensitivity questionnaire (Haidt et al., 1994). f STAI-trait anxiety inventory with 20 items (Spielberger, 1983). n Significant group differences (p 5 .05). b

P227f cathode ray-tube monitor at a 100-Hz refresh rate with a resolution of 1280 ! 1024 pixels. Experiment software was Presentation 10.3 (Neurobehavioral Systems, Inc., Albany, CA). Stimuli were pictures of 30 spiders and 30 mushrooms in 8-bit gray scale and without backgrounds. These 60 pictures were taken from a previous study on the relationship between fear of spiders and snakes and recognition of masked spiders, mushrooms, snakes, and flowers (Wiens, Peira, Golkar, & O¨hman, 2008). As reported in that study, pictures were matched in lowlevel features (number of pixels, contrast, luminance), and for the present study, their screen size was adjusted to fit in a window of 5 ! 3.6 cm (3.6 ! 2.6 degrees, width ! height). During the experiment, screen background was black. Letters in the letter discrimination tasks were shown in gray Arial font. Uppercase letters (e.g., W, K) had a maximum size of 2 ! 1.5 cm (1.4 ! 1.1 degrees) and the lowercase letter o was 0.5 ! 0.5 cm (0.4 ! 0.4 degrees). The letters were shown on an imaginary circle centered on the middle of the screen at a radius of 4.6 cm (6.6 degrees for the whole circle). The EEG was recorded with an Active Two Biosemi system (BioSemi, Amsterdam, The Netherlands) at 512 Hz with a 104Hz high cutoff filter. The electrode cap allowed recordings from 64 active electrodes according to the 10/20 system and two additional electrodes to obtain the built-in ground for the Biosemi system (i.e., common mode sense and driven right leg). Procedure Participants performed a speeded letter discrimination task in which they had to push the space key on a keyboard whenever

they detected the letter N or X on the computer screen. Each trial began with a 1-s fixation cross in the center of the screen. Then, six characters were presented on the screen in an imaginary circle at fixation at positions 2, 4, 6, 8, 10, and 12 o’clock. In the high load condition, the six characters were H, K, M, Z, W, and V. In the low load condition, the six characters were three letters drawn from the same set (randomly on each trial). These three letters were placed either at 2, 6, and 10 o’clock or at 4, 8, and 12 o’clock on the imaginary circle. The remaining positions were filled with the lowercase letter o. In both load conditions, the letter circle was shown for 200 ms. Spider and mushroom pictures were shown at fixation concurrently inside the letter circle. On target trials, one of the letters (i.e., H, K, M, Z, W, or V) was replaced by either N or X (50% each across all target trials). The interval between trials was 1.5 s plus a random delay of up to 500 ms. Before the task, participants received instructions that they should respond as quickly as possible without compromising accuracy. If they made a mistake, they should not worry but focus on the next trial. During the task, they should look toward the middle of the screen and minimize eye blinks. Pictures of spiders and mushrooms would be shown at fixation, but because these pictures would be completely irrelevant to the task, participants should ignore them. The experiment consisted of three blocks of 180 trials each. In each block, half of the 180 trials were trials with high perceptual load and half were trials with low perceptual load (trials in random order). Also, for the 90 trials in each load condition, spiders, mushrooms, and blanks (i.e., no picture) were each shown on one-third of the trials. Further, the 30 spiders and 30 mushrooms

1154 were shown once in each load condition. Last, for each load ! event (spider, mushroom, blank) combination, 10 trials were target trials with the letter N or X. So, across the three blocks, there were a total of 30 target trials and 60 no-target trials for each load ! event combination. Thus, participants had to push the space key on one-third of all trials. For each block and participant, trial order was randomized (hence, load was random over trials). Each block lasted about 8 min and participants were allowed short breaks between blocks. To familiarize participants with the actual task, they performed a practice task with 20 trials and pictures that were similar to those used in the actual task. Before and after the task, participants completed short versions of a mood questionnaire. After the task, participants also filled in a questionnaire about several aspects of the task (see Table 1). They rated difficulty in letter detection, difficulty in ignoring pictures in general, and difficulty in ignoring spiders and mushrooms. These questions were rated on a scale from 0 (extremely easy) to 10 (extremely difficult). Participants also rated the feelings that they experienced when a spider or mushroom was shown on the screen. These questions were rated on a scale from –5 (very negative) to 5 (very positive). Data Analysis Signal detection indexes and reaction time. In each of the three blocks, there were 10 target trials and 20 no-target trials for each load ! event (spider, mushroom, and blank) combination. For each load ! event combination and each block, the signal detection indexes d’ (discrimination ability) and C (response bias) were computed based on hits (i.e., space key was pressed on a target trial) and false alarms (i.e., space key was pressed on no-target trials) (for a review of signal detection analyses for these kinds of data, see Wiens et al., 2008). To compute signal detection indexes, hit and false alarm rates need to be greater than 0 and less than 1. Therefore, hit (and false alarm) rates were defined as the number of hits (or false alarms) 10.5, and this sum was divided by the sum of the maximum number of hits (or false alarms) 11. Preliminary analyses showed that block had only a main effect on d’ (i.e., participants performed better over blocks) and did not interact with spider fear. Therefore, data were collapsed across blocks. Similarly, reaction times to hits were computed for each load ! event combination across blocks. EEG. The software BESA (version 5.2.4.48, MEGIS Software GmbH, Gra¨felfing, Germany, www.BESA.de) was used for offline processing. Inspection of the raw data showed that 9 participants (6 fearful participants) had up to three noisy electrodes, which were interpolated with spherical splines. Because eye movements have strong effects on the EEG topography, a built-in algorithm was used to correct the continuous EEG of each participant (Berg & Scherg, 1994; Ille, Berg, & Scherg, 2002). In this algorithm, horizontal and vertical eye movements and blinks for each participant were estimated from all electrodes by simultaneously modeling each type of eye movements with one source (spatial filter) and brain activity with a surrogate brain model of 15 regional sources. Then, sections that fulfilled criteria for artifacts were used to recompute separate spatial filters for horizontal and vertical eye artifacts. Last, eye movement artifacts in the continuous EEG were corrected by estimating the brain signal from the surrogate brain model.

J. Norberg et al. The analysis of EEG data included only non-target trials of spiders and mushrooms in which participants did not push the space key. Both target trials with responses (hits) and non-target trials with responses (false alarms) were excluded for three reasons. First, because EEG is sensitive to movement artifacts, trials with motor behavior decrease the signal-to-noise ratio and, thus, decrease sensitivity in detecting effects (Luck, 2005). Second, because the focus of this study was on EEG rather than behavioral effects, there were many more non-target than target trials (66% vs. 33%). Thus, the smaller proportion of target trials makes it difficult to obtain reliable estimates for target trials and to compare the conditions. Third, because differences in subjective probability have effects on the P3 (Squires, Donchin, Herning, & McCarthy, 1977), which overlaps with the LPP, the target trials and non-target trials with responses are potentially confounded by effects of subjective probability on P3. Because blanks also had a lower probability than pictures (33% vs. 66%), they were excluded from the EEG analysis for the same reason. For each load ! event combination, there was a maximum of 60 non-target trials, but non-target trials in which participants pushed a space incorrectly (i.e., false alarms) were excluded. Across participants, about 3 of 60 trials were excluded for each load ! event combination. For each trial, the continuous EEG after correction for eye movements (see above) was segmented between 500 ms before to 800 ms after the letter circle onset. Data inspection for each participant was conducted without knowledge of group membership (i.e., blind scoring). To detect nonphysiological artifacts, data were inspected with a 0.30 Hz forward high-pass filter (6 db per octave), and epochs were sorted according to their maximum amplitudes (i.e., max minus min within each epoch) and maximum gradients (i.e., maximum step size for consecutive time points within each epoch). In the initial sample of 34 participants, four participants were excluded because they showed excessive artifacts on more than 50% of all trials. In contrast, the remaining 30 participants in the final sample had fewer than 20% of epochs in each load ! event combination that showed excessive maximum amplitudes and gradients. For these participants, epochs with artifacts were excluded. The remaining epochs were baseline corrected with a 100ms interval before the event. Separate mean ERPs were computed for each load ! event combination. Inspection of ERP responses to pictures (spiders and mushrooms) across all participants revealed clear peaks for P1, N1, P2, and LPP. The LPP was computed as the mean amplitude extracted between 300–550 ms across CP1/2, P1/2, CPz, and Pz electrodes with a linked mastoid reference. These parameters are comparable to those in previous studies. For each participant, mean LPP amplitudes were extracted for each combination of load (high, low) ! picture type (spider, mushroom). In a preliminary analysis, the behavioral data were analyzed in a mixed ANOVA with the between-subjects variable spider fear (yes and no) and the within-subjects variables load (high and low), and event type (spider, mushroom, and blank). However, because spider-fearful participants would respond strongest to spiders with the possibility of a generalized response to mushrooms, a specific contrast analysis was performed only with spiders and blanks (excluding mushrooms) to provide a direct test of whether spider fear affected task performance to spiders. Observed effect sizes are reported as partial eta squared (ZP2) in ANOVAs, and significance levels are reported after Greenhouse-Geisser correction and were considered significant if they were below an alpha level of .05, two-tailed.

Never mind the spider Results Questionnaire Data and Task Ratings Table 1 shows demographics, questionnaire data, and task ratings for the spider-fearful and non-fearful groups. The table also shows results of independent samples t-tests. As expected, spider fear was higher in the spider-fearful than non-fearful group. The two groups did not differ in state positive and negative affect either before or after the task (Watson, Clark, & Tellegen, 1988). But, compared to the non-fearful group, the spider-fearful group showed elevated levels of snake fear, disgust sensitivity (Haidt, McCauley, & Rozin, 1994), and trait anxiety (Spielberger, 1983), which is consistent with previous reports (e.g., Olatunji & Deacon, 2008). As further shown in Table 1, when participants rated difficulty of the letter discrimination task on a 0–10 scale ranging from extremely easy to extremely difficult, both groups gave the task similarly high difficulty ratings of about 7. Compared to the nonfearful group, the spider-fearful group found it more difficult to ignore pictures, particularly so when these were spiders rather than mushrooms (po.05 for the interaction between spider fear and picture, cf. Table 1). Notably, not only the spider-fearful group reported being more distracted by spiders than mushrooms, t(14) 5 4.25, po.001, but also the non-fearful group, t(14) 5 2.31, p 5 .037. This effect was probably caused by differences in physical features of the spiders and mushrooms rather than differences in their emotional valence, as the non-fearful group reported similar feelings towards the spiders and mushrooms, to1. In contrast, the fearful group reported more negative feelings to spiders than mushrooms, t(14) 5 3.70, p 5 .002. Task Performance Figure 1 shows mean letter discrimination ability (d’) across load during spiders, mushrooms, and blanks for the spider-fearful and non-fearful groups. In the preliminary ANOVA of d’ with spider fear (yes and no), load (high and low), and event (spider, mushroom, and blank), there was a main effect of load in that participants in general performed worse during high load (d’ 5 1.97

Figure 1. Effects of spiders, mushrooms, and blanks on mean performance ( " SEM) on the letter discrimination task for the spiderfearful group (n 5 15) and the non-fearful group (n 5 15). The error bars represent standard errors of separate within-subjects ANOVAs for each group (Loftus & Masson, 1994). Asterisks denote 2-tailed significance levels of paired t-tests in each group. nnnpo.001, npo.05.

1155 or 64% hits) than low load (d’ 5 3.31 or 85% hits), F(1, 28) 5 191.78, p 5 .001, ZP2 5 .87. This provides a manipulation check of load. As suggested in Figure 1, the ANOVA showed a main effect of event across participants, F(2,56) 5 3.89, p 5 .029, ZP2 5 .12; however, there was no main effect of group, F(1, 28) 5 3.46, p 5 .074, ZP2 5 .11. Critically, the three-way interaction between spider fear, load, and event was not significant and showed an effect size that was nil, Fo1, p4.95, ZP2o.001. Although the spider fear ! event interaction only approached significance, F(2,56) 5 2.26, p 5 .12, ZP2 5 .08, the specific contrast interaction of spiders versus blanks (i.e., without mushrooms) confirmed that spider fear impaired performance to spiders, F(1,28) 5 4.24, p 5 .05, ZP2 5 .13. As shown in Figure 1, this result demonstrates that spiders were more distracting to spider-fearful participants than non-fearful participants. Critically, this effect did not interact with load; for the 3-way interaction of spiders versus blanks (i.e., without mushrooms), spider fear, and load, Fo1, p4.85, ZP2 5 .001. The ANOVA with spider fear (yes and no), load (high and low), and event (spider, mushroom, and blank) showed no significant main or interaction effect of spider fear on the criterion C, p4.15. Similarly, the ANOVA of RT for hits showed no effects of spider fear, p4.30. However, participants were generally slower during high load (796 ms) than low load (726 ms), F(1,28) 5 33.58, p 5 .001, ZP2 5 .55. This result indicates that there was no speed-accuracy tradeoff between high and low load. Thus, during high load, participants responded more slowly and made more errors than during low load. ERP Results LPP. In the ANOVA of LPP with the variables spider fear (yes and no), load (high and low), and picture type (spider, mushroom), the three-way interaction was not significant and showed an effect size that was nil, Fo1, p4.95, ZP2o.001. In contrast, spider fear interacted with picture, F(1, 28) 5 6.18, p 5 .019, ZP2 5 .18. Thus, the fearful group showed relatively larger LPP mean amplitudes to spiders (5.60 mV) than mushrooms (3.27 mV), t(14) 5 5.16, p 5 .001, compared to the nonfearful group (4.95 mV for spiders and 4.01 mV for mushrooms), t(14) 5 2.87, p 5 .012. This effect is also displayed in Figure 2 that shows mean waveforms for spiders and mushrooms separately for the two spider-fear groups. Across participants, LPP was larger to spiders (5.28 mV) than mushrooms (3.64 mV), F(1,28) 5 34.38, p 5 .001, ZP2 5 .55. Also, LPP was larger during low load (4.96 mV) than high load (3.96 mV), F(1,28) 5 10.88, p 5 .003, ZP2 5 .28. Correlations with other variables. As shown in Table 1, the spider-fearful group showed not only greater spider fear but also greater snake fear, disgust sensitivity, and trait anxiety (e.g., Olatunji & Deacon, 2008; Peira, Golkar, Larsson, & Wiens, 2009). To assess their effects, correlations between spider fear and amplitude differences between spiders and mushrooms were computed for LPP after partialing out snake fear, disgust sensitivity, and trait anxiety. The partial correlation (which gives identical results to an ANCOVA) between spider fear and LPP difference scores was r 5 .19 (zero-order r 5 .43), p4.10. Similarly, snake fear, disgust sensitivity, and trait anxiety did not have a significant effect on LPP difference scores after partialing out spider fear (p4.10). Strictly speaking, these results do not demonstrate that spider fear accounted for the elevated LPP

1156

J. Norberg et al.

Figure 2. Mean ERP waveforms at six electrode sites to spiders and mushrooms (collapsed across load) for the spider-fearful group (n 5 15) and the nonfearful group (n 5 15). Data were referenced to linked mastoid and, for display only, were low-pass filtered at 30 Hz (zero phase shift with slope of 24 dB/ octave). From the six electrodes, the late positive potential (LPP) was computed as the mean amplitude between 300–550 ms (see horizontal line in bold in each panel). Note that the spider-fearful group showed higher LPP to spiders than mushrooms compared to the non-fearful group, which yielded a significant interaction between spider fear and spiders vs. mushrooms, p 5 .019, ZP2 5 .18. F 5 fearful group; N 5 non-fearful group; spi 5 spider; and mush 5 mushroom.

amplitudes to spiders compared to mushrooms; however, they do not suggest either that a variable other than spider fear had an effect independent from spider fear.

Discussion Spider-fearful participants showed elevated LPP amplitudes and impaired performance to task-irrelevant pictures of spiders at fixation. Critically, these effects were not moderated by perceptual load. The absence of a three-way interaction between spider fear, picture type, and load on LPP and task performance suggests that strong emotional stimuli at fixation may resist manipulations of perceptual load. In contrast to the present findings, previous research suggests that emotional processing is reduced, if not eliminated, when perceptual load on a concurrent task is increased (Erthal, de Oliveira, Mocaiber, Pereira, Machado-Pinheiro, et al., 2005; Okon-Singer et al., 2007; Pessoa, 2005). However, most ERP and fMRI studies have used weak emotional stimuli such as facial expressions. So, consistent with the idea that pictures of spiders are potent emotional stimuli for spider-fearful individuals (Globisch et al., 1999), the present findings suggest that these emotional reactions are not easily reduced by perceptual load. Notably, the present findings were obtained even though the emotional pictures were always task irrelevant. Further, because pictures in the present study were always shown at fixation whereas other studies showed task-irrelevant pictures in the periphery, the present study isolated effects of perceptual load on

spatially attended but task-irrelevant emotional pictures (OkonSinger et al., 2007). Because non-significant findings are often problematic, alternative explanations need to be considered for the lack of modulation of emotional processing by perceptual load (i.e., the nonsignificant three-way interaction between fear group, picture type, and perceptual load). First, low power is probably not a problem because the effect sizes were basically zero (ZP2o.001). Second, a failure to obtain an effect of emotion on LPP and task performance is not a problem because significant effects of emotion were obtained across load. Regarding LPP, spider-fearful participants showed greater LPP to spiders than mushrooms compared to non-fearful participants, which was supported by a significant two-way interaction between spider fear and picture type (see Figure 2). Regarding task performance, spider-fearful participants performed worse to spiders than blanks compared to non-fearful participants, which was supported by a significant contrast two-way interaction between spider fear and picture type (see Figure 1). Third, a failure to manipulate load is not a problem because, in general, participants made more errors and responded more slowly under high load (d’ 5 1.97, RT 5 796 ms) than low load (d’ 5 3.31, RT 5 726 ms). Fourth, a weak manipulation of load is unlikely because the high-load condition matched the high-load condition in other studies (e.g., Bishop et al., 2007; Okon-Singer et al., 2007), and because participants in our study rated the task as very difficult. The ANOVA of LPP amplitudes also yielded a main effect of picture type and a main effect of load. The main effect of picture type showed that across fear groups, LPP was greater to spiders than mushrooms. Indeed, even non-fearful participants showed

Never mind the spider

1157

greater LPP to spiders than mushrooms. However, because nonfearful participants rated spiders and mushrooms to be similar in valence (see Table 1), their greater LPP to spiders than mushrooms is unlikely to be evidence for an emotional reaction. Further, the main effect of load showed that, across fear groups, LPP was greater during low load than high load. Unfortunately, because the low load condition had 3 letters and the high load condition had 6 letters, this confound in physical features may have affected LPP (Luck, 2005). Critically, the focus of the study was on the two-way interaction between fear group and picture type and also the three-way interaction with perceptual load. These effects are independent from the main effects of picture type and load and allowed us to isolate the moderating effect of perceptual load on LPP to potent emotional pictures. Spider-fearful participants had not only high spider fear but also elevated snake fear, disgust sensitivity, and trait anxiety. Although correlations among these variables are common (e.g., Olatunji & Deacon, 2008; Peira et al., 2009), this finding raises the question whether the effects on LPP and task performance are specific to spider fear or result from these other variables. In a previous study on spider fear in a change detection task, we found that, although other variables correlated with spider fear, there remained a significant partial correlation between spider fear and reaction time to spiders after controlling for these other variables (Peira et al., 2009). This finding suggests that spider fear was the underlying mechanism that affected reaction time to spiders. However, in the present study, spider fear correlated with enhanced LPP amplitudes to spiders, but after controlling for these other variables, the partial correlation with spider fear was not significant. Conversely, none of these other variables had a significant effect after partialing out spider fear. Therefore, from a strict statistical perspective, the present findings do not allow us to rule out that any of these other variables may account, at least in part, for the relationships between spider fear and enhanced LPP amplitudes to spiders. But, this conclusion may not be an issue. On one hand, because the other three variables also reflect emotional processing, their possible involvement does not challenge the conclusion that strong emotional processing may be immune to high load; instead, findings may imply that,

aside from spider fear, other emotional processes such as trait anxiety may also play a role. On the other hand, previous research found that state anxiety and trait anxiety affected responses to task-irrelevant fearful faces; critically, these effects were eliminated under high load (Bishop et al., 2007). Because it seems reasonable to assume that, for people with high state or trait anxiety, spiders are no more emotionally significant than fearful faces, perceptual load would have been predicted to eliminate LPP to spiders if state or trait anxiety played a major role. Therefore, the most parsimonious explanation for the present findings is that spider fear accounts for the observed relationships between spider fear and enhanced LPP amplitudes to spiders. The findings of enhanced LPP to spiders in spider-fearful participants replicate previous findings of enhanced LPP responses to emotional pictures in general (Olofsson et al., 2008) and to spider pictures by spider-fearful individuals (Flykt & Caldara, 2006; Kolassa et al., 2005, 2006; Miltner et al., 2005; Trippe et al., 2007). The present findings are consistent with the LPP as an index of allocation of attentional resources to motivationally relevant stimuli (Bradley, 2009). However, recent research suggests that the LPP may comprise several independent but overlapping processes. For example, principal component analysis of the LPP to emotional pictures suggests that it consists of at least three independent components (Foti, Hajcak, & Dien, 2009). Further, when emotional and neutral pictures were shown at fixation and were task relevant or irrelevant in different blocks, both emotion and task relevance affected LPP although some differences were observed (Ferrari, Codispoti, Cardinale, & Bradley, 2008; Schupp, Stockburger, Codispoti, Junghofer, Weike, & Hamm, 2007). Therefore, future research needs to isolate the various processing stages that are currently subsumed under the LPP. Because task-irrelevant spiders at fixation elicited enhanced LPP amplitudes in spider-fearful individuals even under high perceptual load, the present findings provide evidence for the processing of task-irrelevant but potent emotional pictures independently of attention. Although these results do not challenge the basic idea of the Load theory (Lavie, 2005), they provide a limit of Load theory in that perceptual load may have little if any effects on processing of strong emotional pictures at fixation.

REFERENCES A˚hs, F., Pissiota, A., Michelga˚rd, A˚., Frans, O¨., Furmark, T., Appel, L., et al. (2009). Disentangling the web of fear: Amygdala reactivity and functional connectivity in spider and snake phobia. Psychiatry ResearchFNeuroimaging, 172, 103–108. Berg, P., & Scherg, M. (1994). A multiple source approach to the correction of eye artifacts. Electroencephalography and Clinical Neurophysiology, 90, 229–241. Bishop, S. J., Jenkins, R., & Lawrence, A. D. (2007). Neural processing of fearful faces: Effects of anxiety are gated by perceptual capacity limitations. Cerebral Cortex, 17, 1595–1603. Bradley, M. M. (2009). Natural selective attention: Orienting and emotion. Psychophysiology, 46, 1–11. Cuthbert, B. N., Schupp, H. T., Bradley, M. M., Birbaumer, N., & Lang, P. J. (2000). Brain potentials in affective picture processing: Covariation with autonomic arousal and affective report. Biological Psychology, 52, 95–111. Eimer, M., & Holmes, A. (2007). Event-related brain potential correlates of emotional face processing. Neuropsychologia, 45, 15–31. Eimer, M., Holmes, A., & McGlone, F. P. (2003). The role of spatial attention in the processing of facial expression: An ERP study of rapid brain responses to six basic emotions. Cognitive, Affective & Behavioral Neuroscience, 3, 97–110.

Erthal, F. S., de Oliveira, L., Mocaiber, I., Pereira, M. G., MachadoPinheiro, W., Volchan, E., et al. (2005). Load-dependent modulation of affective picture processing. Cognitive Affective & Behavioral Neuroscience, 5, 388–395. Ferrari, V., Codispoti, M., Cardinale, R., & Bradley, M. M. (2008). Directed and motivated attention during processing of natural scenes. Journal of Cognitive Neuroscience, 20, 1753–1761. Flykt, A., & Caldara, R. (2006). Tracking fear in snake and spider fearful participants during visual search: A multi-response domain study. Cognition & Emotion, 20, 1075–1091. Foti, D., Hajcak, G., & Dien, J. (2009). Differentiating neural responses to emotional pictures: Evidence from temporal-spatial PCA. Psychophysiology, 46, 521–530. Globisch, J., Hamm, A. O., Esteves, F., & O¨hman, A. (1999). Fear appears fast: Temporal course of startle reflex potentiation in animal fearful subjects. Psychophysiology, 36, 66–75. Haidt, J., McCauley, C., & Rozin, P. (1994). Individual differences in sensitivity to disgustFA scale sampling 7 domains of disgust elicitors. Personality and Individual Differences, 16, 701–713. Holmes, A., Kiss, M., & Eimer, M. (2006). Attention modulates the processing of emotional expression triggered by foveal faces. Neuroscience Letters, 394, 48–52.

