Advances in the Study of Behavior, Volume 12

Advances in

THE STUDY OF BEHAVIOR VOLUME 12

Contributors to This Volume KAREN L. HOLLIS TIMOTHY D. JOHNSTON BARRY R. KOMISARUK CAROLYN A. RISTAU DONALD ROBBINS

Advances in

THE STUDY OF BEHAVIOR Edited by

JAY S. ROSENBLATT Institute of Animal Behavior Rutgers University Newark, New Jersey

ROBERTA. HINDE Medical Research Council Unit on the Development and Integration of Behaviour University Sub-Department of Animal Behaviour Madingley , Cambridge, England COLINBEER Institute of Animal Behavior Rutgers University Newark, New Jersey MARIE-CLAIRE BUSNEL Laboratoire de Physiologie Acoustique Institut National de la Recherche Agronornique Jouy en Josas (78350),France

VOLUME 12 1982

ACADEMIC PRESS A Subsidiary of Harcourt Brace Jovanovich, Publishers

New York London Paris San Diego San Francisco S&o Paulo Sydney Tokyo Toronto

R Y ACADEMIC PRESS,INC. ALL RIGHTS RESERVED. NO PART OF THIS PUBLICATION MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM OR BY ANY MEANS, ELECTRONIC OR MECHANICAL, INCLUDING PHOTOCOPY, RECORDING, OR ANY INFORMATION STORAGE AND RETRIEVAL SYSTEM, WITHOUT PERMISSION IN WRITlNG FROM THE PUBLISHER.

COPYRIGHT @ 1982,

ACADEMIC PRESS, INC.

I11 Fifth Avenue, New York, New

York 10003

United Kirigdoni Ediriofi published by ACADEMIC PRESS, INC. (LONDON) LTD. 24/28 Oval Road, London NWI

7DX

LIBRARY OF CONGRESS CATALOG CARDNUMBER: 64-8031 ISBN 0-12-004512-5 PRINTED IN THE UNITED STATES OF AMERICA 82 83 R4 85

9 8 7 6 5 4 3 2 1

Contents

List of Contributors ........................................................ Preface ................................................................... Erratum ..................................................................

vii ix xi

Pavlovian Conditioning of Signal-Centered Action Patterns and Autonomic Behavior: A Biological Analysis of Function KAREN L . HOLLIS I. I1. I11. IV .

V.

VI .

Introduction ......................................... The Prefiguring Hypothesis ........................... Pavlovian Conditioning and Foraging Behavior .......... Pavlovian Conditioning of Defensive Behavior .......... Pavlovian Conditioning of Reproductive Behavior ....... Ecological Implications of Prefiguring .................. References ..........................................

1

3 6 28 43 48 53

Selective Costs and Benefits in the Evolution of Learning TIMOTHY D. JOHNSTON

I . Introduction ......................................... I1 . Learning and Evolution-Historical Background and Current Concerns .................................... I11. An Ecological Conception of Learning ................. IV . Cost-Benefit Analysis and the Evolution of Adaptations . . V . The Selective Benefits of Learning ..................... VI . The Selective Costs of Learning ....................... VII . Learning and the Adaptive Complex .................... VIII . Implications for the Study of Learning ................. References .......................................... V

65

66 69 70 73 79 92 96 98

vi

CONTENTS

Visceral- Somatic Integration in Behavior. Cognition. and “Psychosomatic” Disease BARRY R. KOMISARUK I . Introduction ......................................... I1. Visceral-Somatic Relationships ....................... I11. Higher Order Integration of Visceral and Somatic Activity ..................................... IV . Visceral Activity and Ideational Imagery ............... V . Toward a Concept of Psychogenic Organic (“Psychosomatic”) Disease ........................... VI . Conclusion .......................................... References ..........................................

107 109 112 123 129 134 135

Language in the Great Apes: A Critical Review CAROLYN A . RISTAU AND DONALD ROBBINS

I. I1. 111. IV . V. VI . VII . VIII . IX . X. XI .

Introduction ......................................... Brief History of the Ape Language Projects ............. Theoretical Issues .................................... The Signing Apes .................................... Artificial Lexicons ................................... Investigations into Meaning ........................... Investigations into Mental States ....................... Relation to Animal Cognition and Natural Animal Communication ...................................... Implications for Human Language and Cognitive Development ........................................ Problems Raised by the Language Research and Suggestions for Future Research ....................... Concluding Statements ............................... References ..........................................

Index ..................................................................... Contents of Previous Volumes ...............................................

142 143 145 155 178 188 207 219 230 237 245 247

257 261

List of Contributors Numbers in parentheses indicate the pages on which the authors’ contributions begin.

KAREN L. HOLLIS,* Animal Behaviour Research Group, Department of Zoology, University of Oxford, Oxford, England ( 1 ) TIMOTHY D. JOHNSTON, Research Branch, North Carolina Division of Mental Health, Raleigh, North Carolina 2761 1 (65) BARRY R. KOMISARUK, Institute of Animal Behavior, Rutgers University, Newark, New Jersey 07102 (107) CAROLYN A. RISTAU, The Rockefeller University, New York, New York 10021 (141) DONALD ROBBINS, Division of Social Sciences, Fordham University at Lincoln Center, New York, New York 10023 (141)

*Present address: Department of Psychology, University of Toronto, Toronto, Ontario, Canada MSS 1Al. vii

This Page Intentionally Left Blank

Preface With the publication of the twelfth volume of Advances in the Study of Behavior, we wish to restate in more contemporary terms the aims stated in the original preface, namely, to serve “. . . as a contribution to the development of cooperation and communication among scientists in our field.” Since that preface was written in 1%5, an increasing number of scientists from disciplines as widely separated as behavioral ecology and the biochemistry of behavior have become engaged in the study of animal behavior, employing the specialized techniques and concepts of their disciplines. Even then, the boundaries of ethology and comparative psychology were no longer distinct: now they have been merged with broader syntheses of social and individual functioning and have together provided the bases for studies of the neural and biochemical mechanisms of behavior. New vigor has been given to traditional fields of animal behavior by their coalescence with closely related fields and by the closer relationship that now exists between those studying animal and human subjects. Scientists engaged in studying animal behavior now range from ecologists through evolutionary biologists, geneticists, endocrinologists, ethologists, and comparative and developmental psychologists, to neurophysiologists and neuropharmacologists. The task of developing cooperation and communication among scientists whose skills and concepts necessarily differ in accordance with the diversity of the phenomena they study has become more difficult than it was at the inception of this publication. Yet the need to do so has become even greater as it has become more difficult. The Editors and publisher ofAdvances in the Study ofBehavior will continue to provide the means by publishing critical reviews of research in our field, by inviting extended presentations of signifcant research programs, by encouraging the writing of theoretical syntheses and reformulations of persistent problems, and by highlighting especially penetrating research that introduces important new concepts.

ix

This Page Intentionally Left Blank

Erratum Advances in the Study of Behavior Volume 11 Page 34

Figure 28 should appear as follows: 20

-----

I5

-

spermotoqeness promo testosieron

- - - - - -- - - - -

MAS

-

wl?nwlOmO/

-

rpcrmolldr/

devzkwd cpdIdymlS

-

wrmomoo

- IpmurldS - P'mory

s9ermotocy~

- wmwrlwpn~ \

'.-

fully rqmrad

FIG.28. Hormonal control of male reproductive behavior in Anolis carolincnsis. Top: Relationship between plasma testosterone levels and behavioral and physiological events in the annual reproductive cycle of the male lizard Anolis carolinensis. Modified from Crews (1975). The number of animals assayed in each monthly sample are as follows: N = 6; D = 10; J = 14; F = 10; A = 10; M = 17; J = 8; S = 12. Tokarz and Crews (unpublished). Bottom: Effect of castration and androgen (testosterone) replacement therapy on the sexual and aggressive behavior of the male lizard, Anolis corolinrnsis. In these experiments. sexually active males were castrated after baseline levels of sexual and aggressive behavior were determined. Two weeks following castration. males were given Silastic implants (0.06 cm i.d. x 0.12 cm 0.d.) containing testosterone subcutaneously. Males were tested daily first with a male intruder and then with a female intruder for 15 min each. Mean and SEM are shown; n = 12. From Crews (1979a) with permission of the Society for the Study of Reproduction. xi

This Page Intentionally Left Blank

Advances in

THE STUDY OF BEHAVIOR VOLUME 12

This Page Intentionally Left Blank

ADVANCES IN THE STUDY OF BEHAVIOR. VOL I2

Pavlovian Conditioning of Signal-Centered Action Patterns and Autonomic Behavior: A Biological Analysis of Function KARENL. HOLLIS* ANIMAL BEHAVIOUR RESEARCH GROUP DEPARTMENT OF ZOOLOGY UNIVERSITY OF OXFORD OXFORD, ENGLAND

I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11. The Prefiguring Hypothesis . . 111. Pavlovian Conditioning and F A. Locomotory Search and A B. Consummatory and Food-Procuring Behavior ...................... C. Pavlovian Processes in the Development of Food Recognition D. Conditioning of Digestive E. Pavlovian Processes in the Rejection-r Ingestion-f Toxins . . . . . . . . IV. Pavlovian Conditioning of Defensive Behavior . . . . . . . . . . . . . . . . . . . . . . . . A. Interspecific (Antipredator) Defense .............................. B. Interspecific Defense and the Backw C. Intraspecific Defense . . . . . . . . . . . . . . . . . . V. Pavlovian Conditioning of Reproductive ...................... A. Courtship ...................... ...... B. Parental Behavior ....................................... VI. Ecological Implication figuring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Prefiguring and Non-Pavlovian Learning B. Naturally Occurring Conditional Stimuli as Learned Releasers . . . . . . . . . C. Concluding Comments . . . ...................... References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..................