1158 Holmes, A., Vuilleumier, P., & Eimer, M. (2003). The processing of emotional facial expression is gated by spatial attention: Evidence from event-related brain potentials. Cognitive Brain Research, 16, 174–184. Ille, N., Berg, P., & Scherg, M. (2002). Artifact correction of the ongoing EEG using spatial filters based on artifact and brain signal topographies. Journal of Clinical Neurophysiology, 19, 113–124. Kolassa, I.-T., Musial, F., Mohr, A., Trippe, R. H., & Miltner, W. H. R. (2005). Electrophysiological correlates of threat processing in spider phobics. Psychophysiology, 42, 520–530. Kolassa, I.-T., Musial, F., Kolassa, S., & Miltner, W. H. R. (2006). Event-related potentials when identifying or color-naming threatening schematic stimuli in spider phobic and non-phobic individuals. BMC Psychiatry, 6, 38. Lavie, N. (1995). Perceptual load as a necessary condition for selective attention. Journal of Experimental PsychologyFHuman Perception and Performance, 21, 451–468. Lavie, N. (2005). Distracted and confused?: Selective attention under load. Trends In Cognitive Sciences, 9, 75–82. Loftus, G. R., & Masson, M. E. J. (1994). Using confidence-intervals in within-subject designs. Psychonomic Bulletin & Review, 1, 476–490. Luck, S. J. (2005). An introduction to the event-related potential technique. Cambridge, MA: MIT Press. Miltner, W. H. R., Trippe, R. H., Krieschel, S., Gutberlet, I., Hecht, H., & Weiss, T. (2005). Event-related brain potentials and affective responses to threat in spider/snake-phobic and non-phobic subjects. International Journal of Psychophysiology, 57, 43–52. Okon-Singer, H., Tzelgov, J., & Henik, A. (2007). Distinguishing between automaticity and attention in the processing of emotionally significant stimuli. Emotion, 7, 147–157. Olatunji, B. O., & Deacon, B. J. (2008). Specificity of disgust sensitivity in the prediction of fear and disgust responding to a brief spider exposure. Journal of Anxiety Disorders, 22, 328–336. Olofsson, J. K., Nordin, S., Sequeira, H., & Polich, J. (2008). Affective picture processing: An integrative review of ERP findings. Biological Psychology, 77, 247–265. Peira, N., Golkar, A., Larsson, M., & Wiens, S. (2009). What you fear will appear: Detection of schematic spiders in spider fear. Experimental Psychology, 57, in press. Pessoa, L. (2005). To what extent are emotional visual stimuli processed without attention and awareness? Current Opinion in Neurobiology, 15, 188–196.

J. Norberg et al. Pessoa, L., Padmala, S., & Morland, T. (2005). Fate of unattended fearful faces in the amygdala is determined by both attentional resources and cognitive modulation. NeuroImage, 28, 249–255. Rinck, M., Bundschuh, S., Engler, S., Muller, A., Wissmann, J., Ellwart, T., et al. (2002). Reliability and validity of German versions of three instruments measuring fear of spiders. Diagnostica, 48, 141–149. Sabatinelli, D., Lang, P. J., Keil, A., & Bradley, M. M. (2007). Emotional perception: Correlation of functional MRI and event-related potentials. Cerebral Cortex, 17, 1085–1091. Schupp, H. T., Stockburger, J., Codispoti, M., Junghofer, M., Weike, A. I., & Hamm, A. O. (2007). Selective visual attention to emotion. Journal of Neuroscience, 27, 1082–1089. Spielberger, C. D. (1983). Manual for the state-trait anxiety inventory. Palo Alto, CA: Consulting Psychologists. Squires, K. C., Donchin, E., Herning, R. I., & McCarthy, G. (1977). On the influence of task relevance and stimulus probability on eventrelated-potential components. Electroencephalography and Clinical Neurophysiology, 42, 1–14. Trippe, R. H., Hewig, J., Heydel, C., Hecht, H., & Miltner, W. H. R. (2007). Attentional blink to emotional and threatening pictures in spider phobics: Electrophysiology and behavior. Brain Research, 1148, 149–160. Vuilleumier, P. (2005). How brains beware: Neural mechanisms of emotional attention. Trends in Cognitive Sciences, 9, 585–594. Vuilleumier, P., Armony, J. L., Driver, J., & Dolan, R. J. (2001). Effects of attention and emotion on face processing in the human brain: An event-related fMRI study. Neuron, 30, 829–841. Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affectFThe PANAS scales. Journal of Personality and Social Psychology, 54, 1063–1070. Wendt, J., Lotze, M., Weike, A. I., Hosten, N., & Hamm, A. O. (2008). Brain activation and defensive response mobilization during sustained exposure to phobia-related and other affective pictures in spider phobia. Psychophysiology, 45, 205–215. Whalen, P. J. (1998). Fear, vigilance, and ambiguity: Initial neuroimaging studies of the human amygdala. Current Directions in Psychological Science, 7, 177–188. Wiens, S., Peira, N., Golkar, A., & O¨hman, A. (2008). Recognizing masked threat: Fear betrays, but disgust you can trust. Emotion, 8, 810–819. (Received June 27, 2009; Accepted November 25, 2009)

Psychophysiology, 47 (2010), 1159–1166. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01020.x

Interactions between systemic hemodynamics and cerebral blood flow during attentional processing

STEFAN DUSCHEK,a HEIKE HEISS,a MARCO F.H. SCHMIDT,b NATALIE S. WERNER,a and DANIEL SCHUEPBACHc a

Department of Psychology, University of Munich, Munich, Germany Department of Developmental Psychology, Max Planck Institute for Evolutionary Anthropology Leipzig, Leipzig, Germany Psychiatric University Hospital Zu¨rich, Zu¨rich, Switzerland

b c

Abstract The study explored interactions between systemic hemodynamics and cerebral blood flow during attentional processing. Using transcranial Doppler sonography, blood flow velocities in the middle cerebral arteries (MCA) of both hemispheres were recorded while 50 subjects performed a cued reaction time task. Finger arterial pressure and heart rate were also continuously monitored. Doppler sonography revealed a right dominant blood flow response. The extent of the increase measured in second two of the interstimulus interval showed a clear positive association with reaction speed. Task-related changes in blood pressure and heart rate proved predictive of changes in MCA flow velocities in limited time windows of the response. Besides an association between cerebral blood flow and attentional performance, the results suggest a marked impact of systemic hemodynamics on the blood flow response. All observed interactions are highly dynamic in time. Descriptors: Cerebral blood flow, Neurovascular coupling, Attention, Doppler sonography

flow increases and task performance. Taking advantage of the high temporal resolution of fTCD, further studies focused on the time dynamics of this relationship. Schuepbach, Boeker, Duschek, and Hell (2007) conducted a component specific analysis of the hemodynamic response in the basal cerebral arteries during mental planning. They reported that the early component of the response, i.e., the change in flow velocity in the left MCA during the second second after task onset, accounted for the largest proportion of variance in planning ability. Similarly, using an attention task Duschek, Schuepbach, and Schandry (2008) obtained the closest association between modulations in bilateral MCA flow velocities and performance during the second and third seconds of task execution. In both studies, later response components were less closely associated or were unrelated to performance indices. On account of this, the authors hypothesized that the connection between the cerebral blood flow response and cognitive performance is dynamic and becomes apparent predominantly in a relatively early time window. Modulations in systemic cardiovascular function during cognition are also well established. A large number of studies revealed, for instance, changes in heart rate, blood pressure, and cardiac contractility during various types of tasks (Bohlin & Kjellberg, 1979; Coles & Duncan-Johnson, 1975; Duschek, Muckenthaler, Werner, & Reyes del Paso, 2009; van Roon, Mulder, Althaus, & Mulder, 2004). Modulations in systemic and cerebral hemodynamics are commonly assumed to be determined by different physiological mechanisms and to occur almost independently of each other. While the increase in cerebral blood flow during neural activation relates to metabolically and neu-

Functional interactions between nerve-cell activity and cerebral blood perfusion were first observed more than a century ago (Mosso, 1881; Roy & Sherrington, 1890). As a result of an augmented metabolic rate of the nerve-cells, neural activation leads to dilation of cerebral arterioles and capillaries followed by increased blood flow in the active tissue (Iadecola, 2004). In addition to flow metabolism coupling, neural mechanisms are involved in blood flow modulation during cerebral activation (Szirmai, Amrein, Pa´lvo¨gyi, Debreczeni, & Kamondi 2005). A fast-acting neural system is assumed to directly trigger dilation of cortical microvessels as a response to brain stem activation (Sa´ndor, 1999; Sato, Sato, & Uchida, 2001). These mechanisms can, for instance, be observed in the enhancement of cerebral blood flow during cognition and other psychological processes (Logothetis, Pauls, Augath, Trinath, & Oeltermann, 2001). A recent line of research investigated relationships between cerebral blood flow modulation and mental performance. For this purpose, functional transcranial Doppler sonography (fTCD) was applied, an ultrasonic technique that enables continuous measurement of blood flow velocities in the basal cerebral arteries (Aaslid, Markwalder, & Nornes, 1982; Deppe, Ringelstein, & Knecht, 2004). Duschek and Schandry (2004, 2006) recorded flow velocities in both middle cerebral arteries (MCA) during the execution of attention and arithmetic tasks, and found correlations between the magnitudes of task-induced Address reprint requests to: Dr. Stefan Duschek, Ludwig-Maximilians-Universita¨t Mu¨nchen, Department Psychologie, Leopoldstr. 13, 80802 Munich, Germany. E-mail: [email protected] 1159

1160 rally transmitted vasodilation, changes in systemic hemodynamics are mediated by a central autonomic network including brain stem nuclei, the basal ganglia, as well as hypothalamic, limbic, and prefrontal areas (Craig, 2002, 2003; Iadecola, 2004). In addition, processes of cerebral autoregulation usually keep brain perfusion widely constant by buffering systemic blood pressure oscillations. In order to ensure stable perfusion, cerebral resistance vessels constrict during increases and dilate during reductions in blood pressure (Paulson, 2002). A number of observations, however, challenged the assumption of complete independence of systemic hemodynamics and cerebral blood flow (Duschek, Werner, Kapan, & Reyes del Paso, 2008). A substantial impact of fluctuations in blood pressure on brain perfusion was reported, for instance, in patients with hypertension and neurological diseases, as well as in persons experiencing posturally related syncopes (Chillon & Baumbach, 1997; Claydon & Hainsworth, 2003; Novak, Novak, Spies, & Low, 1998). In individuals with chronically low blood pressure, markedly reduced cerebral blood flow responses during mental activity were described (Duschek & Schandry, 2004; Stegagno, Patritti, Duschek, Herbert, & Schandry, 2007). In the same population, the extent of the rise in blood pressure during arithmetic processing correlated significantly with the increase in bilateral MCA blood flow velocities (Duschek & Schandry, 2006). Even though the interdependence between systemic and cerebral hemodynamics observed in these studies may be attributed to autoregulatory deficits related to the respective clinical conditions, its occurrence may not be ruled out in healthy normotensive persons either. Modulations in blood pressure and heart rate, which accompany cognitive processes, occur relatively fast, and their time courses tend to be complex, including distinct periods of increases and decreases. This is crucial since the autoregulatory response to such fluctuations is relatively inert, taking place with a delay of several seconds (Florence & Seylaz, 1992; Paulson, Strandgaard, & Edvinsson, 1990; Zhang, Zuckerman, Giller, & Levine, 1998). Thus, at least during specific phases of the cerebral hemodynamic reaction, effects of the systemic circulation may not be fully compensated, and blood flow may rise and fall with blood pressure and heart rate. This dependence of cerebral blood flow modulation during cognition on systemic hemodynamics is of particular interest given the described importance of cerebral hemodynamic adjustment for optimum performance (Duschek & Schandry, 2004, 2006; Duschek et al., 2008; Schuepbach et al., 2007). In order to gain further insight into the interaction between systemic and cerebral hemodynamics and its temporal dynamics, high time resolution analysis is required. For this purpose, fTCD combined with continuous peripheral hemodynamic recording is certainly a valuable tool. In the present study, cerebral and systemic hemodynamics were investigated based on a cued reaction time task. Paradigms of this type focus on the arousal component of attention. In particular, they look at the short-term increase of attentiveness during the anticipation of a significant event, which enables a rapid adjustment to situational requirements. This function of ‘‘phasic arousal’’ is considered to be a specific component of the human attentional system that is undoubtedly of vast importance in everyday life (Johnson & Proctor, 2004; Posner & Rafal, 1987). A large part of the brain areas relevant for the control of attentional arousal, such as the dorsolateral frontal and the inferior parietal lobes, are parts of the perfusion territory of the MCA (Pardo, Fox, & Raichle, 1991; Paus, Zatorre, Hofle, Caramanos, Gor-

S. Duschek et al. man, et al., 1997; Haines, 2007), which can easily be accessed using fTCD recordings. MCA measurements also proved suitable in previous studies based on this type of task (Duschek & Schandry, 2004, Duschek, Hadjamu, & Schandry, 2007; Duschek et al., 2008). Regarding systemic hemodynamics, the execution of a cued reaction time task is accompanied, for instance, by characteristic changes in heart rate, which are related to processes such as attentional focusing and orientation, as well as preparation of the motor response (Bohlin & Kjellberg, 1979; Hugdahl, 2001). Blood flow velocities in the MCA and finger blood pressure were continuously monitored in the study, and changes were quantified in consecutive time windows of 1 s duration. A main research question addressed possible interactions between cognitively induced modulations in blood pressure and heart rate on the one hand and in MCA perfusion on the other. The methodological approach applied should prove particularly useful for analyzing possible differences in these interactions as a function of the time course of the hemodynamic response. Considering the reasoning presented above, we expected positive associations between modulations in systemic and cerebral hemodynamic modulations at least during limited phases of the response. A secondary question pertained to the association between cerebral blood flow modulation and attentional performance. Based on earlier results (e.g., Duschek et al., 2008; Schuepbach et al., 2007), we hypothesized that higher magnitudes of the increase in MCA blood flow velocities, especially during the first seconds of task execution, are associated with better performance.

Methods Participants Fifty subjects (33 women, 17 men) participated in the study. Exclusion criteria comprised severe physical diseases, psychiatric disorders, as well as the use of psychoactive drugs or medication affecting the cardiovascular system. All participants were righthanded according to the Edinburgh Handedness Inventory (Oldfield, 1971). Forty-five of the participants were university students, 3 were employees, and 2 were self-employed. Information regarding age, Body Mass Index (BMI, kg/m2), as well as resting blood pressure and heart rate is presented in Table 1. Task Characteristics The task was presented on a computer using the ‘‘Experimental Runtime System’’ software program (BeriSoft Cooperation, 2000). The white outline of a small cross (6 ! 6 mm) was shown on the screen. After 55 s, an acoustic cue was presented (400-Hz tone of 500 ms duration). Five seconds after the cue, the image was replaced with a full white cross of the same size. This served as the imperative stimulus requiring an immediate keystroke. The task consisted of a total of 20 trials. A constant inter-trial interval may potentiate possible habituation and facilitate automatic task responses. The 55-s duration of the inter-trial interval, however, seemed long enough to prevent such processes. In order to control for laterality effects, half of the participants carried out the first 10 trials with the right hand and the remaining with the left hand. The sequence was reversed in the second half of the participants. The subjects were asked to sit still, not to speak, and to look at the cross for the entire duration of the task. Reaction times were recorded automatically and aggregated by calculating the median for each subject. Medians were used instead of means,

Systemic hemodynamics and cerebral blood flow

1161

Table 1. Age, Body Mass Index, Resting Blood Pressure, and Heart Rate in the Sample

Age in years Body mass index in kg/m2 Systolic blood pressure in mmHg Diastolic blood pressure in mmHg Resting heart rate in beats per min

M

SD

Min

Max

24.1 22.2 111.9 74.7 72.2

4.6 2.9 12.6 8.0 11.4

20 17.1 91.7 61.0 40.0

48 29.4 145.0 95.7 95.3

Note: Means (M), standard deviations (SD), minimal (Min) and maximal (Max) values.

in order to prevent distortion of the data due to outliers. Reaction times for the trials performed with the left and right hands correlated highly (r 5 .75, po.01), hence only the average reaction times across all trials were included in the statistical analysis. Recording of Cerebral Blood Flow Velocities, Blood Pressure, and Heart Rate For the purpose of cerebral hemodynamic recording, a commercially available Doppler sonography device (Multidop L2, DWL Elektronische Systeme, Sipplingen, Germany) was employed. Blood flow velocities were monitored simultaneously in both MCA. The recordings were obtained through the temporal bone windows, using two 2-MHz transducer probes. Insonation of the MCA took place at a depth of 50 mm in all subjects. Following vessel identification, the ultrasonic probes were fixed to the head using a tight rubber band. The spectral envelope curves of the Doppler signal were stored at a rate of 28 samples per second (for insonation technique and validation of fTCD see Dahl, Russell, Nyberghansen, & Rootwelt, 1992; Duschek & Schandry, 2003; Jorgensen, Perko, & Secher, 1992; Larsen, Olsen, Hansen, Paulson, & Knudsen, 1994). Blood pressure and heart rate were monitored continuously using a Finometer device (Model-2, Finapres Medical Systems, Amsterdam, The Netherlands). The cuff of the Finometer was applied to the mid-phalanx of the third finger of the right hand. In order to control for the influence of hydrostatic level errors, the height correction unit integrated in the device was used. To enable periodic recalibration, the ‘‘Physiocal’’ feature (Wesseling, De Wit, Van der Hoeven, Van Goudoever, & Settels, 1995) was put into operation. The signal was digitized at a sample rate of 200 Hz. Procedure The experimental sessions were conducted in a silent, dimly lit room. Prior to the experimental procedure, blood pressure and heart rate were taken using an automatic inflation blood pressure monitor (Omron M9 Premium, Omron Electronics, Schaumburg, IL). Three readings were taken, separated by 1-min rest intervals (Table 1 includes the values averaged across the three measurements.). Following this, the ultrasonic probes and the cuff of the Finometer were mounted, and the cued reaction time task was presented in the described form. Subjects were requested not to drink alcohol or beverages containing caffeine for 3 h prior to the experimental session. Data Analysis The envelope curves revealed by Doppler sonography were analyzed offline using the software AVERAGE (Deppe, Knecht, Henningsen, & Ringelstein, 1997). MCA blood flow velocites were represented by a time and intensity-weighted ‘‘mean flow

velocity index.’’ This score is the least susceptible to artifacts and demonstrates the highest correlation with blood volume flowing through a vessel (Duschek & Schandry, 2003). Flow velocities were integrated over each cardiac cycle and averaged, time locked to the cuing tone. The epochs were set beginning 10 s before the cuing tone and ending 25 s after the imperative stimulus. The mean flow velocity during the 10 s prior to the cuing tone served as a baseline (FVbas). Relative (per cent) changes in flow velocity during task execution (dFV) were calculated for the left and right MCA using the formula dFV 5 [FV(t) ! FVbas] " 100/FVbas, where FV(t) is the flow velocity over the course of time. Following this, mean values of dFV were computed for 30 consecutive response intervals of 1-s duration each. The response intervals covered a period of task execution starting with the onset of the cuing tone (second 1), and ending 25 s after the imperative stimulus (second 30). For the purpose of analyzing the data obtained by continuous blood pressure recording, the program BeatScope 1.1a (Finapres Medical Systems) was employed. Beat-to-beat values of systolic and diastolic blood pressure as well as heart rate were computed and transformed into second-by-second values (c.f. Finapres Medical Systems, 2005). These data were exported and further processed using SPSS 16.0 (SPSS Inc., Chicago, IL). Epochs were created and averaged in the same way as in the procedure concerning the MCA flow velocities. Again, the 10 s before the cuing tone served as a baseline, and percent changes were computed for each of the 30 s of the defined task period. In the first step of the statistic analysis, possible laterality of the cerebral hemodynamic response was determined. For this purpose, t-tests were applied to compare the secondwise flow velocity values between both MCA. Repeated measures analyses of variance (ANOVAs) were computed for MCA flow velocities, blood pressure, and heart rate in order to evaluate intra-subject trends. Associations between task-related modulations in cerebral perfusion and such in systemic hemodynamics were quantified using regression analysis. In order to evaluate differences in the relationships over the course of time, separate models were computed for each of the 1-s response intervals of the defined task period (i.e., 30 separate regression analyses). As predictors, relative values of blood pressure and heart rate were applied. Due to high correlations between systolic and diastolic values in each of the response intervals (mean r 5 .82), mean arterial pressure was used as a single index of blood pressure. ‘‘Forced entry’’ was chosen as a method for entry of predictors. As there was no reason to expect specific effects for the left and right MCA, flow velocity values averaged over both vessels served as dependent variables. Relationships between cerebral blood flow modulation and task performance were quantified by means of regression analysis with the flow velocity values for the 5 s between the cuing tone and the imperative stimulus as predictors and reaction time as dependent variable. Separate models were computed for flow velocity values from the left and right MCA. Correlations between the predictors were high, resulting in strong collinearity (tolerance values o.1 for all flow velocity values). Therefore, the models could be used only to determine the flow velocity value, which accounted for the largest proportion of variance in reaction time. For this purpose, a ‘‘stepwise’’ procedure was applied for entry and removal of predictors, in which the predictor that first enters the model explains the largest portion of criteria variance. In addition, simple Pearson correlations between flow velocity values and reaction time were computed.

1162

S. Duschek et al.

Figure 1 displays the changes in the perfusion of the left and right MCA during execution of the cued reaction time task. The data points represent the relative changes in flow velocities, which were computed for each of the 30 response intervals. A steep bilateral increase was observed after presentation of the cuing tone peaking in second 4. The imperative stimulus was again followed by a rise in flow velocity and a second maximum in second 10. The hemodynamic response was stronger in the right cerebral artery than in the contralateral vessel during a large part of the task period. The statistical analysis indicated significant differences between both MCA for the seconds 3 to 17 (all t[49]42.6, all po.05). In Figure 2, modulations in the systemic hemodynamic parameters are given. Biphasic increases in systolic and diastolic blood pressure were observed with maxima in the seconds 12 (systolic blood pressure) and 11 (diastolic blood pressure). Heart rate decreased in the first seconds. After a transient increase (second 4) it fell once again reaching its minimum in second 7. A biphasic increase occurred during the remaining period with peaks in seconds 10 and 19. The repeated measures ANOVAs revealed significant linear, quadratic, and cubic trends for flow velocities in the left and right MCA, as well as for systolic and diastolic blood pressure (left Cuing Imperative tone stimulus

6 Left MCA Right MCA

5 Changes in %

4 3 2 1

4

HR SBP DBP

3 2 1 0 –1 –2 –3

5

10

15 Time in s

20

25

30

MCA, linear: F[1] 5 65.44, po.01, quadratic: F[1] 5 16.15, po.01, cubic: F[1] 5 46.01, po.01; right MCA, linear: F[1] 5 147.35, po.01, quadratic: F[1] 5 44.90, po.01, cubic: F[1] 5 101.24, po.01; systolic blood pressure, linear: F[1] 5 13.74, po.01, quadratic: F[1] 5 10.72, po.01, cubic: F[1] 5 34.31, po.01; diastolic blood pressure, linear: F[1] 5 26.90, po.01, quadratic: F[1] 5 7.99, po.01, cubic: F[1] 5 54.57, po.01). For heart rate, significant fourth and fifth order trends were also obtained (linear: F[1] 5 60.51, po.01, quadratic: F[1] 5 55.76, po.01, cubic: F[1] 5 7.98, po.01, fourth order: F[1] 5 54.69, po.01, fifth order: F[1] 5 7.49, po.01). The results of the regression analyses concerning the prediction of bilateral MCA blood flow changes by modulations in mean arterial pressure and heart rate are given in Table 2. The table displays the standardized Beta weights and R values for each second of the task period. For heart rate, significant positive Beta weights were found for the initial phase, i.e., the seconds 1 to 6, as well as for the seconds 8 to 10. Significant positive Beta weights for mean arterial pressure were obtained for the seconds 4 to 9, and 14 to 23. Mean reaction time of the participants was 309.4 ms (SD 5 44.9 ms). Two stepwise regression analyses for the prediction of reaction time from flow velocity values from the left and right MCA (seconds 1 to 5) were conducted. For both vessels, the value for second 2 entered the model in the first step (left MCA: standardized Beta 5 ! .35, po.05; right MCA: standardized Beta 5 ! .41, po.01) indicating that bilateral blood flow modulations in second 2 accounted for the largest proportion of variance in reaction time. High collinearity in both models prevented further conclusions. The Pearson correlations between the flow velocity values for the seconds 1 to 5 and reaction time are given in Table 3. Significant negative correlations were obtained for flow modulations in the left MCA for second 2, and for modulations in the right MCA for the seconds 2 to 5.

1

–1 –2

Imperative stimulus

Figure 2. Relative changes in heart rate and blood pressure during task execution (grand average). HR, heart rate; SBP, systolic blood pressure; DBP, diastolic blood pressure.