I.

1

8

19 28 28

43

44 46 48 49

50 51

53

INTRODUCTION

Consider the difficulties an animal faces if it must actively search for food, water, or for a mate. Where and when can food be found and mates secured? Likewise, where and when will predators most likely strike and rivals appear? *Present address: Department of Psychology, University of Toronto, Toronto, Ontario, Canada, M5S 1Al. I

Copyright 0 1982 by Academic FYess. Inc. All rights of reproduction in any form reserved. ISBN 0-12-004512-5

2

KAREN L . HOLLIS

For many animals the strategic difficulties of searching-and remaining vigilant to the search of others-would be insuperable obstacles to survival if the environment were a capricious one. But given that an animal could predict where and when food would be available, such information would provide considerable savings in foraging time and energy reserves. And, if an animal could locate predators based upon, say, some sort of warning signal, the danger of a surprise attack would be negated altogether. Pavlovian conditioning is a mechanism for such predictions. But Pavlovian conditioning appears to involve more than mere “predictability” or “information transfer,” as might be expected. The result of such conditioning is the eventual elicitation of a response, the conditional response or CR, which actually precedes the occurrence of the biologically important event and which, if it is a skeletal behavior, is often directed at the conditionul stimulus (CS).The CR would not only appear to be “jumping the gun” but would appear to be somewhat misdirected. That is, in a typical Pavlovian conditioning experiment (Mackintosh, 1974) the experimenter presents two stimuli to the animal, a CS followed by an unconditional stimulus (US). The US is some biologically relevant event, e.g., food, water, or noxious stimulation. The US was so-named by Pavlov ( 1927) because it unconditionally elicits a stimulus-specific response, the unconditional response (UR). Pecking is a UR to food in pigeons, for example. The CS is sometimes described as a “neutral” stimulus because it does not elicit a conditional response prior to training; that is, the CR is conditional upon the CS-US pairing procedure. The animal need not perform some response to obtain the US, as it must in an instrumental (Skinnerian) conditioning paradigm. (These two paradigms, as well as experimental justification for distinguishing between them, are discussed in a later section.) The procedural hallmark of Pavlovian conditioning is that the CS and US are presented to the animal independently of its behavior. Nonetheless, after sufficient training of this nature, the CS elicits a conditional response which precedes US presentation. This anticipatory conditional response is, more often than not, energetically costly. For example, a territorial male Betra splendens will forcefully attack a stimulus which in the past has preceded the appearance of a rival (Thompson and Sturm, 1965). Similarly, if the illumination of a key light is reliably followed by food, a pigeon will peck quite vigorously at the key, regardless of the fact that no such response is required in this situation (Brown and Jenkins, 1968). Why does conditioning result in an anticipatory response? What advantage does an animal gain by responding to cues which reliably accompany US events when these USs are, themselves, already effective in eliciting the appropriate response? One might ask why the classically conditioned pigeon does not simply go to the food magazine and wait, or why the classically conditioned Betta does not conserve its energy for the real aggressive encounter. That is, while no one would question whether the predictive function of Pavlovian conditioning is advantageous, why

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

3

does the operation of this mechanism also result in a response whose true referent has not yet appeared? In this article I will explore the biological function of the anticipatory conditional response. Psychologists’ views of learning phenomena, including Pavlovian conditioning, have changed radically in recent years. Among other things, learning is no longer believed to provide an endless source of plasticity (Shettleworth, 1972; Hinde and Stevenson-Hinde, 1973). Rather, an ability to learn merely permits the animal to extend its utilization of the species-specific response repertoire to new situations. From a biological standpoint, this ability is, in itself, the product of its species evolutionary history. Ultimately, then, the explanation of all learning phenomena must be made consistent with ecological and ethological considerations. An analysis of the function of Pavlovian conditional responses is but one attempt to do so.

11.

THEPREFIGURING HYPOTHESIS

I suggest that the biological function of classically conditioned responding, which I will call prefiguring, is to enable the animal to optimize interaction with the forthcoming biologically important event (US). The performance of a CR, although energetically costly, allows the animal better to deal with the US event, and, as such, the CR is essentially preparatory.’ The animal is, of course, already physiologically equipped to respond to biologically important events in the absence of their having been signaled. Nonetheless, the prefiguring hypothesis maintains that anticipatory responses to stimulus events (CSs), which in the past have reliably accompanied those USs, provide a selective advantage over and above the ability to respond to the US alone. In this sense the CR is nor an accidental “false start” of an otherwise adequate conditioning process. The anticipatory conditional response is, itself, the evolutionary ruison d’ttre of Pavlovian conditioning. Of course, the performance of a CR is how psychologists have always measured Pavlovian conditioning. Nonetheless, when the question of function has arisen, greater importance has been attached to the predictive power of the CS-US relationship (Mackintosh, 1979) where “prediction” appears to take on the ordinary dictionary meaning of “knowing in advance.” However, to predict that it will rain tomorrow afternoon is one matter; to take my umbrella tomorrow morning is quite another. ’The use of the word preparatory here is purely descriptive and refers to the funcrion of the CR, not the mechanism whereby the CR is produced. That is, I do not intend that the term preparatory be theoretically synonymous with its use in some animal learning contexts, e.g., the “preparatoryresponse” hypothesis (Perkins, 1968; see also Prokasy, 1965). There the term is used to refer to a non-Pavlovian mechanism of learning. This issue is further discussed in Section IV,A,2.

4

KAREN L . HOLLIS

It would seem that a conditional response to experimenter-selected CS events, like illuminated discs and ticking metronomes, would offer little in the way of a selective advantage, and, indeed, in some cases would seem to be maladaptive. In the “long box,” for example (Hearst and Jenkins, 1974), pigeons could be trained in such a way that they invariably missed a brief food presentation because they had been pecking at a CS some distance away (although, of course, no such response was required). However, the fact that we can produce “maladaptive” or “inefficient” behavior in the laboratory by distorting the temporal and/or spatial relationships between stimuli is irrelevant to the functional argument presented here. Naturally occurring signals, with which evolution has dealt, would hardly be so arbitrarily related to US events. The bird which approaches and pecks an illuminated spot on the chamber wall because in the past this CS has been paired with food presentation, is also the bird which, in the wild, is likely to approach and peck certain small “spots” on the forest floor previously associated with food. Here, however, these spots may be the tell-tale cues that buried insects unavoidably leave behind. If a functional approach to Pavlovian conditioning, a phenomenon today almost exclusively concerned with causal mechanisms, somewhat muddles the boundaries between biology and psychology, the idea is not wholly original. Pavlov himself was a functionalist interested in “phenomena of adaptation (1928, p. 83). Pavlov’s writings and those of his students clearly reflect those functional considerations. However, E. A. Culler perhaps said it best:

”

[The] concept of a self-regulating mechanism has been amply documented by Cannon (1932). Constancy of water content, of salt content, . . . maintenance of body temperature are but special forms of a pervasive ‘homeostasis.’ Admirable as these autonomic stabilizers are, they do not approach in range and flexibility the adjustive mechanisms which nature has provided in [Pavlovian]conditioning. . . . [If a] UR were his only recourse, the animal would still be forced to wait in every case for the stimulus to arrive before beginning to meet it. The veil of the future would hang just before his eyes. Nature began long ago to push back the veil. Foresight proved to possess high survival value, and conditioning is the means by which foresight was achieved.. . . The CR, in brief, is nature’s way of getting ready for an important stimulus. The salivaty secretion prepares the mouth for reception of food and gastric secretion for its proper ingestion. . . . Conditioned lid-closure protects the cornea from a blow or jet of air. Conditioned iridic reflex saves the retina from undue stimulation. Conditioned galvanic skin-response . , . prepares for seizing and manipulating the stimulus, and so on (Culler, 1938. pp. 134-136).

And so on, indeed. In the remainder of this article I will attempt to show, specifically, how “nature’s way of getting ready” is accomplished by Pavlovian conditioning of the animal’s repertoire of adaptive responses. This applies not only to the conditioning of autonomic responses, the domain of Pavlovian conditioning known to an earlier era, but to the more recently discovered conditioning of complex skeletal behaviors (autoshaping or signtracking) as well. Moreover,

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

5

for any given US, I will stress the notion that the conditional response is not one but actually a battery of responses, both autonomic and skeletal, some similar to the UR and some opposite in direction, and together as one unit they function to insure optimization of biologically relevant events. A functional analysis of behavior involves determining the ultimate evolutionary advantage of indulging in a particular behavior which, in theory, would be measured in terms of an animal’s inclusive fitness (Hinde, 1975). Thus, the prefiguring hypothesis suggests that, by responding to CSs in anticipation of the US event, an animal increases its own reproductive potential or that of its close kin. Fitness is often difficult to measure, however, because many behaviors are far removed in time from actual reproductive gains. Under such circumstances a currency is chosen which, it is assumed, will ultimately be translated into reproductive success. The behavior in question is then measured in terms of its net contribution to this currency (McCleery, 1978). For example, a behavior which enables an animal to obtain more food is recognized as adaptive because feeding efficiency would eventually be translated into a reproductive advantage. Feeding efficiency and other such currencies will be employed in this article as a means with which to assess the contribution of prefiguring to fitness. Nonetheless, in his treatise on adaptation, G. C. Williams (1966) warns us against “unwarranted uses of the concept of adaptation” (p. 11). A benefit may be the result of chance, not design, and we must be careful to distinguish adaptations from mere fortuitous effects (see also Lewontin, 1979; Maynard Smith, 1978). Thus, although one may be able to demonstrate successfully a benefit which results from the performance of a CR, this in itself is insufficient. One must also demonstrate that conditional responses are not merely accidental by-products of the conditioning process. That is, an animal’s nervous system might be such that it cannot help but anticipate the US, even though anticipation is not the function of Pavlovian conditioning. This simplistic account of the CR-that it is an accidental false start-seems unlikely: We will see that the CR does not always involve the same response system as the UR (Wasserman, 1973; Jenkins et al., 1978) and even where it does, it is not always similar to the UR (Bykov, 1959). Indeed, in some cases the CR is in the opposite direction of the UR (S. Siegel, 1979a). Moreover, the form of the CR changes over the course of conditioning (Wasserman, 1973) and varies with both the intensity (Gray, 1965) and type of the CS (Holland, 1977), and even the timing of CS and US events (Holland, 1980; Rescorla, 1980b). Most importantly, however, the ability of the prefiguring hypothesis to predict successfully such perturbations of the anticipatory conditional response suggests that a functional analysis is correctly centered on the CR. Finally, functional explanations do not compete with causal explanationsthey involve orthogonal questions whose answers complement and support one another ( N . Tinbergen, 1963; see Shettleworth, in press, for a discussion of this

6

KAREN L. HOLLIS

topic with regard to learning). A functional argument cannot explain how Pavlovian conditioning works. On the other hand, as the remainder of this article will reveal, the prefiguring hypothesis is able to make novel predictions about conditioned behavior and to provide a systematic account of the many different “response rules” governing the form, the appearance, and the timing of the Pavlovian conditional response.

111.

PAVLOVIAN CONDITIONING A N D FORAGING BEHAVIOR

Obtaining food and water is a long and complicated process involving a sequence of many different behavior patterns, both skeletal and autonomic. The process begins with physiological changes responsible for hunger or thirst. These are followed by locomotory sequences of behavior, sometimes called appetitive behavior (Craig, 1918), which bring the animal in contact with food and water sources. Later, when food or water is encountered, consummatory behaviors are evoked. Concurrent with these activities, and continuing long past them, physiological changes involving ingestion are taking place. Some of the more familiar of these physiological responses include salivation (in some animals), secretion of gastric and pancreatic fluids, and increased gastrointestinal motility. Pavlovian signaling operations have been demonstrated in all of these various behavioral contexts. Pavlov himself was primarily interested in digestion. But not only was the prototypical Pavlovian conditioning experiment concerned with ingestion; much of the autoshaping literature today is an investigation of conditioned food-related behaviors. Although this section is not intended to provide an exhaustive discussion of the literature, it will attempt to show through selected animal learning and animal behavior experiments how an animal might incorporate a signaling operation in the various aspects of its search for food, including locomotory search and approach behavior, consummatory and food-procuring behavior, and autonomic behavior related to food ingestion. Finally, the importance of Pavlovian conditioning in the development of food selection and in the avoidance-or consumption-of poisons will be discussed. A.

LOCOMOTORY SEARCHA N D APPROACH BEHAVIOR

Because Pavlovian conditioning has been studied within the confines of laboratory apparatus, locomotory search behaviors are necessarily restricted. Nonetheless, a large body of research (for reviews, see Hearst and Jenkins, 1974; Schwartz and Gamzu, 1977) suggests that specific orienting, approaching, and searching behaviors (which I will shorten to “approach”) are elicited by CSs paired with food and water. In pigeons, for example, visually localizable signals like the illumination of a key light which is followed by food presentation, evoke

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

7

approach behavior (Staddon and Simmelhag, 1971; Hearst and Jenkins, 1974). In these experiments, the pigeon is not reinforced for approach in the sense that food is made contingent upon an approach response. The pigeon approaches the key only because key illumination precedes food presentation. Pavlovian conditional stimuli paired with food USs have been reported to evoke approach in several species of fish (Squier, 1969; Woodard and Bitterman, 1974) and birds (Moore, 1973; Neuringer and Neuringer, 1974; Gardner, 1969) plus a wide variety of mammals including rats (Peterson et a l . , 1972; Atnip, 1977), cats (Grastyan and Vereczkei, 1974), monkeys (Sidman and Fletcher, 1968), and dogs (Zener, 1937; Jenkins et a l . , 1978; S . G . Smith and Smith, 1971). Although Pavlov (1927) noted that a novel stimulus would also elicit head and body movements directed toward it, Patten and Rudy (1967) have shown that this “what is it?” or “orienting reflex” declines over repetitions (i.e., habituates) if the CS is not followed by food, whereas approach and orienting behaviors actually increase if the CS signals food presentation. Moreover, approach behavior develops even though, in an artificially arranged situation, such behavior actually interferes with the procurement of food. Grastyan and Vereczkei (1974) trained cats to run from a starting position to the location of a food-delivery device when presented with an auditory signal (clicking sound). The auditory CS remained on until the cats obtained the food. The speaker which produced the CS was, in different groups of animals, either near the food device, or behind the starting position (i.e., 180”from the food source), or in the arm of a T maze leading away from the food site. Approach to this auditory CS developed in all groups, even though this behavior disrupted food acquisition in those animals whose food CS was spatially separated from the US source. These data make a very important point. The finding that the approach CR was established in spite of its adverse consequences in the spatially separated CS conditions makes it hard to believe that the reason these behaviors are established in ordinary Pavlovian conditioning experiments is merely because the behavior has beneficial consequences; that is, approach responses are “rewarded” with food. Rather, these data suggest that approaching a CS associated with food is established because of the temporal relationship between the CS and US. Approach is “uncontrollably” elicited by signals paired with food. Pavlovian conditioning is part of the animal’s adaptive repertoire with which it locates food and, as such, it is difficult to eliminate even when the behavior is counterproductive. Only after extended training were Grastyan and Vereczkei able to do so. (The view that is being maintained here is that behavior is being classically conditioned, not “instrumentally” conditioned through implicit operant contingencies. This Pavlovian vs instrumental distinction will be further explored-and experimentally justified-in the next section.) In the context of the prefiguring hypothesis, naturally occurring CSs for food and water serve, initially, simply to guide and direct an animal’s searching

8

KAREN L. HOLLIS

behavior. If approach is elicited by signals which in the past have reliably accompanied food, then such conditioning can result in the animal searching in highly likely places. Knowing where food items are likely to be found need be little more than moving toward signals which the prevailing hunger motivation has transformed into salient and attractive stimulus events. Although much recent literature has elaborated how the type of CS (e.g., auditory vs visual) may influence the exact form of the resulting CR (Holland, 1977, 1980), discretely localized CSs invariably elicit some sort of taxic approach and/or orienting behavior. Let us imagine an animal which is engaging in its species-specific searching behavior, say, a blue tit turning over leaves in its search for beechnuts, a badger searching a pasture for earthworms, or a honeycreeper probing newly opened flowers for nectar. When in the course of this search the animal encounters food signals, the search pattern is channeled in the direction of the signal location and searching behavior becomes more efficient. It is interesting to note here that hungry animals also withdraw from stimuli that predict the absence of food (Wasserman el d.,1974; Hearst and Franklin, 1977; Staddon and Simmelhag, 197 1 ). Thus, Pavlovian conditioning might be viewed (albeit somewhat metaphorically) as a learned taxic response. Indeed, the term signrracking“behavior that is directed toward or away from a stimulus as a result of the relation between that stimulus and the reinforcer or between that stimulus and the absence of the reinforcer” (Hearst and Jenkins, 1974, p. 4)-also emphasizes the taxic influence of learned signal events. Croze (1 970), Royama (1970), and Murton (1971) present evidence that, in the wild, camon crows, great tits, and woodpigeons, respectively, return to locations where they have found food in the past. Whether this behavior is influenced by Pavlovian food CSs poses an interesting question for research.

B . CONSUMMATORY A N D FOOD-PROCURING BEHAVIOR The function of the approach behavior elicited by CSs is, of course, to guide the animal “to places, or objects, or substances” (Moore, 1973, p. 183) which signal food-that is, to locations of food availability. However, these same stimuli elicit CS-directed consummatory and food-procuring behavior as well. If the function of approach CRs appears obvious, the function of these signal-centered action patterns would seem to require further exploration. In the now classic autoshaping experiment with pigeons (Brown and Jenkins, 1968) illumination of an 8-sec white key light (CS) was followed immediately by access to food for 4 sec (US). Between trials the key was unlighted. After several such CS-US pairings, the pigeon approached the CS and began to peck the lighted key in the manner characteristic of pigeons food-pecking. When, later, the autoshaping experiment was repeated with a water US instead of food (Jenkins and Moore, 1973). the water-reinforced birds were observed to “drink”