Results

0

Cuing tone

Changes in %

Alpha-level was set at .05 in all analyses. Taking into account possible occurrence of Type I error inflation related to multiple statistical testing, the use of lower significance criteria may be considered. In the relatively small sample, this, however, would substantially reduce the power of the tests, i.e., increase the chance of Type II errors and reduce the probability of detecting any effects present. The present sample size is comparable to previous studies on interactions between systemic hemodynamics and cerebral blood flow, where between 40 and 80 subjects (e.g., Claydon & Hainsworth, 2003; Duschek & Schandry, 2004, 2006; Duschek et al., 2008), or even less (Zhang et al., 1998) were investigated. Instead of reducing alpha-level, the number of parallel tests could be limited by using longer response intervals, i.e., combining the 1-s intervals, for instance, to make 5-s intervals. This, in turn, would interfere with the aim of high time resolution analysis, which was crucial in our methodological approach.

Discussion 5

10

15 Time in s

20

25

30

Figure 1. Relative changes in flow velocities in the left and right MCA during task execution (grand average).

The study revealed a biphasic increase in blood flow velocities measured in the MCA during the execution of a cued reaction time task, which was more pronounced in the right than in the left vessel. The extent of the flow increase in the left MCA during

Systemic hemodynamics and cerebral blood flow

1163

Table 2. Regression Analyses for the Prediction of Modulations in Bilateral MCA Flow Velocities from Modulations in Mean Arterial Pressure (MAP) and Heart Rate (HR) Over the Course of the Task Standardized Beta

Standardized Beta

Time (s)

MAP

HR

R

Time (s)

MAP

HR

R

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

.10 .02 .08 .31n .46nn .49nn .34n .26n .26n .20 .19 .14 .19 .26n .31n

.45 .36n .32n .41nn .31n .27n .17 .28n .43nn .39nn .19 .05 ! .12 ! .23 ! .24

.46 .36 .33 .56 .63 .59 .40 .37 .49 .45 .28 .15 .21 .33 .36

16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

.32 .26n .27n .31n .30n .30n .33n .27n .15 .05 .04 .00 .06 .22 .21

! .16 .06 .15 .12 .12 .19 .22 .19 .19 .16 .16 .23 .22 .12 .21

.35 .27 .28 .33 .32 .37 .42 .33 .24 .16 .16 .23 .22 .24 .28

nn

n

Note: Standardized Beta- and R-values for each second of the task period. n po.05; nnpo.01.

second 2 after the cuing tone was positively associated with reaction speed. Flow modulations in the right MCA during almost the entire interstimulus interval correlated with performance, the closest association being obtained for second 2. While the courses of the systolic and diastolic blood pressure responses were similar to those in MCA flow velocities, heart rate modulation showed a more complex pattern including two minima and two maxima, respectively. Task-related changes in mean arterial pressure and heart rate proved predictive of these in bilateral MCA flow velocities; the relationships were, however, limited to specific time windows of the hemodynamic response. The task-induced increase in cerebral blood flow velocities can be ascribed to neural activation processes in structures such as the dorsolateral frontal and the inferior parietal lobes, which are relevant for the control of attentional arousal and form parts of the perfusion territory of the MCA (Haines, 2007; Pardo et al., 1991; Paus, Zatorre, Hofle, Caramanos, Gorman, Petrides, & Evans, 1997). The higher increase in the right MCA is in accordance with the assumption of a dominance of the right hemisphere for arousal regulation (Kolb & Whishaw, 2003; Posner & Petersen, 1990). One may hypothesize that the first peak, which appeared between the cuing tone and the imperative stimulus, related to neural activation associated with increase in attentive-

Table 3. Correlations Between Modulations in MCA Blood Flow Velocities During the Seconds 1 to 5 and Reaction Time Time (s) 1 2 3 4 5 po.05; nnpo.01.

n

Left MCA

Right MCA

! .19 ! .35n ! .26 ! .23 ! .26

! .27 ! .41nn ! .34n ! .29n ! .31n

ness, anticipation of the imperative stimulus, as well as preparation of the response. This response component may relate to the contingent negative variation (CNV) and the readiness potential (‘‘Bereitschaftspotential’’) known from electroencephalography (EEG) research (Andreassi, 2000). Also, the assessment of cerebral blood flow by means of near infrared spectroscopy revealed a similar phenomenon during response preparation (Weber, Lutschg, & Fahnenstich, 2004). The second maximum was observed after the imperative stimulus and the ensuing keystroke, and may be attributed to cortical activation associated with a second increase in alertness due to concentration on the imperative stimulus, as well as activation related to the execution of the motor reaction. The time window of the closest association between MCA blood flow modulation and performance (second 2) fits well with previous studies from the field of attention and executive functions (Duschek et al., 2008; Schuepbach et al., 2007). This underlines the specific relationship between early cerebral hemodynamic modulation and cognition. Flow velocity changes in the right MCA were more closely related to reaction time than those in the left vessel, which again emphasizes the specific role of the right hemisphere for attentional arousal. Given the rapid onset of the cerebral hemodynamic response, one may hypothesize that, in addition to neurovascular coupling, fast-acting neurogenic blood flow regulation was involved in its occurrence. Innervation of the cortical arterioles by fibers originating from the brain stem has been well documented (Hamel, Vaucher, Tong, & St-Georges, 2002; Sa´ndor, 1999). These fibres form part of an intracranial neural vasodilative system consisting of cholinergic and serotonergic neurons that project from the ascending reticular activating system to cortical areas (Sato et al., 2001, Szirmai et al., 2005). In animals, the stimulation of brain stem nuclei was shown to be accompanied by cortical blood flow increase (Biesold, Inanami, Sato, & Sato, 1989). Activation of brain stem areas constitutes an integral part of the cerebral processes related to the increase of attentiveness occurring in a cued reaction time task (Posner & Petersen, 1990; Sturm, de Simone, Krause, Specht, Hesselmann, et al., 1999). Thus, also brain stem activity may have contributed to the modulation of flow velocities registered in the MCA. Although the biphasic blood pressure response was presumably related to the same cognitive and motor processes, the beginning of the rise as well as both maxima were observed 1 to 3 s after the increases in cerebral blood flow velocities. Blood pressure modulations during mental activity result from cardiac and vasomotor adjustment, which are initiated by the central autonomic network and transmitted to the cardiovascular system by autonomic nervous and hormonal pathways (Craig, 2002, 2003; Loewy, 1990). As suggested by the present data, these processes are markedly slower than metabolically mediated cerebral blood flow adjustment. According to classic theories, a variety of psychological processes contribute to heart rate responses during attention and reaction time tasks (Hugdahl, 2001; Lacey & Lacey, 1970). The beginning of the present response was characterized by heart rate deceleration. This is a common observation, which is usually ascribed to cognitive processes of orienting and attentional focusing (Coles & Duncan-Johnson, 1975; Venables, 1991). After a transient return to baseline level, heart rate decreased once again. Given that this decline occurred exactly when blood pressure started to increase, it is plausible to attribute it to the cardiac baroreflex. The baroreflex consists of a negative feedback loop, in which blood pressure fluctuations are responded to by com-

1164 pensatory adjustment of heart rate (Levy & Pappano, 2007; Reyes del Paso, Vila, & Garcia, 1994). After the second decline, heart rate increased by a large amount. This accelerative component of the cardiac response is commonly ascribed to internal cognitive processing as well as emotional and motivational aspects of the task (Coles & Duncan-Johnson, 1975; Hugdahl, 2001). A second and smaller acceleration appeared right after the start of the blood pressure decrease, and thus may again be best explained by counter-regulatory activity of the baroreflex. The most interesting result of the study may be the interactions between peripheral and cerebral hemodynamic modulations and their courses across the task. Changes in mean arterial pressure yielded substantial positive Beta weights in the prediction of modulations in MCA blood flow velocities during a large part of the response (seconds 4 to 23) suggesting a considerable impact of systemic arterial pressure on cerebral perfusion. One may hypothesize that the blood pressure increase facilitated the rise in flow velocities, whereas its subsequent decline supported the return of blood flow toward the initial level. In the seconds 10 to 13, the Beta weights did not reach significance, which may be explained by the specific course of the blood pressure response. The named period included the maximum of the response and therefore only slight changes that were apparently without significant impact on cerebral perfusion. The same holds true for the final part of the task period, during which blood pressure was virtually stable. One should, however, not overlook that, in the present correlative design, the interpretation of a causal effect of systemic hemodynamics on cerebral blood flow may not be drawn with complete certainty. This is underlined by the observation that the peaks of the cerebral blood flow response appeared earlier than the maxima of the heart rate and blood pressure responses. One may thus suppose that, in addition to a causal influence of systemic on cerebral hemodynamics, third variable effects may in part account for the correlations. Here, one may consider central nervous regulatory processes, which affect both cerebral blood flow and peripheral cardiovascular processes. A previous study by Zhang et al. (1998) supports the notion that rapid changes in blood pressure are not fully compensated by cerebral autoregulation. The study explored the link between spontaneous fluctuations in arterial pressure and MCA blood flow velocities. Data obtained by Doppler sonography and finger blood pressure measurement were processed in the frequency domain based on transfer function analysis. Oscillations in MCA perfusion occurring in a relatively high frequency range (0.07– 0.30 Hz) were closely associated with oscillations in blood pressure. This was not the case for lower frequencies. The authors therefore characterized autoregulation as a frequency dependent phenomenon. While it is effective in dampening slower fluctuations, changes in the high frequency range are widely transferred to cerebral blood flow. Cerebral autoregulation is mediated by a number of physiological mechanisms including metabolic, myogenic, and endothelium-related factors (Iadecola, 2004; Paulson, 2002). Though the exact response times of these processes are still unknown, it would appear that the dynamics of the autoregulatory system do not allow the full compensation of either spontaneous or psychologically triggered fast blood pressure modulations. Cerebral autoregulation is considered a protective mechanism preventing brain ischemia during phasic blood pressure decrease, and capillary damage, edema formation, and disruption of the blood brain barrier during blood pressure increase (Paulson,

S. Duschek et al. 2002). The mechanism most efficiently operates in the normotensive range (Chillon & Baumbach, 1997). Hence, one may assume that rapid blood pressure fluctuations are even more strongly transferred to cerebral blood flow in the extremes of the tonic blood pressure spectrum, i.e., hypotension and hypertension. In chronically low blood pressure, reduced mental performance has repeatedly been documented (Duschek, Matthias, & Schandry, 2005; Duschek & Schandry, 2007), and there is evidence linking these deficits to insufficient cerebral autoregulation (Duschek & Schandry, 2004, 2006). Impaired autoregulation is furthermore involved in the genesis of symptoms occurring in patients with orthostatic failure (Novak et al., 1998). Also, in the case of chronically elevated blood pressure, cerebral autoregulation does not operate at its optimum (Chillon & Baumbach, 1997). A number of observations suggested blunted vascular reactivity and diminished cerebral blood flow in hypertension, which among other factors may underlie the reduced cognitive performance related to the condition (Jennings, 2003; Waldstein, Ryan, Manuck, Parkinson, & Bromet, 1991). Dependence of cerebral blood flow on systemic blood pressure may be of special importance in individuals with increased blood pressure variability, a group that is well known to be at risk of cerebrovascular disease (Sloan, Shapiro, Bagiella, Myers, & Gorman, 1999). Here, fluctuations in systemic hemodynamics may result in pronounced destabilization of brain perfusion and reduced protection of the neural tissue interfering with optimal cerebral functioning. On account of this, it would certainly be worthwhile to more intensely study the interaction between systemic and cerebral hemodynamic modulation and possible clinical implications in the named populations. In our study, heart rate modulation was also positively associated with that in MCA blood flow. The link was particularly pronounced in the period between the cuing tone and the imperative stimulus, where an overall slight heart rate reduction occurred. This indicates lower amplitudes of the initial component of the cerebral blood flow response in individuals who experienced stronger heart deceleration than in such in which the decline was less pronounced. Like blood pressure, the predictive value of heart rate modulation for changes in MCA blood flow was reduced while at extreme values, i.e., the phase around the heart rate minimum (seconds 6 to 8). It seems notable that modulations in mean arterial pressure and heart rate were associated with changes in cerebral blood flow velocities during different time windows of the response. In this regard, it is important to bear in mind that heart rate and blood pressure are determined by different physiological factors. Heart rate modulation is a product of rather direct autonomic and hormonal influences on sinus node activity, while changes in mean arterial pressure relate to alterations in multiple cardiac and vascular factors such as vasomotor tone, stroke volume, and peripheral resistance (Levy & Pappano, 2007). Classic psychophysiological models view the decrease in heart rate during attentional processing as an adaptive mechanism (Hugdahl, 2001; Porges, 1992; Thayer & Lane, 2009). According to the ‘‘intake rejection hypothesis,’’ for instance, phasic heart deceleration is associated with a state of reduced sensory thresholds and improved conditions for the detection of external stimuli (Lacey & Lacey, 1970). The present findings propose a somewhat different perspective. Considering the positive association between modulations in heart rate and MCA blood flow, one may hypothesize that pronounced heart rate deceleration interferes with efficient cerebral hemodynamic adjustment. As suggested

Systemic hemodynamics and cerebral blood flow

1165

by the link between initial cerebral blood flow modulation and reaction time, lower degrees of blood flow increase may in turn impede attentional performance. The latter association, however, can certainly not be interpreted unambiguously. It is generally believed that modulations in cerebral blood flow during cognitive activity reflect a perpetual adjustment to the changing metabolic demands of neural activation. In particular, the high rate of aerobic metabolism in the neural tissue requires a constant and sufficient supply of oxygen and glucose (Paulson, 2002). It therefore seems reasonable to assume that a stronger increase in cerebral perfusion is accompanied by improved functional conditions, resulting in enhanced cognitive performance (Duschek & Schandry, 2004). An alternative explanation could emphasize a possible association between levels of neural activation and attentional performance. According to this view, stronger blood flow increase in case of better performance would only constitute an epiphenomenon of enhanced nerve-cell activity. The present data do not make it possible to decide between these interpretations. Recent research, however, suggested that differences in cerebral blood flow may indeed causally modulate cognitive function. It could be demonstrated that experimental increase in brain perfusion induced by pharmacological blood pressure elevation is followed by significant enhancement of attentional performance (Duschek et al., 2007). Facilitation of cognitive functioning as a consequence of efficient cerebral blood flow adjustment thus seems a plausible consideration. A limitation of the study is due to the use of a relatively easy reaction time task with complete predictability of the imperative stimulus. Different types of cognitive paradigms with varying levels of difficulty may evoke hemodynamic reactions differing in time course and magnitude. Also, the generalizability of the association between cerebral hemodynamic modulation and task

performance is limited. In more difficult tasks, non-linear, for instance inverted U-shaped, relationships may also occur. The major limitation of fTCD concerns its low spatial resolution, which is determined by the size of the brain area supplied by the artery under study (Duschek & Schandry, 2003). The perfusion territory of the MCA is relatively large, thus our conclusion about task-related modulations in regional blood flow in specific cortical structures must remain hypothetical. In future studies, flow velocities in cerebral arteries other than the MCA should also be assessed. In this regard, the anterior cerebral arteries are of particular interest since they supply medial regions of the cortex including the anterior cingulate, which is of great importance for attentional processes. Regarding data analysis, possible restrictions arise from multiple statistical testing. The use of separate regression models for each of the 30 response intervals is related to the aim of hemodynamic analysis with high time resolution. Even though in principle this approach constitutes a strength of the study, multiple testing inevitably increases the risk of Type I errors, i.e., false rejection of the null hypothesis. A final methodological limitation is due to the imbalance of the sexes in the sample. This prevented the efficient analysis of possible gender effects on hemodynamic responses, which have recently been proposed (Schuepbach, Huizinga, Duschek, Grimm, Boeker, & Hell, 2009). In conclusion, the present study revealed evidence for pronounced interactions between cognitively induced modulations in systemic hemodynamics and such in cerebral blood flow. Like the connection between cerebral blood flow adjustment and cognitive performance, these interactions are highly dynamic in time. The findings underline the importance of the temporal aspect in the investigation of relationships between cardiovascular and central nervous processes and emphasize the suitability of research techniques enabling high time resolution analyses.

REFERENCES Aaslid, R., Markwalder, T. M., & Nornes, H. (1982). Noninvasive transcranial Doppler ultrasound recording of flow velocity in basal cerebral arteries. Journal of Neurosurgery, 57, 769–774. Andreassi, J. L. (2000). Human Behavior and Physiological Response. London: Lawrence Erlbaum. BeriSoft Cooperation (2000). Experimental Run Time System, Version 3.32. Frankfurt a.M.: BeriSoft Cooperation. Biesold, D., Inanami, O., Sato, A., & Sato, Y. (1989). Stimulation of the nucleus basalis of Meynert increases cerebral cortical blood flow in rats. Neuroscience Letters, 98, 39–44. Bohlin, G., & Kjellberg, A. (1979). Orienting activity in two-stimulus paradigms as reflected in heart rate. In H. D. Kimmel, E. H. van Holst, & J. F. Orlebeke (Eds.), The orienting reflex in humans (pp. 169–198). Hillsdale: Lawrence Erlbaum. Chillon, J. M., & Baumbach, G. L. (1997). Autoregulation of cerebral blood flow. In K. M. A. Welch, L. R. Caplan, D. J. Reis, B. K. Siesjo¨, & B. Weir (Eds.), Primer on cerebrovascular diseases (pp. 51–54). San Diego: Academic Press. Claydon, V. E., & Hainsworth, R. (2003). Cerebral autoregulation during orthostatic stress in healthy controls and in patients with posturally related syncope. Clinical Autonomic Research, 13, 321–329. Coles, M. G., & Duncan-Johnson, C. C. (1975). Cardiac activity and information processing. The effects of stimulus significance and detection and response requirements. Journal of Experimental Psychology: Human Perception and Performance, 1, 418–428. Craig, A. D. (2002). How do you feel? Interoception: The sense of the physiological condition of the body. Nature Reviews of Neuroscience, 3, 655–666. Craig, A. D. (2003). Interoception: The sense of the physiological condition of the body. Current Opinion in Neurobiology, 13, 500–505.

Dahl, A., Russell, D., Nyberghansen, R., & Rootwelt, K. (1992). A comparison of regional cerebral blood-flow and middle cerebral-artery blood-flow velocitiesFSimultaneous measurements in healthy subjects. Journal of Cerebral Blood Flow and Metabolism, 12, 1049– 1054. Deppe, M., Knecht, S., Henningsen, H., & Ringelstein, E. B. (1997). AVERAGE: A Windows program for automated analysis of event related cerebral blood flow. Journal of Neuroscience Methods, 75, 147–154. Deppe, M., Ringelstein, E. B., & Knecht, S. (2004). The investigation of functional brain lateralization by transcranial Doppler sonography. NeuroImage, 21, 1124–1146. Duschek, S., Hadjamu, M., & Schandry, R. (2007). Enhancement of cerebral blood flow and cognitive performance due to pharmacological blood pressure elevation in chronic hypotension. Psychophysiology, 44, 145–153. Duschek, S., Matthias, E., & Schandry, R. (2005). Essential hypotension is accompanied by deficits in attention and working memory. Behavioral Medicine, 30, 149–158. Duschek, S., Muckenthaler, M., Werner, N. S., & Reyes del Paso, G. A. (2009). Relationships between features of cardiovascular control and cognitive performance. Biological Psychology, 81, 110–117. Duschek, S., & Schandry, R. (2003). Functional transcranial Doppler sonography as a tool in psychophysiological research. Psychophysiology, 40, 436–454. Duschek, S., & Schandry, R. (2004). Cognitive performance and cerebral blood flow in essential hypotension. Psychophysiology, 41, 905–913. Duschek, S., & Schandry, R. (2006). Deficient adjustment of cerebral blood flow to cognitive activity due to chronically low blood pressure. Biological Psychology, 72, 311–317.

1166 Duschek, S., & Schandry, R. (2007). Reduced brain perfusion and cognitive performance due to essential hypotension. Clinical Autonomic Research, 17, 69–76. Duschek, S., Schuepbach, D., & Schandry, R. (2008). Time-locked association between rapid cerebral blood flow modulation and attentional performance. Clinical Neurophysiology, 119, 1292–1299. Duschek, S., Werner, N., Kapan, N., & Reyes del Paso, G. A. (2008). Patterns of cerebral blood flow and systemic hemodynamics during arithmetic processing. Journal of Psychophysiology, 22, 81–90. Finapres Medical Systems (2005). BeatScope 1.1a. User’s Guide. Amsterdam: Finapres Medical Systems. Florence, G., & Seylaz, J. (1992). Rapid autoregulation of cerebral blood flow: A laser-Doppler flowmetry study. Journal of Cerebral Blood Flow and Metabolism, 12, 674–680. Haines, D. E. (2007). Neuroanatomy. An atlas of structures, sections, and systems. Philadelphia: Lippincott Williams & Wilkins. Hamel, E., Vaucher, E., Tong, X. K., & St-Georges, M. (2002). Neuronal messengers as mediators of microvascular toe in the cerebral cortex. International Congress Series, 1235, 262–276. Hugdahl, K. (2001). Psychophysiology: The mind-body perspective. London: Harvard University Press. Iadecola, C. (2004). Neurovascular regulation in the normal brain and in Alzheimer’s disease. Nature Reviews Neuroscience, 5, 347–360. Jennings, J. R. (2003). Autoregulation of blood pressure and thought: Preliminary results of an application of brain imaging to psychosomatic medicine. Psychosomatic Medicine, 65, 384–395. Johnson, A., & Proctor, R. W. (2004). Attention: Theory and practice. Thousand Oaks: Sage Publications. Jorgensen, L. G., Perko, G., & Secher, N. H. (1992). Regional cerebralartery mean flow velocity and blood-flow during dynamic exercise in humans. Journal of Applied Physiology, 73, 1825–1830. Kolb, B., & Whishaw, I. Q. (2003). Fundamentals of human neuropsychology. New York: Freeman. Lacey, B. C., & Lacey, J. I. (1970). Some autonomic-central nervous system interrelationships. In P. Black (Ed.), Physiological correlations of emotion (pp. 205–227). New York: Academic Press. Larsen, F. S., Olsen, K. S., Hansen, B. A., Paulson, O. B., & Knudsen, G. M. (1994). Transcranial Doppler is valid for determination of the lower limit of cerebral blood flow autoregulation. Stroke, 25, 1985– 1988. Levy, M. N., & Pappano, A. J. (2007). Cardiovascular physiology. Philadelphia: Mosby Elsevier. Loewy, A. D. (1990). Central autonomic pathways. In A. D. Loewy & K. M. Spyer (Eds.), Central regulation of autonomic functions (pp. 88– 103). New York: Oxford University Press. Logothetis, N. K., Pauls, J., Augath, M., Trinath, T., & Oeltermann, A. (2001). Neurophysiological investigation of the basis of the fMRI signal. Nature, 412, 150–157. Mosso, A. (1881). U¨ber den Kreislauf des Blutes im menschlichen Gehirn. Leipzig: Von Veit. Newell, D. W., Aaslid, R., Lam, A., Mayberg, T. S., & Winn, H. R. (1994). Comparison of flow and velocity during dynamic autoregulation testing in humans. Stroke, 25, 793–797. Novak, V., Novak, P., Spies, J. M., & Low, P. A. (1998). Autoregulation of cerebral blood flow in orthostatic hypotension. Stroke, 29, 104– 111. Oldfield, R. C. (1971). The assessment and analysis of handedness: The Edinburgh inventory. Neuropsychologia, 9, 97–113. Pardo, J. V., Fox, P. T., & Raichle, M. E. (1991). Localisation of a human system for sustained attention by positron emission tomograpy. Nature, 349, 61–64. Paulson, O. B. (2002). Blood-brain barrier, brain metabolism and cerebral blood flow. European Neuropsychopharmacology, 12, 495–501. Paulson, O. B., Strandgaard, S., & Edvinsson, L. (1990). Cerebral autoregulation. Cerebrovascular and Brain Metabolism Reviews, 2, 161– 192. Paus, T., Zatorre, R. J., Hofle, N., Caramanos, Z., Gorman, J., Petrides, M., & Evans, A. C. (1997). Time-related changes in neural systems underlying attention and arousal during the performance of an auditory vigilance task. Journal of Cognitive Neuroscience, 9, 392–408.