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

9

rather than food-peck the key. Whereas food-pecks on the key consisted of sharp, vigorous movements with the beak slightly open on contact, water-pecks involved slower. more sustained key contact with the beak almost completely closed. Water-pecks were frequently accompanied by swallowing movements. CS-directed consummatory and food-procuring behavior has been demonstrated in a number of other species as well. Peterson ef al. (1972) reported that rats gnawed, bit, and licked a retractable lever whose insertion into the chamber signaled food delivery. Interestingly, when, instead, the presentation of another rat signaled the delivery of a food US, subject rats approached the signal conspecific and performed social contact and soliciting behavior (Timberlake and Grant, 1975). Dogs displayed individually distinctive action patterns to various visual and auditory CSs, which included licking the CS source (Pavlov, 1934), or performing CS-directed soliciting or hunting (i.e., sight-pointing and flushing) behavior (Jenkins et al., 1978). Archer fish, which obtain food by directing a forceful squirt of water at insects perched on overhanging plants, also squirted a red light CS, located over the water, when it preceded delivery of an insect onto the surface of the water (Waxman and McCleave, 1978). And, according to Breland and Breland (1966), who attempted to train a variety of animals to perform various tricks using food rewards, turkeys, porpoises, and whales will swallow objects which are associated with food reinforcement. Breland and Breland (1961) also report that when they attempted to train a pig to pick up a wooden coin and drop it into a piggy bank, the animal would “repeatedly drop it, root it, drop it again, root it along the way, pick it up, toss it in the air. . . and so on” (p. 683) while a raccoon, similarly trained, would spend its time washing the coins by rubbing them together. And finally, when chickens were required to stand upon a platform for a short time period before receiving food, they began to ground-scratch the floor of the platform. This is exploited by entertainers who exhibit ”dancing” chickens. This summary, though far from being complete, illustrates the most evident characteristic of the CR when discretely localizable CSs are employed: Anticipatory consummatory CRs are isomorphic with the species’ method of obtaining food. In pigeons, forward pecking accompanied by sideways flicking of the head are the means by which buried insects are found on forest floors (Craig, 1918; Moore, 1973). In fowl, foraging behavior consists of ground-scratching and pecking, both behaviors functioning to uncover food objects. Lorenz (1969) made a similar point regarding a Pavlovian conditioning experiment related to him by Howard Liddell, working in Pavlov ’s laboratory: LThe experiment] consisted siniply in freeing from its harness a dog that had been conditioned to salivate at the acceleration in the beat of a metronome. The dog at once ran to the machine, wagged its tail at i t , tried to jump up to it. barked, and so on; in other words, i t showed as clearly as possible the whole system o f behavior patterns serving, in a number of Curiidur, to beg food from a conspecific. It is, in fact, this whole system that is being conditioned in the classical [conditioning] experiment (Loren.?, 1969. p. 47).

10

K A R E N L . HOLLIS

Jenkins er al. (1978) propose that the signal-centered action patterns of animals in food-related Pavlovian conditioning experiments emerge because the experimental CS-USepisode mimics a naturally occumng episode for which preorganized behavior patterns exist. . . . The response to natural food precursors is the product of evolutionary specialization and previous food signaling encounters and it is these behavior patterns that occur as signal-centered actions in the Pavlovian food signaling experiment. We propose that the artificial signal substitutes for a natural signal. . . (Jenkins et al., 1978, p. 292).

The authors admit that the development of this idea requires an understanding of natural signaling events and the species-specific behavior that these events elicit; at present, the concept lacks predictive power. In effect, what is being called for is an understanding of the function of the conditional response. Of what advantage is this anticipatory signal-directed behavior? My answer is that the CR is, itself, a highly efficient means of foraging for food and, as such, is a behavioral phenomenon readily embraced by optimal foraging theory. Optimal foraging, an idea introduced by MacArthur and Pianka ( I 966), has become a popular theme in the area of behavioral ecology (see Pyke et a f . , 1977, for a review). The term refers to the application of optimality theory, or the idea of optimal choice, to the foraging behavior of animals. The logic of such an application is straightforward: Through natural selection, animals are produced which are maximally efficient at propagating their genes. To do so, of course, would require that they perform all activities, including food harvesting, with maximal efficiency, since all of these activities subserve propagation in the end (Davies and Krebs, 1978). Pavlovian conditioning may be one mechanism whereby this is accomplished. Suppose, for example, that discrete visual cues, like the brightly colored molds and fungi on dead logs, come to be associated with food, in this case insects living just under the bark. Pecking at these cues in the future is likely to uncover food and foraging will thereby be directed to highly likely food locations. In general, anticipatory conditional responding in the presence of naturally occurring food signals would decrease foraging time and energy expenditure. Thus, what has been called foraging behavior in the wild could in some instances be signal-directed conditional responding. This cue-directed feeding behavior would appear to have much in common with the concept of “search image formation,” which has been used to refer to a change in an animal’s ability to detect cryptic food and which results from previous experience with that food item (L. Tinbergen, 1960). Krebs (1978) has identified search image formation with a learning mechanism, albeit an unspecified one, and views this behavior as an adaptation which enables animals to increase the profitability of a particular food item via a decrease in the recognition time of that food item. M. Dawkins (1971) and Pietrewicz and Kamil(l979) have shown that the observed improvement is not merely the result of becoming better at handling the food, or searching in a better location. Dawkins was also

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

11

able to rule out simple peripheral modifications to vision, such as reorientation of the head or eyes. Although search image formation is often referred to as a perceptual change, the phenomenon might also be described as the solution of a difficult Pavlovian discrimination problem in which the CS+ configuration associated with food (visual cues of the prey) is highly similar to the CS- configuration (visual cues of nonprey items). Learning would take place slowly under these conditions but, eventually, approach and consummatory behavior would be directed toward only the CS+ objects. Search image formation is believed by Dawkins to involve selective attention to visual cues of the prey item; and, because of this selective attention, other obvious food items are frequently ignored. Alcock (1973) reports, for example, that previous experience with a particular food type influenced the ability of red-winged blackbirds to locate other types of prey items. The food items were placed within small holes located on wooden perches. When both seeds and mealworms were made available in this manner following a training period in which birds found only mealworms or seeds, mealworm-experienced birds took more mealworms before removing a single seed, and seed-experienced birds tended to locate seeds first. Alcock concluded that selective attention to specific visual cues which signal a food enables birds to filter out irrelevant visual information and, thus, obtain prey more quickly. Pietrewicz and Kamil’s (1979) data would tend to support this conclusion. The ability of blue jays to detect cryptic moths increased over successive encounters with a single prey species. When, however, the experimental situation was arranged such that successive encounters involved mixed prey species whose visual characteristics (cryptic wing patterns) differed markedly, detection ability did not improve. Whether anticipatory conditional responses contribute in any way to search image formation remains to be explored. A recent report (Sinnott et al., 1980) that pecking in red-winged blackbirds, Alcock’s study species, may be elicited by Pavlovian CSs paired with food is encouraging of such an approach. And, interestingly, a controversy has arisen in the search image literature which Pavlovian conditioning might easily address. Search image formation, as noted earlier, is said to involve specific visual cues. Royama (1970), on the other hand, has argued that, more often, birds may simply rely on a mechanism which returns them to profitable locations. Alcock (1973) has shown that both location cues and specific prey cues are employed in the same searching task. Of course, this latter position is consistent with a Pavlovian approach: CSs, either object or location cues, elicit approach and contact behavior. Thus, two mechanisms need not be invoked. Pavlovian conditioning can account for the behavior of animals which return to profitable locations and which locate specific prey items by utilizing specific prey cues. The optimizing strategy which has been suggested for anticipatory responding might also help to explain (functionally, that is) why, in dogs, for example (Jenkins et d.,1978), the same CS event should elicit soliciting behavior in some

12

KAREN L . HOILIS

animals but licking and consummatory behavior in others. The history of an animal’s food procurement-whether it has had to beg or search for food-may determine in part the form of the CR. Woodruff and Williams (1976) suggest that whether the CS elicits appetitive (e.g., soliciting) or consummatory behavior depends upon the extent to which the CS resembles naturally occurring appetitive or consummatory releasers. To this I would add that “resemblance” might be shaped by the individual’s past history. A more fundamental issue that must be raised here is whether it is reasonable to assume that the behaviors which emerge as CRs do so as a direct result of the Pavlovian conditioning relationship between stimuli and not because the CR has beneficial consequences. That is. when an experimenter arranges that a lighted key will be followed by food, independently of the pigeon’s behavior, the paradigm is indisputably Pavlovian. However, the learning process which results in key pecking may not be. Because the lighted key is followed by food, pecks on the lighted key are also followed by food. Might the pigeon be pecking the key because its pecks are reinforced with food; that is, is key-pecking instrumentally reinforced and established through the relationship between response and reinforcer‘? Whether the response-reinforcer (instrumental) relationship is necessary for the development of the CR can be determined experimentally. If the performance of the conditional response depends upon instrumental reinforcement, then manipulations of that instrumental contingency should bring about changes in the response. Specifically, if the occurrence of the CR prevents the appearance of the reinforcer (US), then the CR should cease altogether. This procedure is called omission training. The US follows the CS only on trials in which a CR does not occur. On trials in which the CR occurs the CS is presented alone. In spite of the adverse consequences of a CR, many experiments (e.g., D. R . Williams and Williams, 1969; Herrnstein and Loveland, 1972; Moore, 1973) have shown that omission training with experimentally naive animals does not result in the disappearance of the CR. Typically, what happens in these experiments is that, following the initial pairings of CS and US (Fig. I A) a CR emerges (Fig. I B). The US is prevented on that trial. A CR on a few subsequent trials continues to prevent the US. Finally (Fig. 1C) the CR is omitted and, on that trial, CS and US are again paired. After a few more pairings the CR again emerges, more quickly this time. The entire process of emergence, disappearance, and more rapid emergence repeats itself over and over again. It is important to recognize that, in omission training with experimentally naive animals, the response has never been followed by food; yet, in the absence of such instrumental reinforcement, a CR is elicited. Thus, we may conclude that the instrumental contingency is not necessary; the Pavlovian CS-US (stimulus-reinforcer) relationship is sufficient to produce the CR. (However, under natural conditions, the instrumental contingency may still be important; see Section VI,A.)