S. Duschek et al. Porges, S. W. (1992). Autonomic regulation and attention. In B. A. Campbell, H. Hayne, & R. Richardson (Eds.), Attention and information processing in infants and adults (pp. 201–223). Hillsdale: Erlbaum. Posner, M. I., & Rafal, R. D. (1987). Cognitive theories of attention and the rehabilitation of attentional deficits. In M. J. Meier, A. L. Benton, & L. Diller (Eds.), Neuropsychological rehabilitation (pp. 182–201). Edinburgh: Churchill Livingston. Posner, M. I., & Petersen, S. E. (1990). The attention system of the human brain. Annual Review of Neuroscience, 13, 25–42. Reyes del Paso, G. A., Vila, J., & Garcia, A. (1994). Physiological significance of the defense response to intense auditory stimulation: A pharmacological blockade study. International Journal of Psychophysiology, 17, 181–187. Roy, C., & Sherrington, C. (1890). On the regulation of the blood supply of the brain. Journal of Physiology, 11, 85–108. Sa´ndor, P. (1999). Nervous control of the cerebrovascular system: Doubts and facts. Neurochemistry International, 35, 237–259. Sato, A., Sato, Y., & Uchida, S. (2001). Regulation of regional cerebral blood flow by cholinergic fibres originating in the basal forebrain. International Journal of Developmental Neuroscience, 19, 327–337. Schuepbach, D., Boeker, H., Duschek, S., & Hell, D. (2007). Rapid cerebral hemodynamic modulation during mental planning and movement execution: Evidence of time-locked relationship with complex behavior. Clinical Neurophysiology, 118, 2254–2262. Schuepbach, D., Huizinga, M., Duschek, D., Grimm, S., Boeker, H., & Hell, D. (2009). Rapid cerebral hemodynamic modulation during set shifting: Evidence of time-locked associations with cognitive control in females. Brain and Cognition, 71, 313–319. Sloan, R. P., Shapiro, P. A., Bagiella, E., Myers, M. M., & Gorman, J. M. (1999). Cardiac autonomic control buffers blood pressure variability responses to challenge: A psychophysiological model of coronary artery disease. Psychosomatic Medicine, 61, 58–68. Stegagno, L., Patritti, D., Duschek, S., Herbert, B., & Schandry, R. (2007). Cerebral blood flow in essential hypotension during emotional activation. Psychophysiology, 44, 226–232. Sturm, W., de Simone, A., Krause, B. J., Specht, K., Hesselmann, V., Radermacher, I., et al. (1999). Functional anatomy of intrinsic alertness: Evidence for a fronto-parietal-thalamic-brainstem network in the right hemisphere. Neuropsychologia, 37, 797–805. Szirmai, I., Amrein, I., Pa´lvo¨gyi, L., Debreczeni, R., & Kamondi, A. (2005). Correlation between blood flow velocity in the middle cerebral artery and EEG during cognitive effort. Cognitive Brain Research, 24, 33–40. Thayer, J., & Lane, R. D. (2009). Claude Bernard and the heart-brain connection: Further elaboration of a model of neurovisceral integration. Neuroscience and Biobehavioral Reviews, 33, 81–88. van Roon, A. M., Mulder, L. J. M., Althaus, M., & Mulder, G. (2004). Introducing a baroreflex model for studying cardiovascular effects of mental workload. Psychophysiology, 41, 961–981. Venables, P. H. (1991). Autonomic activity. Annals of the New York Academy of Sciences, 620, 191–207. Waldstein, S. R., Ryan, C. M., Manuck, S. B., Parkinson, D. K., & Bromet, E. J. (1991). Learning and memory function in men with untreated blood pressure elevation. Journal of Consulting and Clinical Psychology, 59, 513–517. Weber, P., Lutschg, J., & Fahnenstich, H. (2004). Attention-induced frontal brain activation measured by near-infrared spectroscopy. Pediatric Neurology, 31, 96–100. Wesseling, K. H., De Wit, B., Van der Hoeven, G. M. A., Van Goudoever, J., & Settels, J. J. (1995). Physiocal, calibrating finger vascular physiology for Finapres. Homeostasis, 36, 67–82. Zhang, R., Zuckerman, J. H., Giller, C. A., & Levine, B. D. (1998). Transfer function analysis of dynamic cerebral autoregulation in humans. American Journal of PhysiologyFHeart and Circulatory Physiology, 274, H233–H241.

(Received July 22, 2009; Accepted November 25, 2009)

Psychophysiology, 47 (2010), 1167–1171. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01022.x

BRIEF REPORT

Temporal stability of the error-related negativity (ERN) and post-error positivity (Pe): The role of number of trials

MICHAEL J. LARSON,a,b SCOTT A. BALDWIN,a DANIEL A. GOOD,a and JOSEPH E. FAIRa a

Department of Psychology, Brigham Young University, Provo, Utah, USA Neuroscience Center, Brigham Young University, Provo, Utah, USA

b

Abstract The error-related negativity (ERN) and post-error positivity (Pe) components of the event-related potential (ERP) are relatively stable over time. The current study further assessed the temporal reliability of ERN and Pe amplitudes for random samples of 2 to 14 trials per participant and the grand mean over a 2-week retest interval. In a replication of previous results, intraclass and zero-order correlations revealed moderate to good temporal stability for participants’ (N 5 20) grand mean ERN and Pe component amplitudes. Adding trials increased test–retest reliabilities; however, the temporal stability of ERN and Pe amplitudes with 14 or fewer trials were modest at best and considerably lower than that for the grand means. Overall, data support the temporal stability of grand-mean ERN and Pe amplitudes and suggest that more than 14 trials are needed to include in ERN and Pe averages for adequate test–retest reliability. Descriptors: Error negativity (Ne), Anterior cingulate, Reliability, Test-retest, Temporal stability

et al. (2010) observed adequate temporal stability of the ERN over short (20 min) and long (3 to 6 weeks) time periods. Olvet and Hajcak (2009a) report high internal consistency (split-half reliability) and test–retest reliability for the ERN and its correcttrial counterpart, the correct-trial negativity (CRN). Test–retest reliability estimates in this study were similar for individuals who committed a low number of errors (M 5 20) and a high number of errors (M 5 36). Olvet and Hajcak (2009b) also demonstrated moderate to high levels of internal consistency with as few as six and eight error trials for ERN and Pe component amplitudes, respectively. These findings were subsequently replicated across the life span (Pontifex et al., 2010). Although six to eight trials may produce adequate internal consistency for the ERN and Pe, other forms of reliability, such as test–retest reliability, may require increased numbers of trials due to more possible sources of error variation (Kaplan & Saccuzzo, 2008). Thus, the primary purpose of this study was to examine the test–retest reliability of the ERN and Pe components with increasing numbers of error trials. We also sought to replicate previous findings of good test–retest reliability of the ERN and Pe components across a 2-week interval.

The reliability of the error-related negativity (ERN) and posterror positivity (Pe) components of the scalp-recorded event-related potential (ERP) is currently the subject of considerable investigation. The ERN is a fronto-central negative-going deflection in the response-locked ERP that is larger following errors than correct trials and peaks within 100 ms after response (Falkenstein, Hohnsbein, Hoormann, & Banke, 1991; Gehring, Goss, Coles, Meyer, & Donchin, 1993). Evidence suggests that the ERN reflects the activity of a performance- and action-monitoring system when there is a mismatch between intended and produced responses or when competing response options are simultaneously activated (Falkenstein et al., 1991; Gehring et al., 1993; Holroyd & Coles, 2002; Yeung, Cohen, & Botvinick, 2004). The Pe is a positive deflection in the ERP that occurs between 100 and 400 ms following participant response and is more positive following error trials than correct trials (Falkenstein et al., 1991; Overbeek, Nieuwenhuis, & Ridderinkhof, 2005). Current theories suggest that the Pe is associated with signaling for post-error adjustments in behavior and the conscious recognition of errors, as Pe amplitudes are decreased when individuals are unaware of performance errors or neurologic deficits (Hajcak, McDonald, & Simons, 2003; Larson & Perlstein, 2009; Nieuwenhuis, Ridderinkhof, Blom, Band, & Kok, 2001). Recent research indicates moderate to good reliability coefficients for the ERN and Pe components. For example, Segalowitz

Method Participants Twenty-eight individuals between the ages of 19 and 29 initially enrolled in the study. Seven participants were excluded because they committed fewer than 14 errors (see Olvet & Hajcak, 2009b)

Address correspondence to: Michael J. Larson, Department of Psychology, Brigham Young University, 244 TLRB, Provo, UT 84602, USA. E-mail: [email protected] 1167

1168 and 1 participant failed to return for the retest session. Thus, final enrollment included 20 healthy, right-handed individuals (eight female), with a mean (!SD) age of 22.35 (2.48) years. Exclusion criteria included history of psychiatric disorder, psychoactive medication use, substance abuse or dependence, neurological disorders, or uncorrected visual impairment. Participants were compensated for study participation. Study procedures were approved by the Institutional Review Board at Brigham Young University. Experimental Task Participants performed a modified version of the Eriksen Flanker Task (Eriksen & Eriksen, 1974) wherein congruent (e.g.,ooooo) and incongruent (e.g.,oo4oo) arrow stimuli were presented centered on a 17-in. computer monitor "20 in. from the participant’s head. Participants were instructed to respond as quickly and accurately as possible with an indexfinger button press if the middle arrow pointed to the left and a middle-finger button press if the middle arrow pointed to the right. Flanker stimuli were presented for 100 ms prior to the onset of the target stimulus, which remained on the screen for 600 ms. To decrease expectancy effects, the intertrial interval (ITI) varied randomly between 800 and 1200 ms, with a mean ITI of 1000 ms. Three blocks of 300 trials (900 total trials) were presented; the distribution of congruent and incongruent trials was equal (450 trials each). Following task completion, a second session was scheduled for approximately 2 weeks later (average of 14.55 [!1.88] days between sessions). Electrophysiological Data Recording and Reduction Electroencephalogram (EEG) was recorded from 128 scalp sites using a geodesic sensor net and Electrical Geodesics, Inc. (EGI; Eugene, OR) amplifier system (20K nominal gain, bandpass 5 0.10–100 Hz). During recording, EEG was referenced to the vertex electrode and digitized continuously at 250 Hz with a 24-bit analog-to-digital converter. Impedances were maintained below 50 kO. Data were average-re-referenced off-line and digitally low-pass filtered at 30 Hz. Eye movement and blink artifacts were corrected using the algorithm described by Gratton, Coles, and Donchin (1983). Following Olvet and Hajcak (2009b), individual-subject response-locked averages were derived spanning 400 ms prior to response and 800 ms following response. Epochs were baseline corrected from # 400 to # 200 ms. Error-trial amplitudes for the ERN were extracted as the average activity from 0 to 100 ms at electrode site FCz. Latency measurements for the ERN were indexed at FCz as the peak negative-going amplitude within the 0–100-ms window. Amplitudes for the Pe were extracted as the average activity from 200 to 400 ms postresponse at electrode Pz. Given the tonic nature of the Pe, no latency times were calculated. Statistical Analysis Median response times (RT), mean error rates, and ERP component amplitude and latency data were analyzed using repeated measures analysis of variance (ANOVA) including the factors congruency (congruent, incongruent) and time (Time 1, Time 2) for RT and error-rate data and the factors accuracy (correct, error) and time for ERP data. Paired-samples t tests were used to decompose significant main effects and interactions. For reliability analyses we followed the procedures outlined by Lew, Gray, and Poole (2007) and Olvet and Hajcak (2009a, 2009b);

M.J. Larson et al. test–retest reliability both of the grand mean and as a function of increasing number of trials was assessed using the single measure intraclass correlation (ICC) with a one-way random-effects model and zero-order correlations. With two time points, the ICC can range from # 1.0 to 1.0. Acceptable values of the ICC vary with different authors. Anastasi (1998) indicated that values of the ICC at or above .60 are adequately reliable, whereas others indicated that ICCs o.40 are poor, ICCs between .41 and .59 are moderate, ICCs between .60 and .74 are good, and ICCs above .75 are excellent (Cicchetti, 2001; Cicchetti & Sparrow, 1981). For zero-order correlations, values above .50 are generally considered reliable for experimental research based on groups (Helmstadter, 1964; Segalowitz et al., 2010). To explore the effect of the number of error trials on the reliability results, we compared the temporal reliability of errortrial ERN amplitude, error-trial ERN latency, and error-trial Pe amplitude separately for random samples of 2 to 14 trials per participant (i.e., we sampled two trials per participant, then three trials per participant, and so on). We used 14 as the maximum number of trials in order to provide a direct comparison with Olvet and Hajcak (2009b). To reduce the impact of sampling error, we replicated the random draw 2,500 times for each number of error trials and computed the mean reliability across the 2,500 replications. Results Response Times and Error Rates Response time (in milliseconds) for congruent and incongruent trials at Time 1 were 363.20 (25.64) and 431.23 (26.87), respectively; RT data for congruent and incongruent trials at Time 2 were 348.48 (25.35) and 406.75 (28.47). A Congruency $ Time ANOVA on RTs revealed the expected main effect of congruency, F(1,19) 5 507.12, po.001, Z2 5 .96, with significantly longer RTs to incongruent relative to congruent trials at both Time 1, t(19) 5 23.11, po.001, and Time 2, t(19) 5 18.99, po.001. A significant main effect of time, F(1,19) 5 38.75, po.001, Z2 5 .67, showed the effect of practice on performance, with decreased (i.e., faster) RTs from Time 1 to Time 2. There was also a significant Congruency $ Time interaction, F(1,19) 5 20.22, po.001, Z2 5 .52, with faster RTs from Time 1 to Time 2 for both incongruent, t(19) 5 7.79, po.001, and congruent trials, t(19) 5 4.19, po.001. Error rates (percent errors) for congruent and incongruent trials at Time 1 were .03 (.02) and .12 (.06), respectively, and .02 (.03) and .09 (.05) for congruent and incongruent trials at Time 2. Analyses of error rates revealed a significant main effect of congruency, F(1,19) 5 49.62, po.001, Z2 5 .72, indicating that participants made more errors to incongruent than congruent trials at both Time 1, t(19) 5 6.29, po.001, and Time 2, t(19) 5 6.60, po.001. The main effect of time and the Congruency $ Time interactions were not statistically reliable, Fso2.9, ps4.11, indicating no overall differences in error rates from Time 1 to Time 2. Event-Related Potential Data Response-locked correct-trial and error-trial waveforms at Time 1 contained an average (!SD) of 728.35 (58.62) trials and 52.45 (29.85) trials, respectively. For Time 2, there was an average of 724.50 (102.23) correct trials and 42.85 (24.42) error trials contained in the averages. The number of error trials in the grand means ranged from 14 to 120. An Accuracy $ Time ANOVA on the number of trials included showed no significant main

ERN test–retest reliability

1169

Figure 1. Grand mean ERP waveforms depicting response-locked correct- and error-related activity for the ERN at electrode FCz and the Pe at electrode Pz at Time 1 and Time 2 as well as a direct comparison of Time 1 and Time 2 error trials.

effect of time, F(1,19) 5 0.37, p4.55, Z2 5 .02, as well as no significant Accuracy ! Time interaction, F(1,19) 5 0.06, p4.81, Z2 5 .003, indicating the number of trials in the grand means did not differ between time points. Average ERP waveforms for correct and error trials reflecting the ERN and Pe at Time 1 and Time 2 are shown in Figure 1. Supplemental analyses showed no relationship between number of error trials and ERP amplitude and latency data. Amplitude measures. An Accuracy ! Time ANOVA on response-locked ERPs for the ERN yielded a significant main effect of accuracy, F(1,19) 5 48.04, po.001, Z2 5 .72, with a significant error relative to correct ERN at both Time 1, t(19) 5 5.21, po.001, and Time 2, t(19) 5 7.27, po.001. The main effect of time was not significant, F(1,19) 5 2.76, p4.12, Z2 5 .13, indicating that the overall magnitude of the ERP amplitudes did not significantly differ between time points. The Accuracy ! Time interaction was also not significant, F(1,19) 5 3.06, p4.09, Z2 5 .14. Paired-samples t tests showed no significant differences between sessions for ERN amplitudes, t(19) 5 " .24, p4.81, but a significant difference between sessions for correct-trial amplitudes, t(19) 5 " 3.66, po.002. Results of an Accuracy ! Time ANOVA on ERN latencies revealed no significant main effects or interactions, Fso.85, ps4.36, indicating there were no differences in ERP latency between correct and error trials and across time points. An Accuracy ! Time ANOVA on Pe amplitudes was similar to that for the ERN. There was a main effect of accuracy, F(1,19) 5 7.62, po.01, Z2 5 .29, with increased amplitude Pe for error trials relative to correct trials at Time 1, t(19) 5 2.65, po.01, and Time 2, t(19) 5 2.20, po.04. There was not a significant main effect of time or a significant Accuracy ! Time interaction, Fso1.23, ps4.28, indicating the amplitude of the Pe generally did not differ between time points. Temporal stability of the grand means. Most important to the current study is the temporal stability of error-related ERP components across time. Analyses indicated statistically reliable temporal stability for grand mean ERN amplitudes, ICC 5 .66,

po.009, and CRN amplitudes, ICC 5 .75, po.001. Zero-order correlations on the grand means between time points supported these results, with significant correlations for the ERN, r 5 .49, po.03, and the CRN, r 5 .72, po.001.1 Grand mean latencies for the ERN showed low test–retest reliability for error trials, ICC 5 .33, p4.18, but good reliability for correct trials, ICC 5 .63, po.02. Zero-order correlations were not significant between Time 1 and Time 2 for ERN latencies, r 5 .25, p4.29; the correlation for CRN latency was significant, r 5 .46, po.04. Analysis of the temporal stability of Pe amplitudes showed moderate retest reliability for error trials, ICC 5 .48, p4.08, and good reliability for correct trials, ICC 5 .68, po.006. Zero-order correlations supported the results of the intraclass correlations, with modest reliability of the error-trial Pe amplitude across time points, r 5 .32, p4.17, but adequate reliability for correct trials, r 5 .59, po.007. Temporal stability with increasing error trials. Mean intraclass correlations and zero-order correlations with 80% confidence intervals across the 2,500 replications for increasing error trials are presented in Figure 2. The results for intraclass correlations and zero-order correlations for ERN amplitude, ERN latency, and Pe amplitude are similar. Not surprisingly, reliability increased with increasing number of error trials up to 14, but never reached the reliability of the grand mean. The benefit of each additional trial is small but does not level off within the scope of 1

We recalculated the grand mean ICCs for fronto-central and centroparietal regions of interest (ROIs) based on the scalp distributions of the current data. Amplitudes of the ERN were averaged across seven frontocentral electrode sites (5, 6, 7, 12, 13, 106, and 112; see Larson, Fair, Good, & Baldwin, 2010, for montage) and seven centro-parietal electrode sites for the Pe (6, 7, 13, 31, 80, 106, and 112). Grand mean amplitudes were more stable over time for the ROIs for the ERN, ICC 5 .77, po.001, and CRN, ICC 5 .82, po.001, relative to amplitudes at FCz for the ERN, ICC 5 .66, po.009, and CRN, ICC 5 .75, po.001. Similarly, ROI grand mean amplitudes were more stable for both error-trial Pe, ICC 5 .79, po.001, and correct-trial Pe amplitudes, ICC 5 .91, po.001, relative to amplitudes for the Pe at site Pz for error trials, ICC 5 .48, po.08, and correct trials, ICC 5 .68, po.006.

1170

M.J. Larson et al.

Figure 2. Average intraclass and zero-order correlations and 80% confidence intervals following 2,500 random samples for the ERN at electrode FCz and Pe at electrode Pz with increasing number of errors and the grand mean (GM).

our sampling scheme. Consequently, we suspect that reliability would continue to increase if we were able to increase the number of error trials beyond 14, although there would eventually be diminishing returns. Importantly, reliability levels for error-trial ERN amplitude, ERN latency, and Pe amplitude do not reach acceptable levels until all error trials are included in the grand means.

Discussion Test–retest reliability for the grand means of both ERN and Pe amplitudes were reliable across time points despite considerable practice effects on the behavioral data leading to faster RTs and a trend toward decreased error rates over time. Changes in behavioral performance are likely due to the effect of practice on behavioral performance; however, error-trial amplitudes for the ERN and Pe did not significantly differ between Time 1 and Time 2, indicating that there was not an attenuating effect of practice on error-related amplitudes across multiple sessions (cf. Schrijvers et al., 2009). Correct-trial ERP amplitudes were more reliable than error-trial ERP amplitudes; this is likely due to the greater number of trials included in all correct-trial averages. We note, however, that overall CRN amplitudes differed between sessions, whereas those for the ERN did not. Measurements of ERN latency were not as reliable as those for ERN amplitudes; however, there was little difference in ERN latency values between sessions. Variability in ERN latency is consistent with previous research and may be due to inconsistency in participant response times and component processing times (see Olvet & Hajcak, 2009a). We also examined the test–retest reliability of the ERN and Pe components of the ERP with increasing numbers of trials. Our findings indicate that the temporal stability of ERN and Pe amplitudes improved with each error trial included in the averages, but that adequate levels of temporal stability were not reached with up to 14 errors. In contrast, analysis of the grand means for

ERN and Pe amplitudes showed moderate to good temporal stability over a 2-week test–retest interval. These results indicate that more than 14 trials are needed to achieve adequate levels of test–retest reliability, but that grand means with an average of 42 or more error trials are temporally stable. Current results replicate previous findings of moderate to good temporal stability for grand mean ERN and Pe component amplitudes over time (Olvet & Hajcak, 2009a; Segalowitz et al., 2010). Our results also augment previous findings that six to eight error trials are enough to achieve adequate internal consistency for ERN and Pe amplitudes (Olvet & Hajcak, 2009b; Pontifex et al., 2010). That is, current data, using the mean of 2,500 samples for measures of test–retest reliability to reduce the effect of sampling error, show that more than 14 error trials are required for ERN and Pe amplitude averages to achieve adequate test–retest reliability. These data do not directly specify the number of trials necessary to achieve reliable temporal stability for these components; however, previous findings indicate moderate to good test–retest reliability for ERN and Pe amplitudes in a sample of individuals that made an average of 20 errors (Olvet & Hajcak, 2009a). Some degree of caution should be taken in interpretation and comparison of findings regarding numbers of trials needed for reliability. Studies differ in EEG acquisition characteristics, task characteristics, and the nature of the sample. For example, data for the current study were collected using a high-impedance EGI system, whereas a previous study examining numbers of trials used BioSemi active electrodes (Olvet & Hajcak, 2009b). Similarly, participants may be more or less motivated to complete the task depending on levels of compensation or examiner characteristics. Thus, whereas temporal stability of the ERN is shown to be good in multiple studies, data should be interpreted within the context of specific laboratory and study procedures. Test–retest amplitude and latency results are consistent with test–retest reliabilities reported in several studies of different ERP components in several different modalities. For example, Lew et al. (2007) showed ICCs ranging from .60 to .80 for the

ERN test–retest reliability

1171

amplitudes of the N1, mismatch negativity, P3, and N4 components of the auditory ERP in healthy individuals. Similarly, test– retest reliabilities are generally in the moderate to good ranges for neuropsychological measures, such as the Rey Auditory Verbal Learning Test, the Controlled Oral Word Association Test, and the Rey-Osterrieth Complex Figure Task, administered to healthy individuals (Strauss, Sherman, & Spreen, 2006). Thus, the temporal stability of electrophysiological measures of error processing is generally consistent with commonly used physiological and cognitive outcome measures. Our results have at least four important implications for future research using ERN and Pe amplitudes. First, and most importantly, researchers should ensure that an adequate number of trials are included in grand means of the ERN to achieve adequate temporal stability. Second, the good temporal stability of ERN and Pe amplitudes allows for the possibility of a physiological measure of change that could provide insight into the neural mechanisms underlying treatment-related changes. Third,

current treatment-related studies of the ERN show a wide range of changes pre- to posttreatment, with no changes in a study of pediatric obsessive-compulsive disorder (Hajcak, Franklin, Foa, & Simons, 2008), some indications for a link with symptom reduction in a study of individuals with depression (Schrijvers et al., 2009), and a clear relationship with 6 weeks of antipsychotic treatment in individuals with schizophrenia (Bates, Liddle, Kiehl, & Ngan, 2004). Findings of adequate temporal stability of the ERN indicate that the variation in treatment-related findings is not due solely to error and suggests the need for future studies. Fourth, findings may help elucidate the role of the ERN as an endophenotype for psychopathology by allowing confidence in multiple measures of ERN amplitude to determine if state-related changes are present in individuals with psychopathology (see Olvet & Hajcak, 2008), although the aforementioned variability in state-related ERN amplitudes associated with treatment indicate a need for considerable research in this regard.