BIOLOGlCAL FUNCTlON OF PAVLOVIAN CONDITIONING

OMISSION

13

TRAINING

FIG. 1 . Omission training. The U S follows presentation of the CS only on trials in which a CR does not occur. The CS is the brief illumination of a key light on the panel wall. The US is the presentation of grain, for a few seconds, from a food hopper just below the key. The top panels depict CS presentation by a darkening of the round key light. During US presentation the pigeon pecks grain from the food hopper which is, at vther times, inaccessible. In the bottom panels, time is represented on the horizontal axis. Onset and offset of stimulus events are depicted by upward and downward deflections, respectively, from the horizontal lines. The interval between successive CSs is variable. Individual pecks on the key light (CRs) are separate vertical deflections.

A simple analogy may make the point more intuitively clear. Suppose that one were to build an animal and suppose further that one determined that the ability to detect cause and effect by this animal would be a useful feature. One way of doing this, certainly, would be to wire the animal such that responses followed by good things would tend to be repeated. This, of course, is an instrumental learning mechanism. However, another way of doing it would be to wire the animal such that stimuli got hooked up with other stimuli if they reliably cooccurred: Events which would precede important stimuli would be associated such that those events would come to elicit similar responses. This is a Pavlovian learning mechanism. Which shall we choose? The answer is that, so far as the outcome is concerned, it would not matter. The outcome of both wiring diagrams is that the animal detects cause and effect. Pavlovian conditioning is neither primitive nor “mindless”; it does not result in the association of any two events that happen to cooccur (Rescorla and Wagner, 1972; Rescorla, 1980a). Yet, so simple a process may result in behavior which, to us, appears to operate “instrumentally” in the animal’s environment. But might it be argued that the results of omission training experiments underline the potential maladaptiveness of anticipatory CRs? This suggestion is an interesting one. On the one hand, one might reasonably refute such arguments on the basis that naturally occurring signals do not reliably accompany events if and only if the animal does not respond to those signals. Omission training experi-

14

KAREN

L.

HOLLIS

ments, despite their usefulness as paradigms for answering causal questions, are clearly distortions of the environment which selected an anticipatory response mechanism. On the other hand, this refutation begs the question of whether the potential maladaptiveness of the CR poses a serious problem for functional interpretations. The answer is that the Pavlovian conditioning mechanism need only produce a net benefit for it to be maintained. Natural selection is concerned with average statistical benefits and sometimes in nature, as in the laboratory, Pavlovian conditioning will have adverse consequences for its possessor. PROCESSES I N THE DEVELOPMENT OF C. PAVLOVIAN FOODRECOGNITION Pavlovian conditioning of consummatory and food-procuring behavior emerges whenever signals reliably accompany the presence of food. The appearance of a food object is understood to constitute the unconditional stimulus. Yet, in many of the species whose conditional behavior was reviewed, the recognition of food itself must be learned (e.g., Wortis, 1969; Hogan, 1973c; Pavlov, 1927). The designation “unconditional” implies that the response to the US does not depend upon previous learning experiences. What implications might this have in the present discussion of Pavlovian conditioning? First, calling the sight of food a US when its recognition must be learned does not undermine research which interprets the emergence of CRs to stimuli paired with food as Pavlovian in origin. Regardless of how food acquired its ability to elicit the pecking response, if stimuli paired with food come to elicit this response, too, we may safely conclude that the process is a Pavlovian one. In this sense the terms US and UR are merely convenient ways of designating the relationship between food stimuli and pecking behavior in an adult bird; as they are more frequently used, the terms make no assumptions regarding the ontogenetic status of the US-UR relationship. Second, there is evidence to suggest that food recognition learning itself is the result of Pavlovian conditioning, the process resulting from natural pairings of food and food stimuli which the parent arranges for the young. Wortis ( 1 969) studied the development of independent feeding in 2-week-old ring dove squabs (Srrepropelia risoria). Prior to 13-15 days of age the squab is completely dependent on male and female parents for nourishment, which it obtains in the form of crop-milk. Crop-milk consists of the sloughed-off cells of the esophageal walls and is transmitted to the young by means of regurgitation. The process of regurgitation feeding undergoes many changes as the squab develops, all of which contribute to the development of independent feeding. One of these changes involves the contents of the regurgitated food. From approximately the third day posthatching until weaning at 21 days, the regurgitated food consists of an increasing quantity of grain and seeds and a decreasing

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

15

quantity of crop-milk itself (Lehrman, 1955). Also, when the squab is 9-10 days of age the parents no longer brood continuously. They begin to leave the young alone in the nest for the first time. At this time begging first appears and consists, among other things, of pecking at the parent’s bill. The parent then delivers the regurgitated food. The transition from this dependent to independent feeding occurs at about 13-15 days of age when squabs first begin to peck at grain themselves. The frequency of parental feeding rejection also increases at this time. However, Wortis’ data suggest that rejection by the parents of squabs begging is not sufficient to explain the occurrence of pecking in 2-week-old squabs. Another parental behavior appears to have greater significance for squabs’ independent feeding. Two-week-old squabs were tested with “foster parents” whose own young were either younger or older than the subject squabs. Foster parents treated the subject squabs as though they belonged to the same age group as their own young. Foster parents whose own squabs were younger succumbed to the subject squabs ’ begging; whereas foster parents whose young were older frequently pecked at grain concurrent with their refusal to feed the begging squab. Groups in which foster parents pecked grain when they refused to feed were the groups in which squabs showed the greatest amount of independent feeding behavior. Control experiments showed that this development was independent of begging refusals per se. Wortis’ interpretation of her observations suggests a clear parallel with the autoshaping experiments. As a result of the experience of regurgitation feeding, the pecking movement which the squab directs at the parent’s bill consists of a bill-opening-and-closingmovement. When the young “xe about 2 weeks old the parent doves respond to these begging pecks by pecking at grain. The squab continues to peck the bill of the parent and this action transfers the squab’s head and bill to the location of grain. The squab’s bill eventually comes into contact with grain itself and, as a result of the opening-and-closing movement, some grain enters the mouth. Grain is, by now, recognized as food since regurgitation feedings have increasingly consisted of grain (Wortis, 1969). Through the actions of the parents, the sight of food (CS) is paired with food ingestion (US) and, following a number of these pairings, grain itself elicits pecking in a hungry ring dove. In Callus species solicitation pecking is not present in the young; nonetheless, food object learning occurs in these species as well (Hogan, 1973a,b; see Hogan, 1973c, for a review). Under natural conditions when a mother hen discovers food, she emits a special call which activates and attracts her chicks (Sherry, 1977). While food-calling the hen pecks at, picks up, and tosses food (see Fig. 2). The chicks peck at these bits of food and tend to pick up and swallow them. Moreover, Hogan (1973~)reports that this pecking seems to generalize to similarly appearing stimuli. Like the parent ring doves, the role of the mother hen is directive. She provides the opportunity for an association between the sight of

FIG. 2. Food object learning in junglefowl (Callus gullus spudiceus). (Top) When the hen discovers food she adopts the characteristic posture shown here and emits a rapid stacatto call. While calling she pecks at the food, picks it up, and drops it repeatedly. (Bottom) This behavior attracts the chicks and directs their pecking to appropriate food objects. (Photographs courtesy of D. Sherry.)

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

17

food (CS) and the positive effects of food ingestion (US). Finally, the description of the development of independent feeding in oystercatchers (Norton-Griffiths, 1969), meerkats, and domestic cats (Ewer, 1963, 1969) clearly suggests a Pavlovian paradigm. Of course, merely because the circumstances in the development of food recognition resemble a Pavlovian conditioning paradigm does not establish that the process is in fact Pavlovian. Many hypotheses exist which attempt to explain how socially transmitted behavior is acquired (Galef, 1976). Before a Pavlovian interpretation is accepted it would be necessary to show that the stimulusreinforcer (CS-US), and not the response-reinforcer, relationship was sufficient for pecking to develop. Omission training experiments could be easily adapted to such purposes. OF DIGESTIVE RESPONSES: A GUT REACTION D. CONDITIONING

In most conventional studies of Pavlovian conditioning, measurement of a CR is restricted to a single response or response system. Not surprisingly, the conditioning of skeletal behavior, for example, is more conveniently measured if animals are not tethered to apparatus necessary for the recording of concurrent autonomic behavior. Nonetheless, concomitant with the conditioning of these signal-directed skeletal behaviors, conditioning of various digestive and metabolic processes have been observed. These anticipatory autonomic responses, llke the skeletal behavior which accompanies them, function to maximize interaction with the food; whereas in the latter case the CR brings the animal in contact with food, in the former instance the CR prepares the animal to digest it.* The prototypical experiment of what, today, represents a large portion of animal learning research-Pavlov ’s demonstration of conditional salivationwas more concerned with digestive processes than with associative ones. Following in the tradition of Pavlov, Soviet research has tended to remain somewhat functionally oriented and has clearly demonstrated the preparatory function of the CR. This preparatory function of the conditional “alimentary reflex” (Pavlov, 1927, p. 17) has long been recognized and will be but briefly summarized here (see Bykov, 1958, 1959, for reviews). Bykov reports that CSs paired with food presentations elicit, in dogs, not only anticipatory salivation, but secretion of gastric and pancreatic juice and increased gastric and intestinal motility. The selective advantage of both increased diges’A number of studies have shown that CSs which are paired with either an injection of insulin (e.g.. S . Siegel, 1972) or an injection of glucose (e.g., Deutsch, 1974) elicit a compensatory glycemic CR. Thus, a compensatory hyperglycemic CR is elicited by CSs which precede insulininduced hypoglycemia and a compensatory hypoglycemic CR is elicited by CSs which precede glucose-induced hyperglycemia (but see, e . g . , Balagura. 1968). Unfortunately, the role of blood glucose levels in the regulation of feeding is not yet fully understood (Friedman and Stricker, 1976). A functional interpretation of these data would thus be premature (but see Nicolaidis. 1977).