REFERENCES Anastasi, A. (1998). Psychological testing (6th ed). New York: Macmillan. Bates, A. T., Liddle, P. F., Kiehl, K. A., & Ngan, E. T. (2004). State dependent changes in error monitoring in schizophrenia. Journal of Psychiatric Research, 38, 347–356. Cicchetti, D. V. (2001). The precision of reliability and validity estimates re-visited: Distinguishing between clinical and statistical significance of sample size requirements. Journal of Clinical and Experimental Neuropsychology, 23, 695–700. Cicchetti, D. V., & Sparrow, S. A. (1981). Developing criteria for establishing interrater reliability of specific items: Applications to assessment of adaptive behavior. American Journal of Mental Deficiency, 86, 127–137. Eriksen, B. A., & Eriksen, C. W. (1974). Effects of noise letters upon the identification of a target letter in a non-search task. Perception & Psychophysics, 16, 143–149. Falkenstein, M., Hohnsbein, J., Hoormann, J., & Banke, L. (1991). Effects of crossmodal divided attention on late ERP components. II. Error processing in choice reaction tasks. Electroencephalography and Clinical Neurophysiology, 78, 447–455. Gehring, W. J., Goss, B., Coles, M. G. H., Meyer, D. E., & Donchin, E. (1993). A neural system for error detection and compensation. Psychological Science, 4, 385–390. Gratton, G., Coles, M. G., & Donchin, E. (1983). A new method for offline removal of ocular artifact. Electroencephalography and Clinical Neurophysiology, 55, 468–484. Hajcak, G., Franklin, M. E., Foa, E. B., & Simons, R. F. (2008). Increased error-related brain activity in pediatric obsessive-compulsive disorder before and after treatment. American Journal of Psychiatry, 165, 116–123. Hajcak, G., McDonald, N., & Simons, R. F. (2003). To err is autonomic: Error-related brain potentials, ANS activity, and post-error compensatory behavior. Psychophysiology, 40, 895–903. Helmstadter, G. C. (1964). Principles of psychological measurement. New York: Appleton-Century Crofts. Holroyd, C. B., & Coles, M. G. H. (2002). The neural basis of human error processing: Reinforcement learning, dopamine, and the errorrelated negativity. Psychological Review, 109, 679–709. Kaplan, R. M., & Saccuzzo, D. P. (2008). Psychological testing: Principles, applications and issues (7th ed). Belmont, CA: Wadsworth. Larson, M. J., Fair, J. E., Good, D. A., & Baldwin, S. A. (2010). Empathy and error processing. Psychophysiology, 47, 415–424.

Larson, M. J., & Perlstein, W. M. (2009). Awareness of deficits and error processing after traumatic brain injury. NeuroReport, 20, 1486–1490. Lew, H. L., Gray, M., & Poole, J. H. (2007). Temporal stability of auditory event-related potentials in healthy individuals and patients with traumatic brain injury. Journal of Clinical Neurophysiology, 24, 392–397. Nieuwenhuis, S., Ridderinkhof, K. R., Blom, J., Band, G. P., & Kok, A. (2001). Error-related brain potentials are differentially related to awareness of response errors: Evidence from an antisaccade task. Journal of Psychophysiology, 19, 319–329. Olvet, D. M., & Hajcak, G. (2008). The error-related negativity (ERN) and psychopathology: Toward an endophenotype. Clinical Psychology Review, 28, 1342–1354. Olvet, D. M., & Hajcak, G. (2009a). Reliability of error-related brain activity. Brain Research, 1284, 89–99. Olvet, D. M., & Hajcak, G. (2009b). The stability of error-related brain activity with increasing trials. Psychophysiology, 46, 957–961. Overbeek, T. J. M., Nieuwenhuis, S., & Ridderinkhof, K. R. (2005). Dissociable components of error processing: On the functional signficance of the Pe vis-a`-vis the ERN/Ne. Journal of Psychophysiology, 19, 319–329. Pontifex, M. B., Scudder, M. R., Brown, M. L., O’Leary, K. C., Wu, C. T., Themanson, J. R., et al. (2010). On the number of trials necessary for stabilization of error-related brain activity across the lifespan. Psychophysiology (in press). Schrijvers, D., De Bruijn, E. R. A., Maas, Y. J., Vancoillie, P., Hulstijn, W., & Sabbe, B. G. C. (2009). Action monitoring and depressive symptom reduction in major depressive disorder. International Journal of Psychophysiology, 71, 218–224. Segalowitz, S. J., Santesso, D. L., Murphy, T. I., Homan, D., Chantziantoniou, D. K., & Khan, S. (2010). Retest reliability of medial frontal negativities during performance monitoring. Psychophysiology (in press). Strauss, S., Sherman, E. M. S., & Spreen, O. (Eds.). (2006). A compendium of neuropsychological tests: Administration, norms, and commentary (3rd ed). New York: Oxford University Press. Yeung, N., Cohen, J. D., & Botvinick, M. M. (2004). The neural basis of error detection: Conflict monitoring and the error-related negativity. Psychological Review, 111, 931–959. (Received May 15, 2009; Accepted November 25, 2009)

Psychophysiology, 47 (2010), 1172–1175. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01023.x

BRIEF REPORT

Further evidence for an association between self-reported health and cardiovascular as well as cortisol reactions to acute psychological stress

SUSANNE R. DE ROOIJ and TESSA J. ROSEBOOM Department of Clinical Epidemiology and Biostatistics, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands

Abstract In a recent study, the association between cardiovascular reactions to acute psychological stress and self-reported health was examined. Participants with excellent or good self-reported health exhibited higher cardiovascular reactivity than those who reported fair or poor health. We investigated this association in a population-based cohort of whom 725 men and women, aged 55–60 years, participated in a standardized psychological stress test. We measured continuous blood pressure and heart rate as well as cortisol reactivity. Good subjective health was associated with higher cardiovascular and cortisol reactions to psychological stress. Results of the present study confirm those of the previously reported study showing that greater cardiovascular reactivity may not always be associated with negative health outcomes. In addition, the same holds for cortisol reactivity. Descriptors: Self-reported health, Acute psychological stress, Cardiovascular reactivity, Cortisol reactivity

were measured. Several measures assessing subjective health were administered, including the same single item measure as used by Phillips et al. (2009), the COOP/WONCA charts, and the SF-36. We used these data to investigate the association between self-reported health and cardiovascular and hypothalamicpituitary-adrenal (HPA) axis reactivity to acute psychological stress in our cohort.

Self-reported health is a strong predictor of future mortality independent of medical, behavioral, or psychosocial risk factors (Idler & Benyamini, 1997). Psychological distress has been suggested as an explanatory variable (Larsson, Hemmingsson, Allebeck, & Lundberg, 2002). Recently, Phillips, Der, and Carroll (2009) reported an association between subjective health and cardiovascular reactions to acute psychological stress. Subjects participated in the paced auditory serial addition test, a mental arithmetic stress test. Surprisingly, it was shown that those who reported excellent or good health had significantly higher systolic blood pressure (SBP), diastolic blood pressure (DBP), and heart rate (HR) stress responses compared to those with fair or poor self-reported health (although HR differences were no longer statistically significant after adjustment for confounders). The study was hampered by a number of limitations though. BP and HR reactivity were assessed by single measurements with a semi-automated blood pressure device. Also, a specific measure of task engagement was missing, and self-reported health was evaluated with a single-item measure. In the Dutch Famine Birth Cohort, 725 participants were subjected to a psychological stress protocol during which cardiovascular reactivity as well as cortisol and perceived stress

Methods Participants and Selection Participants were selected from the Dutch Famine Birth Cohort, which consists of 2,414 men and women born in Amsterdam, the Netherlands, around the time of the Dutch famine. Cohort members were eligible for participation (n 5 1,423) in this study if they lived in the Netherlands on September 1, 2002, and if their address was known to the investigators. The study was approved by the local Medical Ethics Committee. All participants gave written informed consent. We measured socioeconomic status (SES) according to ISEI-92, a numeric scale based on the person’s or his or her partner’s occupation, whichever status was higher. A detailed description of other basal study parameters is given elsewhere (de Rooij et al., 2006).

We thank the participants for their willing cooperation. Address correspondence to: Susanne de Rooij, Ph.D., Department of Clinical Epidemiology and Biostatistics, J1B-210.1, Academic Medical Centre, University of Amsterdam, Meibergdreef 9, PO box 22660, 1100 DD Amsterdam, The Netherlands. E-mail: [email protected]

Self-Reported Health Self-reported health was based on the answer of the participant to the question: How do you rate your health in general? 1172

Self-reported health and stress reactivity (1 5 excellent, 2 5 very good, 3 5 good, 4 5 fair, 5 5 poor). Additionally, the COOP/WONCA charts and the SF-36 were administered. The COOP/WONCA charts (from the Dartmouth Primary Care Cooperative Information Project [COOP Project]/ World Organization of National Colleges, Academies, and Academic Associations of General Practices/Family Physicians) cover several dimensions, including physical fitness, feelings, daily activities, social activities, and pain (Nelson et al., 1987). The five possible responses per question are illustrated by a drawing. Higher scores indicate lower self-perceived health. The SF 36 is the gold standard method for measuring quality of life (Aaronson et al., 1998). Eight main scales are derived from the 36 questions, which can be transformed into two general scores: the physical component summary (PCS) score and the mental component summary (MCS) score. Higher scores indicate better self-perceived health. Psychological Stress Protocol The protocol started with a 20-min baseline period followed by a computerized Stroop (color–word conflict) test, a mirror-tracing test, and a public speech task. Each task took 5 min, with 6 min in between and 30 min of recovery at the end. To measure perceived stress, we asked the participant after each stress task to indicate how committed he or she had felt to performing the stress task, how difficult they found the task, how well they think they had performed, how much they had felt to be in control, how stressed they had felt, and how relaxed they had felt. The last two questions were also asked at baseline and at recovery. A score had to be given on a 7-point scale ranging from 1 (not at all) to 7 (very much). A full description of the three stress tests is given elsewhere (de Rooij et al., 2006). Continuous BP and HR recordings were made using a Finometer or a Portapres Model-2 (Finapres Medical Systems, Amsterdam, the Netherlands). We designated six periods of 5 min each as measuring periods, defined as follows: baseline (15 min into the baseline period), Stroop, mirror-tracing, speech test, Recovery 1 (5 min after completing the speech test), and Recovery 2 (25 min after completing the speech test). We calculated mean SBP, DBP, and HR for each period. Saliva samples were collected using Salivettes (Sarstedt, Rommelsdorf, Germany) at seven time points during the protocol: at 5 and 20 min in the baseline period; at 6 min after completion of the Stroop test and the mirror-drawing test; and at 10, 20, and 30 min after completion of the speech test. Statistical Analyses We took the highest SBP, DBP, and HR values during the 5-min measuring periods as maximum values; the highest maximum value of the six measuring periods was assigned as the peak value during the stress protocol. The highest of the seven cortisol values was assigned as the peak response. The increase from baseline to this peak value was designated as stress reactivity. Baseline cortisol was calculated as the mean of the first and the second cortisol concentrations measured during the baseline period. Perceived stress questionnaire variables were calculated as the sum of the three or five scores. In the analyses, we used the single-item self-reported health variable in a continuous manner including all participants. In line with the publication of Phillips et al. (2009), the data were negatively skewed, with a very small proportion of participants (n 5 9) reporting poor health. For comparison with Phillips et al., we also created a binary variable in which the excellent,

1173 very good, and good categories were taken together as were the fair and poor categories (0 5 good, 1 5 poor; reported in the tables). We did the same for COOP/WONCA, PCS, and MCS, constructing binary variables with the 0 category including those with scores in the lowest quartile and the 1 category including those with scores in the highest quartile. We used linear and logistic regression to compare general characteristics, stress responses, and perceived stress scores between those with good and poor self-reported health (Tables 1 and 2) and associations between stress responses and self-reported health as continuous variable (see Results). Cortisol concentrations were ranked and other variables with a skewed distribution log-transformed to be able to apply regression analysis. In additional analyses we adjusted for the following potential confounders: age, sex, SES, body mass index (BMI), use of antihypertensive medication, and baseline SBP, DBP, HR, and cortisol when appropriate. Variables with skewed distributions are reported as medians and interquartile ranges (IQR). Table 1. General, Clinical, and Stress Test Characteristics According to Status of Self-Reported Health Based on a One-Item Measure Self-reported health

n General characteristics Men (%) Age (years) SESa Use of antihypertensive medication (%) Clinical characteristics Body mass index (kg/m2) Systolic blood pressure (mmHg) Diastolic blood pressure (mmHg) Heart rate (bpm) Stress test characteristics Baseline cortisol (nmol/l)a Peak cortisol (nmol/l)a Cortisol reactivity (nmol/l)a Basal SBP (mmHg) Peak SBP (mmHg) SBP reactivity (mmHg) Basal DBP (mmHg) Peak DBP (mmHg) DBP reactivity (mmHg) Basal HR (bpm) Peak HR (bpm) HR reactivity (bpm) Perceived stress Commitmenta Difficultya Performance Control Stress Relaxeda

Poor

Good

All (SD)

164

561

725

n

42.1 58.3 46 38

48.8 58.3 51n 19n

47.3 58.3 (0.9) 51 (19) 23

725 725 717 725

29.0 138

27.6n 136

27.9 (6.0) 136 (18)

724 722

81

81

81 (10)

722

71

69

70 (10)

722

3.5 4.9 0.9 127.4 171.7 44.9 64.5 84.2 19.9 75.8 85.8 10.2 15 14 9 10 12 10

n

3.8 3.7 (2.6) 647 6.4nn 6.0 (5.6) 691 2.2nn 1.9 (3.6) 645 128.1 127.9 (20.9) 713 176.4 175.4 (28.8) 721 48.4 47.6 (20.7) 713 66.4 66.0 (12.0) 713 87.2 (14.4) 721 88.1nn 21.4 (9.1) 713 21.8nn 73.8nnn 74.2 (10.5) 712 86.1 86.0 (14.2) 721 11.8 (9.5) 712 12.3n 15 15 9 10 11nnn 12nnn

15 (6) 15 (5) 9 (3) 10 (4) 11 (4) 11 (5)

697 697 698 697 698 697

Note: Data are given as means and standard deviations. a Data are given as medians and interquartile ranges. n Statistically significant difference compared to those with poor self-reported health, based on logistic or linear regression analysis (p ! .05). nn Adjusted for sex, age, SES, use of anti-hypertensive medication, BMI, and baseline (p ! .05). nnn Adjusted for sex, age, SES, use of anti-hypertensive medication and BMI (p ! .05).

1174

S.R. de Rooij and T.J. Roseboom

Table 2. Stress Reactivity According to Status of Self-Reported Health Based on the COOP/WONCA and the SF-36 Stress reactivity SBP (mmHg) COOP/WONCA Poor 45.5 (20.2) Good 48.8 (19.9) SF-36: PCS Poor 48.3 (22.1) Good 48.5 (16.9) SF-36: MCS Poor 45.0 (18.2) Good 50.9 (22.0)nn

DBP (mmHg)

HR (bpm)

Cortisol (nmol/l)a

n

20.7 (9.0) 22.0 (9.3)

11.5 (10.3) 13.0 (9.1)

1.1 (3.0) 2.4 (4.7)nn

198 223

21.5 (8.9) 21.8 (7.2)

11.3 (10.3) 12.6 (8.9)

1.3 (2.8) 2.3 (4.5)n

169 169

20.5 (8.2) 21.7 (9.4)

10.7 (8.7) 12.5 (9.9)

1.3 (3.3) 2.5 (3.9)nn

169 169

Note: Data are given as means and standard deviations. PCS: Physical Component Summary; MCS: Mental Component Summary. a Data are given as medians and interquartile ranges. n Statistically significant difference compared to those with poor selfreported health, based on linear regression analysis (p " .05). nn Adjusted for sex, age, SES, use of anti-hypertensive medication, BMI, and baseline (p " .05).

Results Population Characteristics Seven hundred twenty-five participants completed the stress protocol. Data on cardiovascular reactivity were available for 721 subjects, while cortisol data were available for 694 subjects. Self-Reported Health A total of 68 participants (9.4%) considered their health excellent, 77 (10.6%) considered it very good, 416 (57.4%) considered it good, 155 (21.4%) considered it fair, and 9 (1.2%) considered it poor. The median score on the COOP/WONCA was 14 (IQR 5), the median score on the PCS of the SF-36 was 50 (IQR 13), and the median score on the MCS of the SF-36 was 54 (IQR 10). Table 1 shows that the SES of those reporting excellent/good health was higher than the SES of those reporting fair/poor health (po.01). Also, use of antihypertensive medication (po.01), BMI (po.01), and HR in rest condition (p 5 .02) were significantly lower in those reporting excellent or good health compared to those reporting fair or poor health. Stress Reactivity SBP, DBP, HR, and cortisol values increased in response to all of the three psychological stress tests. In a majority of the participants, SBP, DBP, and HR reactivity peaked during the speech task. Maximum reactivity was 47.6 mmHg (95% CI: 46.1–49.1) for SBP, 21.4 mmHg (20.7–22.1) for DBP, and 11.8 bpm (11.1– 12.5) for HR. The mean cortisol response peaked during the first recovery period after the speech test. Maximum reactivity was 1.9 nmol/l (1.5–2.3). Men showed larger cortisol reactivity than women (po.01). SES was positively associated with cortisol (p 5 .05), SBP (po.01), and HR reactivity (po.01). Both use of antihypertensive medication (p 5 .04) and BMI (po.01) were negatively associated with HR reactivity (po.01). BMI was also negatively associated with cortisol reactivity (po.01). Commitment score was positively associated with SBP, DBP, and HR stress reactivity (all p 5 .02). Perceived stress score was positively associated with HR reactivity (p 5 .03).

Self-Reported Health and Stress Reactivity Cortisol stress reactivity significantly increased with a decreasing score (toward better self-reported health) on the one-item selfreported health measure (B 5 37 [95% CI: 20–53], po.01 derived from regression analysis on ranked scores). SBP reactivity to stress increased by 2.7 mmHg (0.9–4.4, po.01) per point decrease in self-reported health, DBP reactivity by 1.2 mmHg (0.5–2.0, po.01), and HR reactivity by 1.2 bpm (0.4–2.0, po.01). After adjusting for possible confounders the effect sizes of the associations became somewhat smaller in most cases but remained significant (B 5 30 [13–47], po.01 for cortisol; 2.5 mmHg [0.7–4.4], po.01 for SBP; 1.3 mmHg [0.4–2.1], po.01 for DBP). However, in the case of HR reactivity, the effect size became much smaller and was no longer statistically significant (0.6 bpm [! 0.2 to 1.5], p 5 .16). Perceived stress and relaxed scores significantly decreased (0.6 points per point [0.3–1.0], po.01) and increased (0.7 points per point [0.4–1.0], po.01) with a decreasing perceptive health score (toward better self-reported health). Commitment did not differ (p 5 .57). Adjusting for commitment, perceived stress, or relaxedness did not alter the associations between self-reported health and stress reactivity. Analyses of the other self-reported health measures showed that cortisol reactivity significantly increased with decreasing COOP/WONCA scores (toward better self-reported health; B 5 6 per point [2–10], po.01) and increased with increasing SF36 scores (toward better self-reported health; B 5 2 per point [1–4], po.01 for PCS; B 5 3 [1–4], po.01 for MCS). In addition, SBP reactivity to stress significantly increased by 0.2 mmHg (0.1–0.4, po.01) per point increase on the MCS and HR reactivity by 0.1 bpm (0.0–0.2, p 5 .05). DBP reactivity increased with a trend toward statistical significance by 0.1 mmHg (0.0–0.1, p 5 .09). After adjusting for possible confounders the effect sizes and p values of the associations remained roughly the same, except in the case of the association between HR reactivity and MCS scores (B 5 0.1 bpm [! 0.0 to 0.1, p 5 .11) and cortisol reactivity and COOP/WONCA scores (B 5 4 per point [0–8], p 5 .08) and PCS scores (B 5 1 per point [0–3], p 5 .012).

Discussion Results showed that reporting good health was associated with higher cardiovascular and cortisol responses to a psychological stress protocol. The difference in cortisol reactivity between good and poor health groups was larger than the difference between smokers and nonsmokers, which was also negatively associated with stress reactivity (data not shown). The differences in DBP reactivity were comparable to smoking, whereas differences in SBP and HR reactivity were about half the size of the smoking effect. Our results confirm the findings by Phillips et al. (2009), who recently reported an association between good subjective health and higher SBP, DBP, and HR reactivity to a psychological stress test. Also in congruence with these results, we showed that associations between self-reported health and BP reactivity survived adjustment for potential confounders, whereas the association with HR reactivity did not. The latter association seemed to be explained by BMI and SES (negatively and positively associated with HR reactivity; data not shown). Unlike the study of Phillips et al. (2009), we did not have longitudinal data at our disposal. We cannot, therefore, confirm

Self-reported health and stress reactivity

1175

their findings on reactivity and future self-reported health status. Another difference is that the age range in our cohort was much smaller. We investigated subjects in the age range of 55 to 60 years whereas Phillips et al. studied individuals from 24 to 63. A final divergence is that the effects of self-reported health on cardiovascular reactivity we found were not always statistically significant, as was found in Phillips’ study. The associations between the physical scale of the SF-36 and the highly correlated COOP/WONCA and cardiovascular reactivity did not reach statistical significance, but associations with cortisol reactivity did. This suggests that cardiovascular stress reactivity may be mainly linked to self-rated health that is predominantly mental health, whereas cortisol reactivity may be linked to both selfrated mental and physical health. An additional limitation to the present study may be that the speech task induced the highest stress reactivity, although this could also be due to the sum reactivity to the three tests adding up. However, talking per se is known to increase BP and HR, and, because we could not adjust the outcomes for a nonstress measure of speech, this may have confounded the results. The contribution of the present study is that the study by Phillips et al. (2009) had a number of limitations that we could address in our study. First of all, we measured BP and HR reactivity continuously with specialized equipment, providing a more accurate and reliable measurement compared to the single measurement with a semiautomated device used by Phillips et al. Also, we administered a self-perceived stress questionnaire during the stress protocol. Part of this questionnaire was an item in which the participant was asked to indicate how committed he or she felt to performing the three stress tasks. Such an item was missing in the study of Phillips et al. Based on the finding that those reporting poor health did not score differently from those reporting good health on this item, we can exclude the

explanation that the reported effect is caused by the fact that participants with poorer subjective health were relatively unmotivated and less engaged with the stress tasks. Interestingly, results of this questionnaire also showed that those reporting good health indicated feeling somewhat less stressed and more relaxed during the stress protocol than those reporting poor health. Finally, besides measuring subjective health with the same sort of single-question item as Phillips et al. used, we also administered the COOP/WONCA and the SF-36, both validated and frequently applied scales for measurement of subjective health. It was again shown that poorer subjective health was associated with lower cardiovascular and HPA axis responses to acute stress, confirming the results of the singlequestion item. The present results fit in a line of a number of studies suggesting that higher cardiovascular and cortisol reactivity to stress is not always associated with poor physical and mental health (see Phillips et al., 2009, for an overview of cardiovascular reactivity and Fries, Hesse, Hellhammer, & Hellhammer, 2005, for cortisol reactivity). This could be due to a failing stress response as a consequence of prolonged stress. Another explanation may be that a diminished stress response could have aversive immunological consequences. The biological stress response stimulates the immune system, which is beneficial for health. A decreased stress response may be less able to do so. Besides this, cortisol is needed to eventually suppress the inflammatory response. A lack of cortisol may thus lead to a prolonged inflammatory state with negative effects on health. In conclusion, in addition to Phillips et al. (2009) we showed that poor subjective health is associated with decreased cardiovascular and HPA axis reactivity to acute psychological stress. This finding lends further support to the notion that high stress reactivity may not always be associated with adverse health.

REFERENCES Aaronson, N. K., Muller, M., Cohen, P. D., Essink-Bot, M. L., Fekkes, M., Sanderman, R., et al. (1998). Translation, validation, and norming of the Dutch language version of the SF-36 Health Survey in community and chronic disease populations. Journal of Clinical Epidemiology, 51, 1055–1068. de Rooij, S. R., Painter, R. C., Phillips, D. I. W., Osmond, C., Tanck, M. W. T., Bossuyt, P. M. M., et al. (2006). Cortisol responses to psychological stress in adults after prenatal exposure to the Dutch famine. Psychoneuroendocrinology, 31, 1257–1265. Fries, E., Hesse, J., Hellhammer, J., & Hellhammer, D. H. (2005). A new view on hypocortisolism. Psychoneuroendocrinology, 30, 1010– 1016. Idler, E., & Benyamini, Y. (1997). Self-rated health and mortality: A review of twenty-seven community studies. Journal of Health and Social Behavior, 38, 21–37.

Larsson, D., Hemmingsson, T., Allebeck, P., & Lundberg, I. (2002). Selfrated health and mortality among young men: What is the relation and how may it be explained? Scandinavian Journal of Public Health, 30, 259–266. Nelson, E., Wasson, J., Kirk, J., Keller, A., Clark, D., Dietrich, A., et al. (1987). Assessment of function in routine clinical practice: Description of the COOP Chart method and preliminary findings. Journal of Chronic Diseases, 40(Suppl 1), 55S–69S. Phillips, A. C., Der, G., & Carroll, D. (2009). Self-reported health and cardiovascular reactions to psychological stress in a large community sample: Cross-sectional and prospective associations. Psychophysiology, 46, 1020–1027. (Received July 23, 2009; Accepted November 28, 2009)

Psychophysiology, 47 (2010), 1176–1181. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01032.x

BRIEF REPORT

Are voluntary switches corrected repetitions?