18

K A R E N L . HOLLIS

tive enzymes and increased gastric and intestinal motility appears to be twofold: First, digestion itself is shortened when food is signaled, a result of the increased flow of digestive enzymes as well as the increased efficiency of the whole digestive process (Bykov, 1959). Second, the anticipatory reflexes actually potentiate metabolic processes, making the greatest portion of food absorbable (Nicolaidis, 1977). Conditional digestive responses do not operate together as an all-or-nothing phenomenon but are individually tuned to their preparatory role. For example, the coordination between salivary and gastric CRs only occurs when the US is digestible; a conditional response which anticipates an acid US does not include the conditional gastric component. Functionally, of course, the CR is in one case a digestive mechanism while in the other it serves to neutralize and expel an undigestible and harmful substance (Pavlov, 1927). Further, both the amount and the composition of anticipatory conditional gastric secretion to a CS eventually becomes adjusted to the specific type of food US (Bykov, 1959). More recently, Zamble has shown that deprived animals which are fed (or given water) at unpredictable intervals are not able to consume as much as animals for whom access to food (or water) is signaled (Zamble, 1973; Zamble et al., 1980, respectively). Rats which were maintained on a restricted diet and were fed dry food at unpredictable intervals lost weight more rapidly than did animals for which the 30-min feeding was signaled by a 15-min visual or auditory CS. Groups which received signaled presentations were able to consume as much as 20% more food than unsignaled control groups. Similarly, waterdeprived rats which received signaled presentations of water were also able to consume more than animals which received water at random intervals, although these differences were less dramatic than in the feeding experiments. Dating from an early experiment by Sheffield and Campbell (1954), there have been several demonstrations that CSs paired with food elicit “general activity” increases in both rats and pigeons (Slivka and Bitterman, 1966; Zamble, 1967, 1969). Zamble has interpreted his more recent data to suggest that this energizing role actually plays a part in the regulation of food and water uptake. Physiological research would indicate that anticipatory digestive (Bykov, 1959) and neuroendocrine CRs (Nicolaidis, 1977) also play a substantial role and may even be directly responsible for the energizing effects themselves. Digestive CRs should be especially important in the consumption of dry food where, as would be predicted, the largest differences between signaled and unsignaled conditions were obtained. Finally, Sheffield (1965) has shown that digestive CRs (salivation) will develop and be reliably maintained on an omission schedule, wherein a CS signals food provided that the dog refrains from salivating in the presence of that CS. Thus the criticism that digestive CRs are not Pavlovian in origin but are instrumentally reinforced via the positive consequences of anticipatory salivation is not

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

19

supported by Sheffield’s data. Like the skeletal CRs with which they are correlated, conditional digestive responses occur “reflexly” to CSs paired with food and only the reliable pairing of CS and US is necessary for the appearance of a CR. The animal itself (or, more precisely, its nervous system) need not be sensitive to the beneficial consequences of these CRs for them to develop. Natural selection has, by providing the animal with a Pavlovian conditioning mechanism, already seen to that. PROCESSES I N THE REJECTION-ORINGESTION-OF E. PAVLOVIAN TOXINS One special problem of locating food, especially in omnivores, is discriminating the edible from the inedible. Although almost all animals have evolved specialized receptors which enable them to reject toxic substances on the basis of taste, errors are inevitable and many substances are ingested which profoundly disrupt homeostatic maintenance in a number of physiological systems. In humans, for example, these substances tend to produce bitter and/or chemical imtant sensations. Conversely, substances which produce these taste-mediated or trigeminally mediated sensations tend to be harmful (Rozin et al., 1979). At the same time that one body of literature, food aversion learning, has demonstrated that Pavlovian conditioning processes may enable an animal to avoid a particular toxic food altogether via the associated cue properties of that substance, another body of literature has demonstrated that, when certain chemical toxins are ingested chronically, the Pavlovian conditional response prepares the animal for toxin ingestion by an anticipatory counterreaction, or “compensatory CR,” which attenuates the impending chemical insult. In both cases, the anticipatory nature of the conditional response is an important feature of optimizing (in this case, minimizing) an interaction with a toxic US.

I . Food Aversion Learning and Poison Rejection When a hungry rat is allowed to consume some distinctively novel substance and then is made ill, typically by injecting the animal with lithium chloride or exposing it to toxic doses of X-irradiation, the animal avoids ingesting that substance in the future. Control animals, which have experienced the internal malaise without prior exposure to the flavored test substance (or which have experienced the test flavor without exposure to the sickness), do not avoid the novel substance (for reviews, see Mackintosh, 1974; Rozin, 1976). Evidence that rats can associate gustatory cues with subsequent sickness has been available for some time (Rzoska, 1953; Garcia et al., 1955). However, the past 14 years have seen a burgeoning interest in food aversion learning, so much so that it is one of the most well-known phenomena of the animal learning literature. It would be virtually impossible in this short space to review the many important findings in

20

KAREN L. HOLLIS

this field or to discuss the many theoretical issues which it raises. This is not my purpose, however. The importance of food aversion learning to a functional understanding of Pavlovian conditioning lies in the answers to two questions: First, is the phenomenon itself the result of the stimulus-reinforcer, that is, the Pavlovian, relationship? Second, does the CS (i.e., the taste, the smell, or the sight of food) elicit a response whose function is to prevent ingestion? Revusky and Garcia (1 970) have argued that aversion results from an association being formed between the response of drinking or eating and the adverse consequences of that response, sickness; that is, the learning is of the instrumental, response-reinforcer variety. However, Mackintosh (1974) favors a Pavlovian point of view on the basis of evidence that aversion may develop to a flavor (CS) paired with illness (US) without its being ingested (Domjan and Wilson, 1972). Furthennore, the association responsible for food aversion can develop in an anesthetized animal (Roll and Smith, 1972; Millner and Palfai, 1975) suggesting to some (Rozin, 1976) that it is “subcortical.” Currently, food aversion is considered by most to be a Pavlovian conditioning phenomenon. In rats, internal malaise is preferentially associated with gustatory CSs (Garcia and Koelling, 1966; see Nachman et a l . , 1977, for a review of the differential participation of sensory modalities in food aversion). Olfactory cues may serve as effective conditional stimuli (Larue, 1975; Domjan, 1973); however, stronger aversions develop to taste cues than to odor cues (Palmerino et al., 1980; Durlach and Rescorla, 1980). And, although visual and auditory stimuli may be effective in producing a learned aversion (Garcia et al., 1956; Best et al., 1973; Rozin, 1969) these stimuli are overshadowed if gustatory stimuli are present (Garcia and Koelling, 1966; Garcia et al., 1968; Best et al., 1973). Many birds, on the other hand, have a poorly developed gustatory system relative to their visual system and select food on the basis of visual cues. Chickens (Capretta and Moore, 1970), quail (Czaplicki e t a l . , 1976; Wilcoxon, 1977), and blue jays (Brower, 1969) have been shown to develop food aversions to visual CSs. Moreover, chicks (Gillette et al., 1980) and quail (Wilcoxon et al., 1971) preferentially form aversions to visual cues rather than gustatory or auditory CSs, just the reverse of rats. The CR, an avoidance response based upon withdrawal from illness-associated CSs, is clearly an adaptive response whose function it is to optimize (here, minimize) interaction with the US. Kalat (1977) considers how the many specialized features of food aversion learning may further contribute to the adaptive significance of the avoidance CR. He suggests, for example, that food aversion learning “plays a major role in the animal’s selection of foods as well as its avoidance of poisons” (p. 83). In mammals, the need for one nutrient, glucose, to cite an example, is dependent upon how much the animal can consume of another, thiamine, since thiamine is necessary for the metabolism of glucose. Thiamine-deficient rats avoid glucose (Richter et af., 1938) presumably because the

BIOLOGICAL FUNCTION OF

PAVLOVIA” CONDITIONING

21

taste CS has been paired with the aversive effects of the inability to digest the sugar. Rozin (1976) details the biological significance of food aversion learning for food selection in generalist feeders, such as the rat and man. Tolerance of long delays between CS and US-as much as 12 hr (J. C. Smith and Roll, 1967)-as well as the preferential association of certain CSs with illness are but some of the features of the mechanism which can solve “the whole host of nutritional and poison-avoidance problems” (Rozin, 1976, p. 34). Cue-elicited avoidance itself would seem perfectly adequate to eliminate contact with poisons. However, there is evidence to suggest that the CR may be more involved. Rozin and Kalat (1971) argued that, “In taste aversion learning, the animal’s perception of the taste itself or of its affective value may change” (p. 478). A similar point has been made elsewhere (Garcia et al., 1970; Rozin et al., 1979). Observations of animals have tended to suggest that as a result of food aversion learning a hedonic change actually takes place and the formerly poisoned food, at one time obviously palatable, comes to taste bad. Rozin (1967) reports that rats spill a food previously associated with sickness, a behavior which they exhibit with innately unpalatable food, such as quinine. Recently,

FIG. 3. This photograph, taken from a videotape of an experimental testing session, shows a ring dove (Srrcppropelia risoria) retching (note hunched back) in the presence of a distinctively colored food substance (grain, indicated by an arrow). This behavior occurred immediately upon introduction to the testing arena. The grain had been paired with illness 24 hr earlier. (Photograph courtesy of D. Lendrem.)

22

K A R E N L. HOLLIS

Lendrem (1980, unpublished results) has obtained even more striking evidence of such a hedonic change. Ring doves were made ill with lithium chloride after consuming a small quantity of distinctly colored grain. Two of Lendrem’s four subjects were observed to retch at the mere sight of the colored grain on their next exposure to it, 24 hr later, although they had long recovered from the poisoning experience. A photograph of this behavior appears in Fig. 3. All of these conditioned behaviors insure, of course, that toxic food items are not consumed. Hedonic changes further protect the animal by guaranteeing that only a negligible amount-one taste-of the poisonous food is ingested. As well, the now unpalatable taste probably increases the animal’s ability to generalize to other potentially poisonous food substances which taste similarly. Further, behaviors like spilling and burying of the poisoned food in rats (Wilkie et al., 1979) serve the additional function of protecting the young from ingesting the toxic substance. This is to be contrasted with the outcome of the same Pavlovian conditioning process but in which the conditional response prepares the organism for poison consumption, discussed in the next section. 2 . Pavlovian Compensaror?,Responses and Poison Ingestion In humans, many toxic substancesdrugs-are consumed regularly and in quantities sufficiently large that the homeostatic maintenance of many physiological systems is stressed. Some of the most intensively studied drugs include the opioid peptides (heroin, morphine) and amphetamines, as well as ethanol, caffeine, and nicotine. The unconditional responses to these substances include both autonomic and skeletal behavior disturbances, some of which are illustrated in Table I. Not surprisingly, these drug-induced homeostatic disturbances reflexively elicit physiological and skeletal counterreactions, or compensatory responses, which restore the organism to normal levels of functioning. Under some, albeit rather restricted, conditions (e.g., continuous administration of a drug) these drug-elicited compensatory responses may play a role in the phenomenon of drug tolerance (Hinson and Siegel, 1980). Tolerance refers to an actual decrease in the homeostatic disturbance produced by a drug over the course of repeated administrations (Goldstein et al., 1974), a mechanism of obvious adaptive value. Drug-elicited compensation is but a minor part of the picture of tolerance, however. S . Siegel (1975, 1976, 1977a, 1978; Hinson and Siegel, 1980) has accumulated a most impressive body of data in support of a Pavlovian conditioning analysis of the tolerance phenomenon. In Seigel’s model of tolerance the drug administration corresponds to a Pavlovian conditioning paradigm (an observation made by Pavlov, 1927, himself, but not heretofore subjected to such rigorous analysis). The drug is the US and the drug-induced homeostatic upset is the UR. Cues which reliably predict the systemic stimulation induced by the drug, including environmental cues and, of course, the drug ritual itself, function as CSs. The conditional response is a compensatory one; that is, unlike

23

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

TABLE I COMPENSATORY CONDITIONAL PHARMACOLOGICAL RESWNSES~ Unconditional stimulus Ethanol Epinephrine

Glucose Insulin Nicotine Atropine Chlorpromazine Amphetamine Dinitrophenol Histamine Methyl dopa Lithium chloride Nalorphine Morphine

Unconditional response

Conditional response

Hypothermia Tachycardia 4 Gastric secretion Hyperglycemia Hyperglycemia Hypoglycemia Hyperglycemia Antisialosis 4 Activity T O2 consumption f O2 consumption Hyperthermia Hypothermia 4 Blood pressure Drinking Tachycardia Bradycardia Hyperthermia Analgesia

Hyperthermia Bradycardia T Gastric secretion Hypoglycemia Hypoglycemia Hyperglycemia Hypoglycemia Hypersalivation t Activity .1 O2 consumption . 1 0 2 consumption Hypothermia Hyperthermia T Blood pressure f Drinking Bradycardia Tachycardia Hypothermia Hyperalgesia

"After S . Siegel, 1979a. See S . Siegel (1979a) for details.

many CRs in which the conditioning procedure results in a response similar to the UR, the CR here is opposite in direction to the unconditional effect of the drug. Compensatory CRs provide a medhanism whereby the animal can minimize both the magnitude and duration of drug-induced disturbances. Figure 4 illustrates how, with repeated pairings of the environmental CS and drug US, compensatory conditional pharmacological responses may attenuate impending chemical insult. On first exposure to a drug (Fig. 4A), prior to conditioning, the drug's pharmacological effects are experienced at full intensity. However, with repeated pairings of predrug cues and drug administration (Fig. 4B and C), the drug UR is increasingly attenuated by the compensatory conditional response. That is, the net effect of the drug-induced UR and the opposing signal-induced CR is a reduction in the pharmacological effect of the drug (S. Siegel, 1979s). The existence of the compensatory CR such as that found in Fig. 4C is typically detected on a test trial by administering a placebo, instead of the drug, in the usual drug context. The environmental CSs give rise to a compensatory response which, in the absence of the drug, is unmodulated by the drug UR. Thus, if hypotherrnia is repeatedly produced by administration of ethanol, the

24

KAREN L . HOLLIS

,DRUG UR

TIME

FIG. 4. Time course of the net drug effect (stippled area) resulting from the interactions between the pharmacological unconditional response (UR) and the compensatory conditional response (CR) during successive stages in conditioning. (After S. Siegel, 1979a).

administration procedure followed by a placebo leads to a hyperthermia response (Uet al., 1979). Epinephrine produces a decrease in gastric secretion, an elevation of blood glucose concentration, and an increase in heart rate. In subjects with a history of epinephrine administration, an injection of a placebo in the usual drug context produces a heart rate decrease (Subkov and Zilov, 1937), an increase in gastric secretion (Guha et al., 1974), and a depression in blood glucose concentration (Russek and Piiia, 1962). Table I summarizes a number of investigations which report similar instances of compensatory conditional responses which attenuate the drug UR and which would decrease the homeostatic disturbance produced by the drug over repeated administrations, that is, would produce tolerance.

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

25

Other interpretations of tolerance postulate the operation of physiological changes within the organism (see S. Siegel, 1976, for a review); however, these interpretations cannot account for experiments which confirm the predictions of a wholly associative (learning) model: ( I ) Drug tolerance may be “extinguished” with a Pavlovian conditioning extinction paradigm. If environmental cues previously associated with a drug are presented without the drug, established tolerance is suddenly attenuated (S. Siegel, 1975, 1977a, 1978). (2) Tolerance is subject to latent inhibition. That is, if the administration procedure, without the drug, is repeatedly presented prior to its pairing with the drug, the acquisition of tolerance is retarded (S. Siegel, 1977a). (3) And, finally, tolerance is subject to the decremental effects of Pavlovian partial reinforcement. If CS-US sessions (environment-drug) are interspersed with CS-only sessions (environmentplacebo), tolerance is slow to develop (S. Siegel, 1977a, 1978). The prefiguring hypothesis maintains that the adaptive function of Pavlovian conditioning lies in the anticipatory nature of the CR. Obviously, the anticipatory compensatory CRs of drug conditioning which attenuate homeostatic imbalance are beneficial. However, to claim that this is an example of an adaptive mechanism at work may seem highly irregular given the short time period, evolutionarily speaking, of drug exposure. But this is to look for adaptation in the wrong place. I would suggest that compensatory responding did not have to evolve in response to recent drug use; this particular mechanism of regulating toxicity was already present. Drug-compensatory conditional responses are merely the inevitable by-products of a mechanism designed to regulate all forms of homeostatic imbalance. One potential source of homeostatic imbalance results from the exploitation of toxic food resources. Food sources, both plant and animal, have evolved various means by which they avoid being eaten. One of these means is by producing chemical irritants or toxic substances which protect them from potential predators. But the clever chemical defense of the pursued is eventually penetrated by the yet more clever offense of the predator. And, thus, the predator-prey arms race has resulted in the evolution of various physiological mechanisms to handle toxin ingestion. For example, the livers of many mammals, birds, and fish contain enzymes which detoxify a variety of drugs and pesticides (R. T. Williams, 1959). And the nasal glands of marine birds are elegantly specialized for a single function-to excrete the salt of ingested seawater which would otherwise upset the homeostatic balance of body fluids (Schmidt-Nielsen, 1959). Certainly, the insect world offers even greater marvels of poison detoxification mechanisms, but Whittaker and Feeny (1971) suggest that vertebrates may be less specialized than insects because they may rely more on learning abilities than genetic mechanisms. Although Whittaker and Feeny were referring specifically to food aversion leaming, the compensatory responding of selected (i.e., toxin-targeted) physiological systems is an additional mechanism whereby animals might reduce toxicity. By canceling any impending physiological disruption, such a mechanism would,