KIMBERLEY VANDAMME, ARNAUD SZMALEC, BAPTIST LIEFOOGHE, and ANDRE´ VANDIERENDONCK Department of Experimental Psychology, Ghent University, Ghent, Belgium

Abstract While recent years have witnessed a growing interest in Voluntary Task Switching (VTS), the control mechanisms that are required in order to switch tasks on a voluntary basis remain to be identified. Starting from the finding that in VTS the proportion of task repetitions is usually higher than the proportion of task switches (task-repetition bias), the present electrophysiological study tests and confirms the hypothesis that, during VTS, one initially re-selects the previously executed task, before correcting this bias and selecting the alternative task. On the one hand, these findings allow us to describe how people switch cognitive tasks voluntarily. On the other hand, our approach underlines the usefulness of electrophysiological measures in understanding the processes by which voluntary behavior occurs. Descriptors: Voluntary task switching, Executive control, LRP

underlying voluntary task-switching remains to this day poorly specified. Voluntary task switching, or mental flexibility in general, is experienced as effortful because it essentially requires a person to overrule the perseveration of actions. A common finding in VTS, which indicates that we do not always succeed in doing so, is that the proportion of task repetitions is higher than the proportion of task switches (i.e., the task-repetition bias). Mayr and Bell (2006, p. 774) interpreted this bias in terms of ‘‘the degree to which voluntary control can override the pull toward sticking with the currently most active task-set.’’ This idea is supported by the finding that the task-repetition bias increases for short compared to long inter-trial intervals (ITI). For instance, Arrington and Logan (2004) observed that the proportion of task repetitions was .68 for an ITI of 100 ms and .60 for an ITI of 1000 ms (see also Arrington, 2008; Arrington & Logan, 2005). In addition, for an ITI of 100 ms, Mayr and Bell (2006) found a larger repetition bias for blocks in which the target stimuli repeated on 50% of the trials compared to blocks in which the target stimuli repeated less frequently. Both these situations promote the perseveration of the previous task, and the larger task-repetition bias observed under these circumstances may lead to the hypothesis that in VTS the previous task is initially activated, and participants have to overcome this bias when performing the alternative task. The aim of the present study was to test this hypothesis by measuring the Lateralized Readiness Potential (LRP), which has already proven its usefulness in standard task switching (Gladwin, Lindsen, & De Jong, 2006; De Jong, Gladwin, & ‘t Hart, 2006). The LRP is a negative shift in the event-related brain potential (ERP) of the motor cortex, which is assumed to reflect handspecific motor activation or preparation (Coles, 1989). Regarding the present study, two different LRPs are of interest: the

Cognitive control refers to the human ability to produce meaningful, goal-directed behavior and to flexibly adapt this behavior according to changing situations (e.g., Botvinick, Braver, Barch, Carter, & Cohen, 2001). An influential framework for studying executive control is the task-switching paradigm, which requires participants to switch between imposed cognitive tasks (Jersild, 1927; Rogers & Monsell, 1995). Although the task-switching paradigm has proven to be useful in order to investigate the nature of the switch cost (i.e., impaired performance on task switches compared to task repetitions), it remains only partially informative about mental flexibility. The reason is that, in realworld multi-tasking environments, cognitive tasks are often not imposed, and one has to decide on a voluntary basis which task to perform when. A promising paradigm in the context of voluntary behavior is Voluntary Task Switching (VTS), in which participants are free to select which task they want to perform as long as they perform each task equally often and in a random order (Arrington & Logan, 2004, 2005; Liefooghe, Demanet, & Vandierendonck, 2009; Mayr & Bell, 2006). Because VTS requires endogenously driven task switching, it is usually seen as a more ecologically valid approach towards investigating cognitive control processes. Nevertheless, the control mechanism

This research was made possible by Grant No. 3G.0010.05 of the Special Research Fund of Ghent University. We are grateful to Gabriele Gratton, Chandramallika Basak, Frederick Verbruggen, and Sander Los for their useful comments and suggestions on a previous version of this manuscript. We would also like to thank Els Severens and Pascal Mestdagh for their help in the EEG analyses. Address correspondence to: Kimberley Vandamme, Department of Experimental Psychology, Ghent University, Henri Dunantlaan 2, 9000 B-Ghent, Belgium. E-mail: [email protected] 1176

Are voluntary switches corrected repetitions? foreperiod LRP and the late LRP. The foreperiod LRP (e.g., Wild-Wall, Sangals, Sommer, & Leuthold, 2003) is typically elicited when information about the required response is available prior to target presentation (Osman, Moore, & Ulrich, 1995; Ulrich, Leuthold, & Sommer, 1998), whereas the late LRP reflects motor preparation in answer to the target. Our use of lateralized motor readiness for measuring taskselection processes in VTS relies on the following rationale. If participants are asked to alternate randomly between two tasks mapped onto different hands, the task-selection process becomes a between-hand response selection process and, hence, the LRP can be used to trace the task-selection process. Accordingly, we designed an experiment in which participants alternated voluntarily between two tasks: digit classification (i.e., odd/even) and letter classification (i.e., consonant/vowel), with each task assigned to a different hand. If there is an initial bias towards repeating the previous task that must be overridden in order to perform the alternative task, we expect that on task-switch trials, a foreperiod LRP towards the previously activated response hand will precede the late LRP activation of the alternative response hand. For task-repetition trials, we anticipate that both foreperiod and late LRP will reflect activation towards the same hand as on the previous trial (i.e., the biased hand).

Method Participants Ten students (8 females, age range: 18 to 28 years) of Ghent University participated for payment (h25). All were righthanded, free of neurological and psychological disorders, and reported to have normal or corrected-to-normal vision. Apparatus and Stimuli Stimuli were presented in white against a black background on a 17-inch color monitor, placed at a distance of approximately 60 cm from the participant. Stimuli consisted of a letter and digit compound (e.g., E7; see also, Rogers & Monsell, 1995). The digit in the compound could be 1, 2, 3, 4, 6, 7, 8, or 9, and the letter G, K, M, R, A, E, I, or U. On every trial, the compound was presented in the centre of the screen, subtending a visual angle of 1.051 horizontally and 0.671 vertically. To discourage strategic behavior, the left-right positions of the letter and the digit varied randomly on a trial-to-trial basis. Depending on the task, participants judged whether the digit was odd or even, or whether the letter was a consonant or vowel. Procedure Participants were seated in front of a table in a sound-attenuated and electrically shielded room. A response device with horizontally arranged keys was placed on the table in front of the participant. The experiment began with written and verbal instructions. Participants were asked to select a task on each trial in such a way that they performed each task an approximately equal number of times and did generate unpredictable task sequences. Unpredictability was explained by translating the ‘‘coin toss’’ instructions of Arrington and Logan (2004) into Dutch. Participants were instructed to respond to the letters with the index and middle finger of the right hand, and to the digits with the fingers of the left hand. Furthermore, participants were asked to maintain fix-

1177 ation on the centre of the screen, and speed as well as accuracy were stressed. Participants started with a practice phase of 256 trials. On each trial, the digit-letter compound was presented for 500 ms, followed by a blank screen until the participant pressed a response key or 1000 ms had elapsed. Reaction times (RTs) were measured from the onset of the stimulus to the response, so that participants were given 1500 ms to respond. The next stimulus appeared 1000 ms after the response to the previous trial (i.e., ITI 5 1000 ms). If an incorrect response was given, the screen turned immediately red for 250 ms before the ITI started. Only participants with less than 15% errors and a minimum switch rate of 30% could continue the experiment. Electrode placement (approx. 20 min) followed the practice phase. Next, participants completed four blocks of 256 trials (similar to the practice phase) with a 5-min break after each block. During each break, accuracy and choice probabilities of each task (letter, digit) and each transition (repetition, switch) were presented on screen. The entire procedure lasted approximately 2 h. Electrophysiological Recordings The electroencephalogram (EEG) was continuously recorded (sample rate 5 512 Hz) from 8 Ag/AgCl electrodes located at standard electrode positions (Fpz, Fz, Fcz, C3, C4, Cz, Cpz, and Pz) of the 10–20 International System. Vertical electro-oculogram (vEOG) was registered (bipolar) from above and below the left eye. Horizontal electro-oculogram (hEOG) was recorded bipolar from the outer canthi of both eyes. All impedances were kept below 5 kO. Data Analysis For each trial, we first categorized the performed task on the basis of the hand that participants used to respond. Then we looked whether the response key was used that corresponded with the chosen task and the presented stimulus. To categorize the error trials, we assumed that participants always correctly selected the hand that corresponded with their task choice (e.g., digit task 5 left hand), but chose the wrong finger within that hand (Scheffers & Coles, 2000). The assumption that the choice of hand indeed corresponded with the selection of the appropriate task was tested and confirmed in an additional analysis on the behavioral data1. Once the trials were categorized according to the selected task, they were subsequently categorized into repetition and switch trials on the basis of the task performed on the previous trial. All error trials and trials following an error, the first trial of each block, and trials with RTs less than 150 ms were discarded from task probability, RTs, and ERP analyses. Note that, although trials following an error were deleted from 1 In an additional analysis, we compared performance on stimulus congruent and stimulus incongruent trials. On stimulus congruent trials, the same finger is required for both tasks (for instance, both stimuli require a key press with the index finger), whereas on stimulus incongruent trials, the two tasks require different fingers. If it were the case that participants responded with the hand that was allocated, for instance, to task A but used the task-rules associated with task B, it can be expected that incongruent stimuli (i.e., stimuli that require different finger presses) will lead to more incorrect responses and/or longer RTs. Our data show that the difference in RT between stimulus congruent (M 5 636, SD 5 56) and stimulus incongruent (M 5 633, SD 5 59) trials and the difference in error rate between stimulus congruent (M 5 .09, SD 5 .04) and stimulus incongruent (M 5 .08, SD 5 .04) trials was, however, not reliable (both Fo1).

1178 analysis, we registered which task was chosen on this trial in order to categorize the subsequent trial as a repetition or a switch trial. For the RT analyses, per participant RTs above 2.5 standard deviations from the mean of the condition were removed from the data set. EEG Analysis All electrodes were referenced offline to an average of the two ears, with the forehead serving as ground. EEG and EOG were filtered with a 0.16–30 Hz band-pass filter. Epochs exceeding 20 standard deviations (40 for EOG) of the signal amplitude were discarded. Prior to averaging, the data were corrected for eye movement artifacts by a subtraction of vEOG propagation factors, based on PCA-transformed EOG components (Nowagk & Pfeifer, 1996). For the stimulus-locked analyses, the EEG and EOG were segmented in epochs running from 1400 ms before to 1000 ms after stimulus presentation. The 200-ms stimulus-locked baseline started 1000 ms before stimulus presentation. For the response-locked analyses, the epochs started 1000 ms before until 200 ms after response onset. The 200-ms baseline referred to a 200-ms pre-stimulus baseline. The LRP was computed using the following formula: ((C3C4)right1(C4-C3)left)/2, calculated per participant and per condition at the average C3 and C4 waveforms (Coles, 1989). Positive and negative deflections correspond to incorrect and correct response activation, or to a task switch and task repetition, respectively, according to the hand lateralization logic used in the present study. After inspection of the grand average waveforms, the foreperiod LRP was measured in a window of 800 to 400 ms prior to stimulus onset of the stimulus-locked grand average waveform, whereas the late LRP was measured in a window of 200 ms prior to the response in the response-locked grand average waveform. The highest amplitude value in the corresponding time-window was taken as the amplitude of the LRP. More specifically, for the foreperiod LRP amplitude on switch trials, the highest positive amplitude value in the corresponding time-window was taken. On repetition trials, the highest negative amplitude value in the corresponding time-window was taken. Horizontal electro-ocular (hEOG) activity was computed analogously to the LRP, in order to assess possible EOG artifacts in the LRP. Based on visual inspection, the average stimuluslocked hEOG (S-hEOG) amplitude was assessed in a window of 1000 to 200 ms prior to stimulus onset, while the response-locked hEOG (R-hEOG) was measured in a window of 100 ms prior to the response. Results Behavioral Results Separate one-way analyses of variance (ANOVAs) with a 5 .05 were conducted on the RTs and error rates as a function of trial type (repetition trials vs. switch trials). RTs on switch trials (M 5 677.40; SD 5 76.01) were longer than on repetition trials (M 5 601.53; SD 5 46.41), F(1,9) 5 16.03, Z2p ¼ :64. More errors were made on switch trials (7%) than on repetition trials (5%), F(1,9) 5 6.74, Z2p ¼ :43, and the probability of a repetition, .54 " .05 (95% confidence interval), was larger than the probability of a switch, .46 " .05. Note that Arrington (2008, Experiment 1) also reported a relatively small repetition bias (.54) using a long RSI (1000 ms). In order to control for bottomup influences from the stimulus location, we also evaluated the possibility that participants stayed focused to only one side of the

K. Vandamme et al. digit-letter compound and used the stimulus on the focused side to choose a task. To do this, we calculated the likelihood of choosing a given task as a function of stimulus location (digitletter vs. letter-digit). We found that, when the digit was presented on the right and the letter on the left, the probability of choosing the digit task (i.e., left hand) was .51 " 0.3. When the letter was presented on the right and the digit on the left, the probability of choosing the letter task (i.e., right hand) was .51 " 0.2. This analysis thus indicates that the choice of task was not influenced by stimulus location. Electrophysiological Results The average stimulus- and response-locked LRP waveforms are shown in Figures 1 and 2. The mean amplitude of the foreperiod LRP (see stimulus-locked waveform) on task repetition trials (M 5 # 1.50, SD 5 .81) was reliably different from the zero baseline (t(9) 5 # 5.84), as was the mean amplitude of the foreperiod LRP on switch trials (M 5 1.90, SD 5 1.20; t(9) 5 4.99). The difference in foreperiod LRP amplitude between repetition and switch trials was reliable,2 F(1,9) 5 29.59, Z2p ¼ :77. For the late LRP (see Figure 2), amplitude values for both repetition (M 5 # 1.57, SD 5 1.24) and switch trials (M 5 # 3.80, SD 5 1.05) were different from baseline, t(9) 5 # 4.01, and, t(9) 5 # 11.47, respectively. The difference in late LRP amplitude between both kinds of trials was significant, F(1, 9) 5 25.87, Z2p ¼ :74. To assess possible contributions of horizontal eye movements to the LRP, the stimulus- and response-locked hEOG were statistically analyzed in the same way as the LRP. The analysis showed that the average stimulus-locked hEOG for repetition (M 5 0.20, SD 5 0.64) and switch trials (M 5 0.03, SD 5 0.50) did not differ from baseline, t(9) 5 0.97, p 5 .36 and t(9) 5 0.22, p 5 .83, respectively. Concerning the average response-locked hEOG, the repetition (M 5 # 0.17, SD 5 1.06) and switch trials (M 5 # 0.13, SD 5 1.37) were also not different from baseline, t(9) 5 # 0.51, p 5 .63 and t(9) 5 # 0.30, p 5 .77, respectively. These results indicate that our observed LRP effects were not contaminated by eye-movement activity. In addition, we evaluated the possibility that voluntary task switches may simultaneously occur with spatial shifts of attention to the left or right of the stimulus compound, which may in turn affect our electrophysiological results. We found no evidence for such concurrent spatial shifts of attention. Finally, we observed that also non-lateralized ERP activity, before and after the stimulus onset, showed switch-related effects. In a time window of 600 ms prior to the stimulus onset, our data reveal a larger contingent negative variation (CNV) negativity for switch than for repetition trials (see Figure 3, upper panel), F(1,9) 5 5.87, po.05, which according to the literature reflects increased preparatory processes (Hsieh & Chen, 2006; Lorist, Klein, Nieuwenhuis, De Jong, Mulder, & Meijman, 2000). The fact that this differentiation between switch and repetition trials occurs prior to stimulus onset is in line with the current view that top-down control mechanisms of voluntary task-selection are engaged in an anticipatory fashion as opposed to being more stimulus driven. After stimulus onset, we found 2 An additional analysis showed that there was no reliable difference in the average ‘‘absolute’’ foreperiod amplitude between repetition and switch trials, F(1,9) 5 1.31, p4.20. Furthermore, when we sorted our data in order to include the switch/repetition on the previous trials (i.e., Rep-Rep, Sw-Rep, Rep-Sw, and Sw-Sw), the results showed that there is foreperiod LRP activation after a repetition as well as after a switch trial.

Are voluntary switches corrected repetitions?

1179

Figure 1. Stimulus-locked LRPs for repetition and switch trials.

that the peak amplitude of the P3 (450–600 ms) was smaller on switch than on repetition trials (see Figure 3, lower panel), F(1,9) 5 18.47, po.01. A modulation of P3 amplitude has already been observed in standard task switching studies (e.g., Kieffaber & Hetrick, 2005; Mueller, Swainson, & Jackson, 2007) where it is assumed that a larger P3 amplitude on repetition trials reflects the consolidation of task-related stimulus evaluation processes due to task repetition. In the context of this study however, this interpretation must be made with care because of the baseline differences prior to stimulus onset.

Discussion Using an electrophysiological approach, the current study tested the hypothesis that voluntary task switching comprises a tendency towards selecting the same task as on the previous trial, and that this tendency needs to be overruled in order to switch deliberately towards the alternative task. An experiment was de-

signed in which two tasks were mapped onto different hands, so that the LRP could be used to gain insight into the nature and time course of the voluntary task-selection mechanism. The results support our hypothesis (see Figure 1): on task-switch trials, a foreperiod LRP towards the previously activated response hand preceded a late LRP activation of the alternative response hand. On task-repetition trials, the foreperiod and the late LRP were activated towards the same hand as on the previous trial. When considering our electrophysiological results in more detail, a number of observations deserve some further comments. As depicted in Figure 1, both on task-repetition and task-switch trials, the deflection of the LRP associated with the response to the previous stimulus n-1 can be seen around 1000 ms prior to the onset of the current stimulus n. Note that this is due to the fact that we used a fixed ITI of 1000 ms. After that, the LRP wave moves back to baseline, before developing a foreperiod LRP which peaks around 600 ms prior to stimulus onset, in the same direction as on the previous trial. Finally, a late LRP, which

Figure 2. Response-locked LRPs for repetition and switch trials.

1180

K. Vandamme et al.

Figure 3. Stimulus-locked ERPs for repetition and switch trials as recorded from FCz (upper panel) and Pz (lower panel).

according to Figure 2 peaks around 100 ms before the response, announces the response to stimulus n. It can be noticed that the foreperiod LRP has the same ‘‘absolute’’ amplitude on taskrepetition and task-switch trials. This means that the activation of the previous task is similar for both trial types, suggesting that, at this particular point in time, the decision of switching or repeating tasks has not been implemented yet. This bias, as well as the observed difference in CNV-like negativity between repetition and switch trials, shows that the alternative task is selected later, presumably through the intervention of top-down control in line with the instructional demands. It is also important to mention that the late LRP amplitude on repetition trials was smaller than on switch trials. This pattern of results is typically observed in tasks that require the overriding of prepotent responses, such as in the flanker task (Kopp, Rist, & Mattler, 1996; Willemssen, Hoormann, Hohnsbein, & Falkenstein, 2004) and the Simon task (Van der Lubbe, Jas´ kowski, Wauschkuhn, & Verleger, 2001). In the flanker task, participants are asked to respond to a centrally presented target stimulus, flanked by congruent (same response) or incongruent (different response) distracter items (Eriksen & Eriksen, 1974). When the flankers are presented prior to the target, there is foreperiod LRP activation towards the correct response hand on congruent trials, while on

incongruent trials the foreperiod LRP shows activation towards the incorrect response hand. Furthermore, these foreperiod LRPs are followed by a late LRP, which is smaller in amplitude on congruent than on incongruent trials. The latter finding is similar to the one we observed in the present study with VTS. The most plausible interpretation for these results in the flanker paradigm and VTS is that, due to the fact that on repetition trials in VTS and on congruent trials in the flanker task foreperiod LRP activation towards the correct response hand emerges, fewer resources are needed to eventually select the same hand when responding to the target (i.e., late LRP). In conclusion, the foreperiod LRPs measured in the present study indicate that voluntary switches are instantiated after an initial bias towards the previous task. Because the LRP is an average waveform, the current data do not necessarily indicate that, on each individual trial, the cognitive system resides in a mode to repeat tasks. Accordingly, it is more appropriate to conclude that, due to a repetition bias, the foreperiod LRP findings reveal an average tendency to repeat the same task. On a sufficient number of trials, this tendency must be overruled in order to switch tasks; these switches are then clearly corrected repetitions. Our explicit demonstration of the VTS mechanism also offers direct empirical evidence for Mayr and Bell’s (2006)

Are voluntary switches corrected repetitions?

1181

interpretation of the task repetition bias. In addition, the finding that processes underlying VTS can be measured online by means of activation in the motor cortices offers important new perspectives in research on VTS, because it constitutes the basis of a

research strategy in which new insights can be gained without altering the minimalistic environmental support that is required when investigating endogenous processes in task switching, or voluntary behavior in general.

REFERENCES Arrington, C. M. (2008). The effect of stimulus availability on task choice in voluntary task switching. Memory & Cognition, 36, 991–997. Arrington, C. M., & Logan, G. D. (2004). The cost of a voluntary task switch. Psychological Science, 15, 610–615. Arrington, C. M., & Logan, G. D. (2005). Voluntary task switching: Chasing the elusive homunculus. Journal of Experimental Psychology: Learning, Memory, and Cognition, 31, 683–702. Botvinick, M., Braver, T., Barch, D., Carter, C., & Cohen, J. (2001). Conflict monitoring and cognitive control. Psychological Review, 108, 625–652. Coles, M. G. H. (1989). Modern mind-brain reading: Psychophysiology, physiology, and cognition. Psychophysiology, 26, 251–269. De Jong, R., Gladwin, T. E., & ‘t Hart, B. M. (2006). Movement-related EEG indices of preparation in task switching and motor control. Brain Research, 1105, 73–82. Eriksen, B. A., & Eriksen, C. W. (1974). Effects of noise letters upon the identification of a target letter in a nonsearch task. Perception and Psychophysics, 16, 143–149. Gladwin, T. E., Lindsen, J. P., & De Jong, R. (2006). Pre-stimulus EEG effects related to response speed, task switching and upcoming response hand. Biological Psychology, 72, 15–34. Hsieh, S., & Chen, P. (2006). Task reconfiguration and carryover in task switching: An event-related potential study. Brain Research, 1084, 132–145. Jersild, A. T. (1927). Mental set and shift. Archives of Psychology, 89, 1–81. Kieffaber, P. D., & Hetrick, W. P. (2005). Event-related potential correlates of task switching and switch costs. Psychophysiology, 42, 56– 71. Kopp, B., Rist, F., & Mattler, U. (1996). N200 in the flanker task as a neurobehavior tool for investigating executive control. Psychophysiology, 33, 282–294. Liefooghe, B., Demanet, J., & Vandierendonck, A. (2009). Is advance reconfiguration in voluntary task switching affected by the design employed? Quarterly Journal of Experimental Psychology, 65, 850– 857. Lorist, M. M., Klein, M., Nieuwenhuis, S., De Jong, R., Mulder, G., & Meijman, T. F. (2000). Mental fatigue and task control: Planning and preparation. Psychophysiology, 37, 614–625.

Mayr, U., & Bell, T. (2006). On how to be unpredictable: Evidence from the voluntary task-switching paradigm. Psychological Science, 17, 774–780. Mueller, S. C., Swainson, R., & Jackson, G. M. (2007). Behavioural and neurophysiological correlates of bivalent and univalent responses during task switching. Brain Research, 1157, 56–65. Nowagk, R., & Pfeifer, E. (1996). UNIX implementation of the ERP evaluation package (EEP 3.0). In A. D. Friederici & D. Y. Von Cramon (Eds.), Annual report of Max-Planck-Institute of Cognitive Neuroscience. Leipzig, Germany: Max-Planck-Institute. Osman, A., Moore, C. M., & Ulrich, R. (1995). Bisecting RT with lateralized readiness potentials: Precue effects after LRP onset. Acta Psychologica, 90, 111–127. Rogers, R. D., & Monsell, S. (1995). Costs of a predictable switch between simple cognitive tasks. Journal of Experimental Psychology: General, 124, 207–231. Scheffers, M. K., & Coles, M. G. H. (2000). Performance monitoring in a confusing world: Error-related brain activity, judgments of response accuracy, and types of errors. Journal of Experimental Psychology:Human Perception and Performance, 26, 141–151. Ulrich, R., Leuthold, H., & Sommer, W. (1998). Motor programming of response force and movement direction. Psychophysiology, 35, 721– 728. Van der Lubbe, R. H. J., Jas´ kowski, P., Wauschkuhn, B., & Verleger, R. (2001). Influence of time pressure in a simple response task, a choiceby-location task, and the Simon task. Journal of Psychophysiology, 15, 241–255. Wild-Wall, N., Sangals, J., Sommer, W., & Leuthold, H. (2003). Are fingers special? Evidence about movement programming and preparation from event-related brain potentials. Psychophysiology, 40, 7–16. Willemssen, R., Hoormann, J., Hohnsbein, J., & Falkenstein, M. (2004). Central and parietal event-related lateralizations in a flanker task. Psychophysiology, 41, 762–771.

(Received May 15, 2009; Accepted January 6, 2010)

Psychophysiology, 47 (2010), 1182–1191. Wiley Periodicals, Inc. Printed in the USA. Copyright r 2010 Society for Psychophysiological Research DOI: 10.1111/j.1469-8986.2010.01029.x

ERP correlates of the irrelevant sound effect

RAOUL BELL, SANDRA DENTALE, AXEL BUCHNER, and SUSANNE MAYR Department of Experimental Psychology, Heinrich-Heine-University, Du¨sseldorf, Germany

Abstract The irrelevant sound effect refers to a decrement in serial-recall performance when auditory distractors are played during encoding or retention of the to-be-remembered items. We examined the event-related brain potentials (ERPs) that were elicited in response to the auditory distractors during encoding and retention of visually presented target sequences. Changing-state distractor sequences that consisted of several different distractor items interfered more with serial recall than steady-state sequences that consisted of repetitions of a single distractor item. The ERP responses that were elicited in response to the distractors comprised the exogenous N1 component and were further characterized by a subsequent positive wave, and a late negativity. The changing-state effect was associated with an increased N1 and a P3a. The results support the attention-capture account of the irrelevant sound effect. Descriptors: Auditory evoked potentials, Auditory distraction, Attentional capture, Working memory

equivalent irrelevant sound effects for speech and non-speech distractors (Jones & Macken, 1993; Tremblay, Nicholls, Alford, & Jones, 2000). It is assumed that irrelevant sound interferes with the maintenance of the to-be-remembered items in working memory because the magnitude of interference does not change regardless of whether the to-be-ignored stimuli are played during encoding or retention of the target material (Buchner, Rothermund, Wentura, & Mehl, 2004; Miles, Jones, & Madden, 1991). Currently, several working-memory models compete for the best explanation of the irrelevant-sound effect (Cowan, 1995; Jones, 1993; Neath, 1999; Salame´ & Baddeley, 1982). These theories fall into one of two categories depending on whether they specify a role for attention in the maintenance of information or not (Elliott, 2002). Both the modular working memory model (Baddeley & Hitch, 1974) and the object-oriented episodic record model (Jones, 1993) imply that (a) attention is not needed for the maintenance of information within working memory and (b) attentional distraction does not play a role in the irrelevant sound effect. The embedded processes model (Cowan, 1995), in contrast, implies that the irrelevant sound effect is due to attentional capture. According to this view, unexpected changes in the auditory modality elicit orienting reactions that draw the focus of attention away from the primary task of maintaining the representations of the target items in a highly accessible state. According to this model, steady-state distractors cause less memory disruption than changing-state distractors because the orienting response habituates to repeated stimulation. Only new or changing sounds have to be attended because they may be important to the organism. Thus, the habituation of the orienting response serves as an attentional filter. Another prediction of the model is that auditory distractors that are of relevance for the individual should be more likely to capture attention, thereby increasing the interference effect. Consistent with this hypothesis, the content of the auditory

The irrelevant sound effect refers to the disruption of serial recall due to the presentation of auditory distractors. In the standard paradigm, participants are required to recall lists of items (digits, consonants, or words) that are sequentially presented in the visual modality. During presentation of these items or during a short retention phase, irrelevant sounds are played. Participants are required to recall the items in the order of their presentation. Typically, irrelevant sound decreases serial recall considerably relative to a silent-control condition. It is well established that the main determinant of the size of the irrelevant sound effect is the number of ‘‘changing states’’ (roughly defined as abrupt changes in pitch and amplitude) in the auditory channel (Campbell, Beaman, & Berry, 2002; Jones & Macken, 1995; Jones, Madden, & Miles, 1992). For instance, continuous random pitch glides fail to disrupt serial recall, but the same glides when interrupted by periods of silence interfere with recall (Jones, Macken, & Murray, 1993). Steady-state sequences that consist of repetitions of a single distractor disrupt serial recall to a lesser extent than changing-state sequences that consist of several different distractors (Jones et al., 1992). Besides, disruption is enhanced when the to-be-ignored sound sequence contains a single deviant that differs significantly from the other distractors in the sequence (Lange, 2005). Although speech stimuli usually interfere with serial recall more than non-speech stimuli (sine wave tones, environmental sounds; Buchner, Bell, Rothermund, & Wentura, 2008; LeCompte, Neely, & Wilson, 1997), those studies that have controlled the transient characteristics of the stimulation found The research reported in this article was supported by a grant from the Deutsche Forschungsgemeinschaft. Address correspondence to: Raoul Bell, Institut fu¨r Experimentelle Psychologie, Heinrich-Heine-Universita¨t, D-40225 Du¨sseldorf, Germany. E-mail: [email protected] 1182

ERP correlates of the irrelevant sound effect distractors enhances interference when it is of relevance to the individual (Bell, Mund, & Buchner, in press; Buchner, Mehl, Rothermund, & Wentura, 2006; Buchner et al., 2004). A separate line of research has also advanced our understanding of auditory distraction and may provide some insights into the mechanisms underlying the performance decrement in the irrelevant-sound paradigm. There are a number of studies examining event-related brain potentials (ERPs) that occur in response to abrupt changes in the to-be-ignored auditory modality. Typically, the ERPs elicited by rare, novel, or varying stimuli in the ignored auditory modality comprise an N1 component as is typical for auditory stimuli in general and are further characterized by a subsequent positive wave (P3a), and a late negativity. It has been proposed that these components are sensitive to, or might even reflect, different stages of involuntary attention switching (Bendixen, Roeber, & Schro¨ger, 2007; Escera, Alho, Winkler, & Na¨a¨ta¨nen, 1998; Escera & Corral, 2007). First, the N1 is elicited even if the auditory modality is to be ignored, and participants focus on a visual primary task such as reading a book or watching a silent movie (Na¨a¨ta¨nen, 1990; Na¨a¨ta¨nen & Picton, 1987; Na¨a¨ta¨nen & Winkler, 1999). It is distributed mostly over frontocentral areas of the scalp. The N1 response is primarily determined by the amount of physical change in the auditory environment. When the amount of ‘‘changing states’’ is controlled, different types of auditory stimuli such as tones, speech, and environmental sounds elicit very similar N1 responses (Na¨a¨ta¨nen & Picton, 1987). The N1 is modulated by stimulus-nonspecific and stimulus-specific refractoriness effects. Stimulus-nonspecific refractoriness refers to a decrement of the N1 due to any prior acoustic stimulation. A large N1 response is elicited by the first auditory stimulus in a sequence of auditory events after a long period of silence. N1 responses to subsequent stimuli are usually much smaller (Na¨a¨ta¨nen, 1990; Na¨a¨ta¨nen & Picton, 1987). Stimulus-specific refractoriness refers to a decrement of the N1 to an auditory stimulus due to the presentation of identical or very similar preceding stimuli (Na¨a¨ta¨nen & Picton, 1987; Opitz, Schro¨ger, & von Cramon, 2005; Schro¨ger & Wolff, 1998). N1 refractoriness increases and N1 amplitude attenuates with stimulus repetition. Functionally, the N1 is often associated with a ‘‘call for attention.’’ It is assumed that the N1 generator triggers an attention switch to the auditory stimulus when the N1 response exceeds a threshold (e.g., Campbell, Winkler, Kujala, & Na¨a¨ta¨nen, 2003; Na¨a¨ta¨nen, 1990; Na¨a¨ta¨nen & Picton, 1987; Rinne, Sa¨rkka¨, Degerman, Schro¨ger, & Alho, 2006). If the to-be-ignored auditory stimulus deviates from a preceding repetitive stimulus sequence, a mismatch negativity (MMN) is elicited (Schro¨ger & Wolff, 1998). The MMN refers to the difference between the ERP to the unexpected deviant and the repetitive standard stimulus. The MMN is a frontocentral component and has a somewhat longer latency than the N1. The MMN has also been associated with a call for attention (e.g., Schro¨ger & Wolff, 1998). In response to novel, unexpected, or changing task-irrelevant auditory stimuli, the N1/MMN complex is often followed by a subsequent positive wave with a frontocentral scalp distribution. This P3a is associated with an orienting response to the eliciting stimulus (Friedman, Cycowicz, & Gaeta, 2001; Polich, 2007). It is assumed that this orienting response coincides with a conscious evaluation of the auditory stimulus. The P3a is often observed alongside higher error rates and increased reaction times in the primary tasks (Escera et al., 1998; Escera, Yago, & Alho, 2001; Schro¨ger & Wolff, 1998). In some studies, the P3a response was

1183 followed by a late negativity with a frontal maximum (e.g., Escera et al., 2001). This negative deflection has been termed reorienting negativity (RON) based on the assumption that the component mayFat least in partFreflect a reorienting of attention back to the primary task (Bendixen et al., 2007; Berti, 2008a; Escera et al., 2001). There are only a few studies that have examined electrophysiological correlates of the irrelevant sound effect (Campbell, Winkler, & Kujala, 2007; Campbell et al., 2003; Kopp, Schro¨ger, & Lipka, 2004, 2006; Martı´ n-Loeches & Sommer, 1998; Weisz & Schlittmeier, 2006). Most of these studies examined brain responses associated with the processing of the target stimuli, but not the distractor stimuli. Recently, Campbell et al. (2003) observed that the N1 to changing-state distractors was increased in comparison to the N1 elicited by steady-state distractors. The difference between the changing-state and the steady-state condition was explained by a stimulus-specific refractoriness of the supratemporal N1 component. In a subsequent study, Campbell et al. (2007) compared the ERPs elicited by changing-state distractors with those elicited by rare deviants in steady-state sequences. Again, they found an increased N1 elicited by changing-state sounds (as compared to the N1 elicited by steady-state distractors). A MMN was observed only in the deviant condition, consistent with other results suggesting that the MMN is only found if a neural model that is based on the regularities in the auditory environment is violated. In both studies, the auditory distractors elicited no P3a. Based on this finding, the attention-capture account of the irrelevant sound effect was rejected. More specifically, Campbell et al. (2007) suggest that the lack of an increase in P3a alongside with the increase in irrelevant-speech interference means that the memory disruption observed in their study ‘‘relies on different mechanisms than those commonly observed for distraction in studies employing the oddball paradigm [and] may be taken as a sign that the current form of memory disruption does not require attentional capture’’ (p. 538; see Campbell et al., 2003, for a similar claim). However, the lack of a P3a could also be attributed to the fast presentation rate that was used in these studies. In Campbell et al.’s (2007) study, the auditory distractors were presented for 100 ms with a silent inter-distractor-interval of 227 ms. In other words, the interval between the onsets of two consecutive stimuli was 327 ms. Thus, it is possible that the components of consecutive distractors may have overlapped, which may have decreased the likelihood of finding significant differences in the amplitudes of later components such as a P3a with a to-beexpected latency of about 250 ms (e.g., Escera et al., 2001). The finding of a P3a due to changing-state distractors in the irrelevant-speech paradigm would fit to previous results showing a P3a and a RON in response to varying changing-state distractor stimuli comprising no regularity (Bendixen et al., 2007), which were interpreted in terms of distraction. Thus, it is possible that similar effects can also be obtained in the irrelevant speech paradigm (for evidence that the P3a is not necessarily confined to the oddball paradigm, see, for example, Berti, 2008b). The purpose of the present study was to replicate the findings of Campbell et al. (2003, 2007) using a somewhat slower presentation rate so as to allow for a better measurement of later ERP components such as the P3a. The experiment was a typical irrelevant-sound experiment. Participants were required to serially recall lists of digits. To-be-ignored sequences of auditory distractors were played during encoding and retention. The distractor sequences consisted either of repetitions of a single

1184 one-syllable distractor word (steady state) or of seven different distractor words (changing state). We expected to find an irrelevant sound effect, that is, a decrement of serial-recall performance due to the presentation of auditory distractors. We also expected to replicate the changing-state effect, that is, worse recall performance in the changing-state condition than in the steady-state condition. With respect to the ERP data, we expected to replicate the finding of Campbell et al. (2003, 2007) that the increase in interference due to changing-state distractors is accompanied by an increase in the N1 amplitude. The most interesting question was whether we would observe a P3a response that would be indicative of an attention switch to the auditory modality. Method Participants EEG recordings and behavioral data were obtained from 40 persons. Six data sets were excluded from the analyses because of muscular artifacts in almost every trial. The remaining 34 participants (24 female, 10 male) were aged between 18 and 40 years (M 5 25). Participants were German native speakers and had no history of neurological disorders or hearing disabilities. Stimuli and Procedure. The visually presented, to-be-remembered lists consisted of eight digits sampled randomly without replacement from the set f1, 2, 3, 4, 5, 6, 7, 8, 9g. A total of 50 such lists were generated for each participant, with 15 lists in each of the three different conditions (silent control, steady state, changing state), and five lists for the training trials. The items were presented at the center of a 22-inch CRT screen. The numbers were written in white Arial font on a black background. Viewing distance was approximately 100 cm, although head position was not constrained. At this distance, each target digit subtended about 1.01 horizontally and 1.51 vertically. Distractor sounds were seven one-syllable German nouns (Bug [bu:k], Eid [a t], Norm [n m], Reiz [ ], Sieb [zi:p], Tausch [ta ], Term [t m]; German pronunciation in brackets) with a mean frequency of 8/1,000,000 according to the German language corpus available in the CELEX database (Baayen, Piepenbrock, & van Rijn, 1993). Information about the valence and concreteness of the distractor words was obtained in an independent norming study (N 5 34). Mean valence of the distractor words (on a scale ranging from -10 [extremely unpleasant] to 110 [extremely pleasant]) was 0 (SD 5 1). Mean concreteness of the distractor words (on a scale ranging from 1 [very abstract] to 20 [very concrete]) was 10 (SD 5 4). All word recordings were spoken by the same female voice, digitally recorded at 44.1 kHz using 16-bit encoding, edited to last 570 ms, and normalized to minimize amplitude differences among the words. The sounds were delivered binaurally through headphones (Sennheiser OMX 90 VC Style, Wedemark, Germany). The average sound level was about 65 dB(A). The experiment began with five practice trials, in which eight visually presented digits had to be remembered after a short retention phase. Each of the eight target digits was presented for 800 ms with a 200-ms blank inter-stimulus-interval. The retention phase after each list was 8,000 ms long (i.e., it was as long as the encoding phase during which the targets were presented). Throughout the retention phase, a fixation cross was shown. Following the retention phase, eight question marks appeared on the screen, corresponding to the eight serial positions of the visual targets. This was the signal for the participants to commence

R. Bell et al. recalling the list items in the order of presentation. The digits were entered via the number keys of the computer keyboard. Typing the first digit replaced the first question mark with that digit, typing the second digit replaced the second question mark, and so on. Participants were required to press a button labeled ‘‘don’t know’’ (the ‘‘0’’ key on the number keypad) for each digit they could not recall. As is usual in many irrelevant-sound experiments, including our own (Bell & Buchner, 2007; Buchner et al., 2008; Buchner & Erdfelder, 2005; Buchner, Irmen, & Erdfelder, 1996; Buchner et al., 2004), participants were allowed to correct their responses. The arrow keys of the computer keyboard could be used to move the current selection to another position at which any prior entry could be replaced. After replacing all of the question marks by numbers or ‘‘don’t know’’ responses, the participants were asked to initiate the next trial by pressing the spacebar. If the spacebar was pressed before all question marks were replaced, a 1,500-ms visual warning was shown. In the distractor conditions, the first distractor started 170 ms prior to the presentation of the first target stimulus to ensure that the onsets of the target and distractor stimuli were uncorrelated (see Figure 1). The auditory distractors were presented for 570 ms with a silent 230-ms inter-distractor interval. Presentation of the distractors continued throughout the retention phase. Ten distractors were played in the encoding phase, and 10 distractors were played in the retention phase. In the steady-state condition, the to-be-ignored sequences consisted of 20 repetitions of a single distractor word. The distractor word was randomly selected from the set of seven distractor words. In the changing-state condition, the to-be-ignored sequences consisted of all seven words of the distractor set that were randomly ordered and repeated until all 20 distractors were presented. EEG recordings and data analysis. An elastic cap with predefined electrode positions (Falk-Minow-Services, Munich, Germany) was mounted on the participant’s head. The 30 active silversilver chloride electrodes were referenced to linked earlobes, with impedance kept below 5 kO. Vertical and horizontal electro-oculograms (EOGs) were recorded to control for ocular artifacts. The difference potential between two electrodes placed above and below the right eye provided the vertical EOG. The horizontal EOG was calculated as the difference potential between two electrodes placed at the outer canthi of the right and left eye. Biosignals were recorded continuously (NuAmps 40 channel digital DC EEG amplifier, Neuroscan, Singen, Germany) sampled at 500 Hz, and online band pass filtered (0.1 to 40 Hz). Offline, EEG data were filtered (0.5–30 Hz, ! 24 dB cut-offs), segmented according to each distractor sound onset ( ! 200 to 570 ms epoch length), and baseline corrected ( ! 200 to 0 ms). Ocular (vEOG, hEOG) artifacts were corrected based on the algorithm proposed by Gratton, Coles, and Donchin (1983). Single electroencephalogram (EEG) sweeps containing muscular artifacts were removed based on visual inspection. The remaining sweeps were averaged according to the distractor condition (silent control, steady state, changing state), presentation phase (encoding phase, retention phase), and electrode position. ERP analysis of serial position had to be omitted due to insufficient segment numbers. Three time windows were determined as regions of interest (N1: 80–170 ms, P3a: 200–350 ms, late negativity: 380–490 ms). Mean reference-to-baseline amplitudes within these windows were computed for each participant, distractor condition, presentation phase, and electrode. To examine topographical effects systematically, the electrodes were split according to their

ERP correlates of the irrelevant sound effect

1185

Figure 1. Schematic illustration of a trial. The upper row depicts an example of a sequence of visual target events. The lower rows depict examples for the sequences of auditory distractors in the silent-control condition (second row), steady-state condition (third row), and changing-state condition (fourth row).

caudality (anterior, central, posterior) and laterality (left, medial, right) into nine clusters of electrodes of approximately the same size (left anterior [LA; Af3, F3, F7], medial anterior [MA; Fz, Fc1, Fc2], right anterior [RA; Af4, F4, F8], left central [LC; Fc5, C5, C3, Cp5], medial central [MC; Cz, Cp1, Cp2], right central [RC; Fc6, C4, C6, Cp6], left posterior [LP; P3, P7, O1], medial posterior [MP; Pz, Po3, Po4, Oz], and right posterior [RP; P4, P8, O2]), as is often done in studies examining auditory selective attention (Mayr, Niedeggen, Buchner, & Orgs, 2006; Mayr, Niedeggen, Buchner, & Pietrowsky, 2003). This has been proven to be a good compromise between spatial resolution and reliability of the measures.

Design The independent variables for the behavioral data were distractor condition (silent control, steady state, changing state) and serial position. The dependent variable was serial-recall performance, which was scored according to a strict serial-recall criterion. For the ERP data, only two levels of the distractor-condition variableFthe steady-state condition and the changing-state conditionFwere of theoretical interest. Given that in most irrelevantspeech experiments the auditory distractors are presented concurrently with the visual, to-be-remembered items (Bell et al., in press; Buchner & Erdfelder, 2005; Buchner et al., 1996, 2006; LeCompte et al., 1997; Salame´ & Baddeley, 1982), we were most interested in what happens during the encoding phase, which is why the statistical analysis will focus primarily on the encoding phase. The design included distractor type (steady state, changing state), electrode caudality (anterior, central, posterior), and electrode laterality (left, medial, right) as independent variables. This 2 ! 3 ! 3 design was considered separately for each of the three different time windows (80–170 ms, 200–350 ms, 380–490 ms). A multivariate approach was used for all within-subjects comparisons. In our applications, all multivariate test criteria correspond to the same (exact) F statistic, which is reported.

Results Serial-Recall Performance Figure 2 illustrates the serial-recall performance in the three experimental conditions. A 3 ! 8 repeated measures multivariate analysis of variance (MANOVA) with distractor type and serial position as independent variables showed significant main effects of distractor type [F(2,32) 5 38.55, po.01, Z2 5 .71], and of serial position [F(7,27) 5 35.20, po.01, Z2 5 .90]. We used orthogonal contrasts to test more specific hypotheses about the effects of the different distractor types. The first of these orthogonal contrasts showed that the difference between the silentcontrol condition and the two distractor conditions combined was significant [F(1,33) 5 76.05, po.01, Z2 5 .70], confirming that there was a typical irrelevant sound effect. Performance in the changing-state condition was significantly worse than performance in the steady-state condition, showing that there was also a changing-state effect [F(1,33) 5 19.56, po.01, Z2 5 .37]. ERP Data First the ERPs evoked by the auditory distractors in the encoding phase were analyzed. The grand-averaged ERPs are depicted in the upper panel of Figure 3. In line with previous findings examining the ERPs to auditory distractors (Escera et al., 1998; Escera & Corral, 2007), the electrophysiological activity generated by the auditory distractors was characterized by a prominent N1 peak, a subsequent positivity, and a late negative component. Mean amplitudes of the ERPs with t statistics for significant differences from zero are shown in Table 1. The ERP components differed between the steady-state and the changingstate distractor conditions. The difference waveforms between these two conditions are shown in Figure 4. Mean amplitudes were submitted to a distractor type (steady state, changing state) ! caudality (frontal, central, posterior) ! laterality (left, medial, right) MANOVA, separately for each temporal ERP epoch. In order to keep the Results section concise, we report only main effects of the distractor-type variable

1186

Figure 2. Mean proportion of digits correctly recalled as a function of distractor type and serial position (left panel) and averaged across serial positions (right panel). The error bars represent the standard errors of the means.

and significant interactions involving the distractor-type variable which are theoretically most relevant. First, we examine the N1 negativity. N1 amplitudes were larger for changing-state distractors than for steady-state distractors [F(1,33) 5 5.12, p 5 .03, Z2 5 .13]. There was also a significant interaction between distractor type, caudality, and laterality [F(4,30) 5 4.04, p 5 .01, Z2 5 .35]. This interaction reflects the fact that the effect of distractor type was maximal at medial central and medial anterior electrodes. The N1 was followed by a subsequent positivity (see Table 1) that was most pronounced at frontal electrodes. Most importantly, there was a much larger positive deflection for changingstate distractors than for steady-state distractors [F(1,33) 5 21.44, po.01, Z2 5 .39]. A significant distractor type ! caudality interaction [F(2,32) 5 33.28, po.01, Z2 5 .68] primarily reflected the fact that the increase in positivity in the changing-state condition was most pronounced at frontal sites. The positive wave was followed by a late negativity (see Table 1) that was maximal at central and medial electrodes. There was no significant effect of distractor type [F(1,33) 5 0.53, p 5 .47, Z2 5 .02], but a significant interaction between distractor type and caudality [F(2,32) 5 7.12, po.01, Z2 5 .31], and a significant three-way interaction between distractor type, caudality, and laterality [F(4,30) 5 3.06, p 5 .03, Z2 5 .29], indicating a more pronounced negativity at medial, central electrodes in the changing-state than in the steady-state condition. The lower panel of Figure 3 displays the grand-averaged ERPs evoked by the auditory distractors in the retention phase.

R. Bell et al. As is evident, N1 amplitudes were much smaller in the retention phase than in the encoding phase. There may be several reasons for this finding. In part, the reduction of the N1 in the retention phase may be due to stimulus-nonspecific refractoriness effects (see Na¨a¨ta¨nen, 1990; Na¨a¨ta¨nen & Picton, 1987; Woods & Elmasian, 1986) given that the encoding-phase distractor sequence was presented after a period of silence, whereas the retentionphase distractor sequence immediately followed the encodingphase distractor sequence. Note that even the changing-state distractors had several features in common (e.g., voice, duration, inter-stimulus interval), which might have amplified the refractoriness effect. Furthermore, given that the silent-control baseline was more negative in the encoding phase, the concurrent presentation of the visual target items in the encoding phase as opposed to concurrent target rehearsal during the retention phase may have led to a positive shift of the N1 component. Speculatively, this may also be due to a shift of the focus of attention from an external orientation (encoding phase) to an internal orientation (retention phase). Note that the result fits with the findings of Valtonen and colleagues (Valtonen, May, Ma¨kinen, & Tiitinen, 2003) who found a decrement of the magnetic counterpart of the N1 (i.e., the N1m wave) in response to the auditory distractors that were played during retention (in comparison to the N1m elicited by the encoding-phase distractors). However, the most critical aspect of the results is the difference between the changing-state and the steady-state conditions. With respect to this property, encoding phase and retention phase are very similar (see Figure 4 for difference waves between the two distractor conditions) at a descriptive level. This impression was confirmed by the statistical analysis of the results. Mean amplitudes of the difference potentials were submitted to a presentation phase (encoding phase, retention phase) ! caudality (frontal, central, posterior) ! laterality (left, medial, right) MANOVA, separately for each temporal ERP epoch. The mean amplitude of the difference wave in the 80–170 ms time window was different from zero [F(1,33) 5 14.88, po.01, Z2 5 .31]. The analysis of the difference waves in the 80–170 ms time window revealed significant main effects of caudality [F(2,32) 5 4.28, po.01, Z2 5 .36], and laterality [F(2,32) 5 8.76, po.01, Z2 5 .35], and a significant caudality ! laterality interaction [F(4,30) 5 5.07, p 5 .01, Z2 5 .24], confirming that the difference between changing state and steady state was most pronounced at medial anterior and medial central electrodes. Most importantly, there was no main effect of presentation phase [F(1,33) 5 1.28, p 5 .27, Z2 5 .04], suggesting that the difference between changing state and steady state was approximately of the same size in both the encoding phase and the retention phase. Descriptively, the difference between the changing-state and the steady-state condition was even more pronounced in the retention phase than in the encoding phase. This was to be expected given that it can be assumed that stimulus-specific refractoriness increases somewhat with the number of preceding steady-state auditory events. The two-way interactions between presentation phase and caudality [F(2,32) 5 1.48, p 5 .24, Z2 5 .08], and between presentation phase and laterality [F(2,32) 5 1.06, p 5 .36, Z2 5 .06], and the three-way interaction among these variables [F(4,30) 5 0.71, p 5 .59, Z2 5 .09] were not significant, suggesting similar scalp distributions of difference waves in the encoding and in the retention phase. The mean amplitude of the difference wave in the 200–350 ms window was different from zero [F(1,33) 5 28.02, po.01, Z2 5 .46]. An analysis of the difference waves in the 200–350 ms window revealed significant main effects of caudality

ERP correlates of the irrelevant sound effect

1187

Figure 3. Grand average ERPs for the three distractor conditions (silent-control, steady-state, changing-state), separately for the two presentation phases (upper panel: encoding phase; lower panel: retention phase) as a function of caudality and laterality, that is, for left anterior (LA; Af3, F3, F7), medial anterior (MA; Fz, Fc1, Fc2), right anterior (RA; Af4, F4, F8), left central (LC; Fc5, C5, C3, Cp5), medial central (MC; Cz, Cp1, Cp2), right central (RC; Fc6, C4, C6, Cp6), left posterior (LP; P3, P7, O1), medial posterior (MP; Pz, Po3, Po4, Oz), and right posterior (RP; P4, P8, O2) sites. In the silent-control condition, we averaged across the same time windows as in the other two conditions, although no distractor sounds were played in this condition.

[F(2,32) 5 44.31, po.01, Z2 5 .74], confirming that the difference between the changing-state condition and the steady-state condition was most pronounced at frontal electrodes. Again, there was no main effect of presentation phase [F(1,33) 5 0.71, p 5 .41, Z2 5 .02], suggesting that the mean difference amplitude Table 1. Mean Amplitudes of the Event-Related Potentials Medial central 80–170 ms Condition Steady state Changing state nn

po.01.

mV

2

t(33) Z

Medial frontal

Medial central

200–350 ms

380–490 ms

mV

t(33) Z

2

mV

t(33) Z2

! 1.10 5.15nn .46 1.16 3.89nn .31 ! 0.74 3.64nn .29 ! 1.61 6.81nn .58 2.64 8.69nn .70 ! 1.18 4.70nn .40

was of similar size in the encoding and the retention phase. Furthermore, there were no two-way interactions between presentation phase and caudality [F(2,32) 5 1.15, p 5 .33, Z2 5 .07], and between presentation phase and laterality [F(2,32) 5 0.17, p 5 .84, Z2 5 .01], and the three-way interaction among these variables was also not significant [F(4,30) 5 1.18, p 5 .34, Z2 5 .14], suggesting similar scalp distributions of the difference waves in the encoding phase and the retention phase. An analysis of the difference waves in the 380–490 ms window revealed significant main effects of caudality [F(2,32) 5 6.88, po.01, Z2 5 .30], and laterality [F(2,32) 5 5.42, po.01, Z2 5 .25], confirming a somewhat more pronounced difference between the steady-state condition and the changing-state condition at medial, central electrodes. Again, there was no main effect of presentation phase [F(1,33) 5 0.01, p 5 .94, Z2o.01], and there were no two-way interactions between presentation

1188

R. Bell et al.

Figure 4. Difference waveforms (changing state–steady-state distractor condition) for the two presentation phases as a function of caudality and laterality, that is, for left anterior (LA; Af3, F3, F7), medial anterior (MA; Fz, Fc1, Fc2), right anterior (RA; Af4, F4, F8), left central (LC; Fc5, C5, C3, Cp5), medial central (MC; Cz, Cp1, Cp2), right central (RC; Fc6, C4, C6, Cp6), left posterior (LP; P3, P7, O1), medial posterior (MP; Pz, Po3, Po4, Oz), and right posterior (RP; P4, P8, O2) sites.

phase and caudality [F(2,32) 5 0.84, p 5 .44, Z2 5 .05] and between presentation phase and laterality [F(2,32) 5 2.51, p 5 .10, Z2 5 .14]. The three-way interaction among these variables was also not significant [F(4,30) 5 1.40, p 5 .26, Z2 5 .16]. Thus, the differences in waveforms between distractor conditions were the same in the encoding and the retention phase.

Discussion The present results can be summarized as follows: (1) Serial recall was decreased by the auditory distractors, especially by changingstate distractors. In other words, a typical irrelevant sound effect and a typical changing-state effect were observed. (2) Irrespective of whether the distractors were played during encoding or retention, the N1 component was larger in the changing-state condition than in the steady-state condition, indicating that distractor repetition attenuated the auditory N1 component. (3) The N1 was followed by a positive wave. The difference waves between the two distractor conditions showed a positive deflection between 200 and 350 ms. This could be taken as evidence that the changing-state distractors elicited a P3a and that these distractors captured attention. (4) The positivity was followed by a late negative wave. This result fits to findings in auditory distraction paradigms in which a late negativity was associated with a reorienting of attention to the primary task (e.g., Escera et al., 2001). Thus, this late negativity could be associated with an attentional reorienting response. However, previous studies have found a frontal distribution of the RON. Thus, the central scalp distribution of the late negative component observed here should not be counted as clear evidence of the reorienting hypothesis. Furthermore, the RON has been found to be time-locked to the visual target stimuli (Escera et al., 2001), which may be a reason why this component is less well defined in the present experiment. Berti (2008a) has suggested that a late negativity with parieto-central maximum following an unexpected change in the auditory environment may reflect a post-

stimulus evaluation process that is activated by the conflict between the automatically triggered response to the auditory event and the task at hand. Alternatively, the late negativity might reflect further processing of the speech distractors. The finding that the N1 amplitude in the changing-state condition is enhanced compared to the N1 response to steady-state distractors replicates the findings of Campbell et al. (2003, 2007) that the N1 increases with the number of distinct distractor tokens in the to-be-ignored sequence. This changing-state effect can be explained by stimulus-specific refractoriness of the N1. It has been established that part of the N1 is generated by neuron populations that respond to specific features of the auditory stimuli (frequency, duration, etc.; Na¨a¨ta¨nen, 1990; Na¨a¨ta¨nen & Picton, 1987). The responsiveness of these neuronal populations attenuates with stimulus repetition (Opitz et al., 2005; Schro¨ger & Wolff, 1998). Thus, the more changing states (distinct distractors) in a sequence, the less feature-specific refractoriness occurs. As a consequence, the N1 response to changing-state distractors is amplified in comparison to the N1 response to steady-state distractors (Campbell et al., 2003, 2007). It is often assumed that the N1 response to auditory distractors reflects a call-for-attention mechanism that triggers an attention switch if it exceeds a certain threshold (Na¨a¨ta¨nen, 1990; Na¨a¨ta¨nen & Picton, 1987). Consistent with this assumption, a large P3a response followed the early negativity in the changingstate condition. This P3a is typically associated with an orienting response to the eliciting stimulus (Friedman et al., 2001) and thus presumably indicates an attention switch to the auditory modality. The P3a to auditory distractors often occurs alongside a decrement in primary task performance (Escera et al., 1998, 2001; Escera & Corral, 2007). Consistent with these earlier findings, the present experiment shows that changing-state distractors elicited a P3a and also disrupted working-memory performance more than steady-state distractors. Thus, the results suggest that an attention switch to the auditory distractors contributes to the irrelevant sound effect.

ERP correlates of the irrelevant sound effect This interpretation of the results is consistent with the conclusion of Campbell and colleagues that ‘‘supratemporal N1 generators may play a role in memory disruption, but the elicitation of this component is not a sufficient condition for memory disruption’’ (Campbell et al., 2003, p. 44). Inconsistent with the assumption that the N1 generating processes are sufficient to elicit an irrelevant sound effect, Campbell and colleagues (2003) found a dramatic increase in N1 amplitude when the distractor set size (that is, the number of unique distractor items in a to-beignored sequence) was increased from 1 to 2, but no corresponding increase in interference. A significant increase in interference was found when distractor set size was increased from 2 to 5, which also caused a further but much less pronounced increase in N1 amplitude. Campbell and colleagues concluded that the processes underlying the N1 may be a necessary rather than sufficient precondition for irrelevant-sound disruption. The present results suggest that the supratemporal N1 generators represent a call for attention, but the decrement in serial-recall performance may be due to the attention switch that is reflected in the P3a component and that is only elicited when the N1 exceeds a certain threshold (see also Rinne et al., 2006). In our data, positivity onset was relatively early (200 ms). One concern might thus be that the early part of the P3a may overlap with a P2 component of the event-related potential. The function of the processes reflected in the P2 is less clear. The P2 in response to target stimuli is often interpreted as reflecting a stimulus classification process preceding the P3. In selective attention paradigms, the P2 is frequently thought to reflect inhibition of information processing that serves to protect against interference (see Crowley & Colrain, 2004, for an overview). Thus, one may speculate whether the enhanced positive deflection in the early part of the 200–350 ms time window was caused by an increase in the P2 in response to changing-state distractors. However, this interpretation is not supported by the present data for a number of reasons. First, it has been previously suggested that the P2 can be distinguished from the P3a by their scalp topography. The P2 has a centro-parietal maximum, whereas the P3a is a more fronto-central component (Ceponiene, Lepisto¨, Soininen, Aronen, Alku, & Na¨a¨ta¨nen, 2004; Ceponiene, Rinne, & Na¨a¨ta¨nen, 2002). Figure 5 illustrates the topographical distribution of the ERP response associated with steady-state and changing-state distractors. As can be seen, the ERP response to the steady-state distractors has a fronto-central maximum. The ERP response in the changing-state condition shows a different scalp topography with a more pronounced frontal maximum. Note that the difference between the steady-state condition and the changing-state condition extends over an interval of 150 ms (and is clearly significant in the 275–350 ms window [F(1,33) 5 10.39, po.01, Z2 5 .24]). Thus, although it cannot be excluded that the positive deflection immediately following the N1 to steady-state distractors may reflectFin partFprocesses associated to a P2 component, the increase in positivity in the changing-state condition is most plausibly due to a P3a. This interpretation of the results fits to findings in the oddball paradigm in which the P2 to repetitive standard stimuli and the P3a to deviants or novel stimuli differ in their scalp topographies (Ceponiene et al., 2004, 2002). Using magnetoencelophalographic recording, Alho, Winkler, Escera, Huotilainen, Virtanen, et al. (1998) have shown that the early P3a to deviants and the P2 to standard stimuli have different generator source locations. Furthermore, our P3a correlate in the changing-state condition fits to results previously reported in the literature. Parallel to the

1189

Figure 5. The maps illustrate the topographical distribution of the eventrelated potentials in response to steady-state and changing-state distractors between 200–350 ms (200–274 ms, and 276–350 ms).

present findings, it has been found that varying (changing-state) tones comprising no regularity can elicit a P3a response and a decrease in behavioral performance in a modified oddball paradigm (Bendixen et al., 2007). Interestingly, in the Bendixen et al. study the P3a to varying (changing-state) tones had also a rather early latency that was in the order of magnitude of the latency observed here. A similarly early latency of the P3a to auditory deviants has been observed under focused attention to the visual channel (MullerGass, Macdonald, Schro¨ger, Sculthorpe, & Campbell, 2007). In line with our interpretation of the present results, these results were interpreted as providing support for the hypothesis that varying auditory stimuli can draw attention away from a primary task. Another interesting question is why Campbell et al. (2003, 2007) failed to observe a P3a response to the auditory distractors, although they obtained (changing-state) irrelevant sound effects. The simplest explanation for this discrepancy seems to be that the fast presentation rate that was used in those studies (the onsets of two consecutive stimuli was 350 ms and 327 ms, respectively) reduced the chance of finding significant differences in the amplitudes of later components. However, alternative explanations should also be considered. Campbell et al. (2007) used nonspeech tones as auditory distractors, whereas the present study used speech distractors. It is commonly observed that speech interferes more with serial recall than other distractor material (Buchner et al., 2008; LeCompte et al., 1997). Thus, the speech sounds used in the present study may have been more attentiongrabbing than the sine wave sounds used by Campbell et al. (2007). Note, however, that, at least in terms of the standardized effect size, the changing-state effect was of comparable size in the present study (Z2 5 .37) and in Campbell et al. (2007; Z2 5 .38). The decreased presentation rate of the auditory distractors

1190

R. Bell et al.

in our study may also have amplified the N1 amplitudes in comparison to previous studies (by decreasing the refractoriness of the N1 generators) and thereby may have increased the likelihood of eliciting attention switches to these stimuli. It would be interesting to manipulate the features of the irrelevant sound (distractor type, duration, presentation rate) directly to see how these variables relate to the ERP correlates and to the amount of memory disruption. This might help to determine whether attention capture in terms of the P3a effect is the main determinant of irrelevant-sound disruption or whether processes that are associated with the N1 may also contribute to the interference effect. By showing that an increase in P3a amplitude accompanies the memory disruption, the present results support the attentional-capture account of the irrelevant sound effect that is based on the embedded-processes model (Cowan, 1995). According to this view, changes in the auditory environment capture the focus of attention. Distractors that are of relevance for the individual may be more prone to capture attention than other distractors (Buchner et al., 2004, 2006). Attention capture interferes with serial recall because attentional resources are needed to keep the to-be-remembered items in a highly accessible state. By supporting an attentional interpretation of the irrelevant sound effect, the present results are consistent with other studies examining

electrophysiological correlates of the irrelevant sound effect. For example, Weisz and Schlittmeier (2006) found a decrease in the N1 to the visual targets and a subsequently reduced theta response at right prefrontal electrodes in a changing-state distractor condition. The findings were interpreted as a sign of reduced attentional resources that were available for the processing of the to-be-remembered items during encoding and retention. Based on these findings, Weisz and Schlittmeier suggested that attention may play a key role in explaining the irrelevant sound effect. In summary, the present results replicate the findings of Campbell et al. (2003, 2007) that an increase in the disruption of serial recall due to a greater variability in the irrelevant auditory stream covaries with an increase in the auditory N1. Extending previous findings (Campbell et al., 2003, 2007), we also found that the amplitude of the subsequent P3a wave increased with increasing amounts of interference, supporting the attentioncapture account of the irrelevant sound effect. Given the obvious parallels between the factors influencing the amount of interference in the irrelevant-sound paradigm and the factors determining ERP responses to acoustic sound changes, we think that it is important to integrate these two lines of research. We hope that the present study may provide a step forward in this direction.

REFERENCES Alho, K., Winkler, I., Escera, C., Huotilainen, M., Virtanen, J., Jaeaeskelaeinen, I. P., et al. (1998). Processing of novel sounds and frequency changes in the human auditory cortex: Magnetoencephalographic recordings. Psychophysiology, 35, 211–224. Baayen, R. H., Piepenbrock, R., & van Rijn, H. (1993). The CELEX lexical database (Release 1). Philadelphia, PA: Linguistic Data Consortium, University of Pennsylvania [Distributor]. Baddeley, A. D., & Hitch, G. J. (1974). Working memory. In G. Bower (Ed.), The psychology of learning and motivation: Advances in research and theory (pp. 47–89). New York: Academic Press. Bell, R., & Buchner, A. (2007). Equivalent irrelevant-sound effects for old and young adults. Memory & Cognition, 35, 352–364. Bell, R., Mund, I., & Buchner, A. (in press). The disruption of short-term memory by distractor speech: Does content matter? Quarterly Journal of Experimental Psychology. Bendixen, A., Roeber, U., & Schro¨ger, E. (2007). Regularity extraction and application in dynamic auditory stimulus sequences. Journal of Cognitive Neuroscience, 19, 1664–1677. Berti, S. (2008a). Cognitive control after distraction: Event-related brain potentials (ERPS) dissociate between different processes of attentional allocation. Psychophysiology, 45, 608–620. Berti, S. (2008b). Object switching within working memory is reflected in the human event-related brain potential. Neuroscience, 434, 200–105. Buchner, A., Bell, R., Rothermund, K., & Wentura, D. (2008). Sound source location modulates the irrelevant sound effect. Memory & Cognition, 36, 617–628. Buchner, A., & Erdfelder, E. (2005). Word frequency of irrelevant speech distractors affects serial recall. Memory & Cognition, 33, 86–97. Buchner, A., Irmen, L., & Erdfelder, E. (1996). On the irrelevance of semantic information for the ‘‘irrelevant speech’’ effect. Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 49, 765–779. Buchner, A., Mehl, B., Rothermund, K., & Wentura, D. (2006). Artificially induced valence of distractor words increases the effects of irrelevant speech on serial recall. Memory & Cognition, 34, 1055–1062. Buchner, A., Rothermund, K., Wentura, D., & Mehl, B. (2004). Valence of distractor words increases the effects of irrelevant speech on serial recall. Memory & Cognition, 32, 722–731. Campbell, T., Beaman, C. P., & Berry, D. C. (2002). Auditory memory and the irrelevant sound effect: Further evidence for changing-state disruption. Memory, 10, 199–214. Campbell, T., Winkler, I., & Kujala, T. (2007). N1 and the mismatch negativity are spatiotemporally distinct ERP components: Disruption

of immediate memory by auditory distraction can be related to N1. Psychophysiology, 44, 530–540. Campbell, T., Winkler, I., Kujala, T., & Na¨a¨ta¨nen, R. (2003). The N1 hypothesis and irrelevant sound: Evidence from token set size effects. Cognitive Brain Research, 18, 39–47. Ceponiene, R., Lepisto¨, T., Soininen, M., Aronen, E., Alku, P., & Na¨a¨ta¨nen, R. (2004). Event-related potentials associated with sound discrimination versus novelty detection in children. Psychophysiology, 41, 130–141. Ceponiene, R., Rinne, T., & Na¨a¨ta¨nen, R. (2002). Maturation of cortical sound processing as indexed by event-related potentials. Clinical Neurophysiology, 113, 870–882. Cowan, N. (1995). Attention and memory: An integrated framework. London: Oxford University Press. Crowley, K. E., & Colrain, I. M. (2004). A review of the evidence for P2 being an independent component process: Age, sleep and modality. Clinical Neurophysiology, 115, 732–744. Elliott, E. M. (2002). The irrelevant-speech effect and children: Theoretical implications of developmental change. Memory & Cognition, 30, 478–487. Escera, C., Alho, K., Winkler, I., & Na¨a¨ta¨nen, R. (1998). Neural mechanisms of involuntary attention to acoustic novelty and change. Journal of Cognitive Neuroscience, 10, 590–604. Escera, C., & Corral, M. J. (2007). Role of mismatch negativity and novelty-P3 in involuntary auditory attention. Journal of Psychophysiology, 21, 251–264. Escera, C., Yago, E., & Alho, K. (2001). Electrical responses reveal the temporal dynamics of brain events during involuntary attention switching. European Journal of Neuroscience, 14, 877–883. Friedman, D., Cycowicz, Y. M., & Gaeta, H. (2001). The novelty P3: An event-related brain potential (ERP) sign of the brain’s evaluation of novelty. Neuroscience & Biobehavioral Reviews, 25, 355–373. Gratton, G., Coles, M. G. H., & Donchin, E. (1983). A new method for off-line removal of ocular artifact. Electroencephalography & Clinical Neurophysiology, 55, 468–484. Jones, D. M. (1993). Objects, streams, and threads of auditory attention. In L. Weiskrantz & A. D. Baddeley (Eds.), Attention: Selection, awareness, and control: A tribute to Donald Broadbent (pp. 87–104). New York: Oxford University Press. Jones, D. M., & Macken, W. J. (1993). Irrelevant tones produce an irrelevant speech effect: Implications for phonological coding in working memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 19, 369–381.

ERP correlates of the irrelevant sound effect Jones, D. M., & Macken, W. J. (1995). Phonological similarity in the irrelevant speech effect: Within- or between-stream similarity? Journal of Experimental Psychology: Learning, Memory, and Cognition, 21, 103–133. Jones, D. M., Macken, W. J., & Murray, A. C. (1993). Disruption of visual short-term memory by changing-state auditory stimuli: The role of segmentation. Memory & Cognition, 21, 318–328. Jones, D. M., Madden, C., & Miles, C. (1992). Privileged access by irrelevant speech to short-term memory: The role of changing state. Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 44, 645–669. Kopp, F., Schro¨ger, E., & Lipka, S. (2004). Neural networks engaged in short-term memory rehearsal are disrupted by irrelevant speech in human subjects. Neuroscience, 354, 42–45. Kopp, F., Schro¨ger, E., & Lipka, S. (2006). Synchronized brain activity during rehearsal and short-term memory disruption by irrelevant speech is affected by recall mode. International Journal of Psychophysiology, 61, 188–203. Lange, E. (2005). Disruption of attention by irrelevant stimuli in serial recall. Journal of Memory and Language, 53, 513–531. LeCompte, D. C., Neely, C. B., & Wilson, J. R. (1997). Irrelevant speech and irrelevant tones: The relative importance of speech to the irrelevant speech effect. Journal of Experimental Psychology: Learning, Memory, and Cognition, 23, 472–483. Martı´ n-Loeches, M., & Sommer, W. (1998). Testing models of the irrelevant speech effect on working memory with event-related potentials. Zeitschrift fuer Psychologie, 206, 1–22. Mayr, S., Niedeggen, M., Buchner, A., & Orgs, G. (2006). The level of reaction time determines the ERP correlates of auditory negative priming. Journal of Psychophysiology, 20, 186–194. Mayr, S., Niedeggen, M., Buchner, A., & Pietrowsky, R. (2003). ERP correlates of auditory negative priming. Cognition, 90, B11–B21. Miles, C., Jones, D. M., & Madden, C. (1991). Locus of the irrelevant speech effect in short-term memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 17, 578–584. Muller-Gass, A., Macdonald, M., Schro¨ger, E., Sculthorpe, L., & Campbell, K. (2007). Evidence for the auditory P3a reflecting an automatic process: Elicitation during highly-focused continuous visual attention. Brain Research, 1170, 71–78. Na¨a¨ta¨nen, R. (1990). The role of attention in auditory information processing as revealed by event-related potentials and other brain

1191 measures of cognitive function. Behavioral and Brain Sciences, 13, 201–288. Na¨a¨ta¨nen, R., & Picton, T. W. (1987). The N1 wave of the human electric and magnetic response to sound: A review and an analysis of the component structure. Psychophysiology, 24, 375–425. Na¨a¨ta¨nen, R., & Winkler, I. (1999). The concept of auditory stimulus representation in cognitive neuroscience. Psychological Bulletin, 125, 826–859. Neath, I. (1999). Modelling the disruptive effects of irrelevant speech on order information. International Journal of Psychology, 34, 410–418. Opitz, B., Schro¨ger, E., & von Cramon, D. Y. (2005). Sensory and cognitive mechanisms for preattentive change detection in auditory cortex. European Journal of Neuroscience, 21, 531–535. Polich, J. (2007). Updating p300: An integrative theory of P3a and P3b. Clinical Neurophysiology, 118, 2128–2148. Rinne, T., Sa¨rkka¨, A., Degerman, A., Schro¨ger, E., & Alho, K. (2006). Two separate mechanisms underlie auditory change detection and involuntary control of attention. Brain Research, 1077, 135–143. Salame´, P., & Baddeley, A. D. (1982). Disruption of short-term memory by unattended speech: Implications for the structure of working memory. Journal of Verbal Learning and Verbal Behavior, 21, 150–164. Schro¨ger, E., & Wolff, C. (1998). Behavioral and electrophysiological effects of task-irrelevant sound change: A new distraction paradigm. Cognitive Brain Research, 7, 71–87. Tremblay, S., Nicholls, A. P., Alford, D., & Jones, D. M. (2000). The irrelevant sound effect: Does speech play a special role? Journal of Experimental Psychology: Learning, Memory, and Cognition, 26, 1750–1754. Valtonen, J., May, P., Ma¨kinen, V., & Tiitinen, H. (2003). Visual shortterm memory load affects sensory processing of irrelevant sounds in human auditory cortex. Cognitive Brain Research, 17, 358–367. Weisz, N., & Schlittmeier, S. J. (2006). Detrimental effects of irrelevant speech on serial recall of visual items are reflected in reduced visual N1 and reduced theta activity. Cerebral Cortex, 16, 1097–1105. Woods, D. L., & Elmasian, R. (1986). The habituation of event-related potentials to speech sounds and tones. Electroencephalography and Clinical Neurophysiology: Evoked Potentials, 65, 447–459. (Received November 17, 2009; Accepted December 22, 2009)