26

KAREN L . HOLLIS

like the nasal gland and the special detoxifying enzymes, effectively increase the range-and thus the net amount-of potential prey items. The compensatory drug CR is a food-related behavior, originally designed to protect animals from the chemical defenses of their prey. Unfortunately, however, humans, as well as wild animals, can develop addictions to the toxic substances found in these food items (R. K. Siegel, 1979). In humans, addictions are characterized by uncomfortable withdrawal symptoms and craving for the substance if its administration is terminated. Although this clinical definition of addiction is difficult to apply to the behavior of wild animals, many birds and mammals will return to feed on plants which contain psychoactive alkaloids or fermented fruit which contains alcohol. As S. Siegel (1979a) notes, a growing epidemiological and psychological literature has shown that the phenomena of drug tolerance and drug dependence “develop, persist and disappear together, as though they were reflections of the same underlying biological change” (Goldstein et al., 1974, p. 610). It is believed that the same mechanism which protects animals from toxins is responsible for the drug dependence, or addiction. That is, the drug withdrawal experience and the craving for the drug are a result of the elicitation of compensatory responses in the presence of predrug cues, which are not followed by the drug. Under these conditions, the compensatory responses, normally modulated by the drug, would achieve full expression and would presumably be very uncomfortable (S. Siegel, 1979a). This position is supported by the fact that the syndrome of withdrawal symptoms is, for the most part, identical to the syndrome of compensatory conditional responses, or “drug preparation” symptoms (S. Siegel, 1979b), and by data which show that experimental manipulations which produce changes in compensatory responding produce concomitant changes in withdrawal (S. Siegel, 1977b, 1979a). Considered in this light, one might argue that a compensatory conditional response mechanism is maladaptive because it risks dependence upon (i.e., addiction to) any chronically ingested toxic substance, including food items. Assuming that severe addictions reduce an individual’s fitness? this argument is nonetheless a straw man. As noted in an earlier discussion of the potential maladaptiveness of an animal’s learning ability, natural selection guarantees that mechanisms which yield a net benefit to the individual will be maintained in the population. Exposure to a drug does not necessarily produce a severe fitnessreducing addiction. Many individuals who are regularly exposed to alcohol or caffeine, for example, are able to restrict their consumption. They have become physiologically tolerant to that limited amount and their dependence upon it is similarly restricted. Despite the deleterious effects of these and other drugs on many animals, a sufficient number of more “temperate” individuals, for whom 3But koalas have managed to do quite well despite the fact that they are believed to be “addicted” to the narcotizing prussic acid present in the leaves of Eucalyptus. upon which koalas feed exclu-

sively (R.K . Siegel, 1979).

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

27

the compensatory conditional response is advantageous, insures that this mechanism remains in the population.

3. Whether to Reject or to Ingest: A Pavlovian Approach Perhaps a more intriguing question at the functional level of analysis concerns the relationship between CRs mediating poison avoidance and CRs mediating poison consumption. Cappell and LeBlanc (1977) note that the same drugs which animals will self-administer in some circumstances are capable, at equivalent dosages, of inducing food aversions in other circumstances. In man, as Rozin et al. (1979) note, although on first exposure to foods containing highly unpalatable toxic substances (e.g., alcohol, tobacco), human infants, young children, and uninitiated adults display an aversion, “most adult humans reverse their natural rejections and end up with strong positive affective responses to at least one innately unpalatable substance” (p. 1001). Under what circumstances, or parameters of toxicity, do the anticipatory avoidance responses of food aversion learning become the compensatory responses of drug conditioning? The importance of choice (self-administered versus experimenter-administered), the duration of drug effect (the longer the duration the more likely an aversion will develop), and, perhaps, the route of administration (intravenous with self-administration versus intraperitoneal with food aversion) have been suggested as causal determinants of whether a drug is avoided or consumed (see Cappell and LeBlanc, 1977, for a review). On another, functional, level the prefiguring hypothesis would predict that whether a toxic substance is ultimately avoided or consumed would depend upon whether compensatory CRs developed. Of course, the finding that aversions were perfectly correlated with the absence of compensatory CRs leaves the causal questions unanswered. One would still want to know what factors were responsible for the development of the “drug preparation” syndrome. Nonetheless, if compensatory responding were the determining factor, our understanding of both poison avoidance and poison consumption would be further advanced. Experimentally, the prefiguring hypothesis would predict that (1) aversion reversals of the sort described by Rozin et al. (1979) would be coincident with the formation of compensatory CRs; (2) conversely, if compensatory CRs were prevented, aversion reversals would not occur; (3) poison avoidance would be ameliorated-or prevented-by predrug treatments whose effects mimicked a poison preparation syndrome, i.e., whose effects were compensatory to the poison; and (4) drug tolerance and addiction would be attenuated by predrug treatments whose effects canceled the compensatory CR. Of course, ( 5 ) conditioned skeletal behaviors-approach or withdrawal-would be expected to coincide with compensatory conditional and food aversion responses, respectively. In sum, the function of the Pavlovian conditional response is to optimize interaction with the forthcoming US event. If the US is a food substance contain-

28

K A R E N L. HOLLIS

ing necessary nutrients and/or calories, both autonomic and skeletal preparatory behaviors will serve to maximize interaction with that US. In some cases-where the food substances also contain toxins-the preparatory CR will involve responses which serve to compensate for any ill effects of the toxin together with responses which maximize caloric and/or nutritional intake. Under other circumstances the preparation will involve behaviors, again both skeletal and autonomic, which serve to prevent consumption and thus minimize interaction with the US.

Iv. PAVLOVIAN CONDITIONING OF DEFENSIVE BEHAVIOR At the same time that some animals are searching for food, they must be alert lest they become food for their predators. Antipredator (interspecific) defense strategies may assume many varied forms (Edmunds, 1974); nonetheless, to be successful these behaviors must occur with minimal delay. Be it fleeing or freezing, a strategy which is deployed too slowly risks predation. Broadly speaking, an animal may be said to engage in intraspecific “defense” as well, this term being used to refer to territorial defense or to intraspecific competition between males. Here, too, the shorter the latency to respond, the more effective the defensive maneuver. Thus, one might expect that knowing where and when predators will strike and rivals will appear would enable an animal to optimize its defensive strategies. Such prescience could, in part, be based upon classically conditioned “warning signals,” in the presence of which, it is hypothesized, anticipatory conditional responses would function to optimize interactions with predators and rivals. Predators should be evaded, rivals vanquished, and both with as little energy lost or as little physiological disruption as possible. This section will review the evidence for classically conditioned interspecific and intraspecific defensive behavior and attempt to show the adaptive value of prefiguring. DEFENSE A. INTERSPECIFIC (ANTIPREDATOR) Despite the enormous variation in different species’ defense strategies (Edmunds, 1974; Harvey and Greenwood, 1978), the demands made upon the physiologies of even as diverse groups as mammals and birds are nearly identical (Assenmacher, 1973). Whether the defense strategy involves mobbing (Hinde, 1954; Curio, 1975), “protean” (i.e., erratic) escape flight (Humphries and Driver, 1971), or elaborate distraction displays (Simmons, 1952), the rapid mobilization of energy reserves is imperative in all antipredator strategies except passive immobility (which will be discussed later in this section). The physiological response subserving this commonality of demand and adaptive function is a

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

29

massive and general internal reaction which prepares the animal for “flight or fight” (Gray, 1971).4 Briefly, heart rate, blood pressure, and respiration all increase. Together with the release of stored red blood cells and the mobilization of blood glucose, these physiological changes permit highly oxygenated and hyperglycemic blood to be redistributed to vital centers directly involved with the flight or fight response, that is, the skeletal musculature, brain, and liver. Conversely, digestive and reproductive functions cease. In mammals the palmar skin surfaces become moistened (the galvanic skin response), thus increasing the surface friction necessary for movement and/or hand manipulation. The autonomic nervous system responses which constitute this “stage of alarm” (Selye, 1952) occur within seconds of the detection of a predator. Clearly, evolution has provided the animal with a mechanism for shortening its reaction time to potentially dangerous events. Yet, even a shortened reaction time may not be short enough: The physiological response of the pursued, which prepares it for a fast escape, also prepares the pursuer for a fast chase. Reaction time may be made even shorter, however, by an additional mechanismlearning. And, one of the ways in which learning may be manifested is by anticipatory autonomic responses to environmental signals which reliably precede danger, that is, Pavlovian conditioning. The advantage of the conditional response is obvious. Any decrease in the latency to respond to the predator effectively increases the prey’s lead in the race.5 Better still, if the prey could in fact respond early enough to escape counterdetection by the predator, a strenuous chase could be avoided altogether. 4“Fight” here refers to inrraspecific (i.e., typically called aggressive) encounters. The autonomic responses discussed here in the context of inrerspecific defense are also those of aggression; however. their discussion will be omitted in the section on conditioned aggression because Pavlovian conditioning studies have been limited to situations resembling interspecific, as opposed to intraspecific, defense. Pavlovian conditioning methodology could, of course, be easily adapted to measure the anticipatory autonomic responses of aggressive encounters. Such research would provide a fertile ground for investigating functional questions. 5 J ~ sas t the prey can learn to respond to signals which accompany the predator, so, too, can the predator learn to “predict” the location of the prey (see Section 111). If this learning were perfectly reciprocal-i.e., both predator and prey signals were equally availablelpredictive and the animals were equally capable of making associations under all circumstances-an unstable runaway escalation or “arms race” would result (R. Dawkins and Krebs, 1979)until the detection time of both predator and prey presumably reached some lower limit. Even under these conditions, Pavlovian conditioning need not become antiquated weaponry; i s . , the advantage of early detection would not be hindered by a symmetry between predator and prey. Nonetheless, most predator-prey arms races are inherently asymmetric since prey are under stronger selection pressures to develop better strategies: “A fox may reproduce after losing a race against a rabbit. No rabbit has ever reproduced after losing a race againstafox”(R. DawkinsandKrebs, 1979,p.493).Theinterestingquestionarisesas to whether the asymmetry in the arms race is at all manifested in the conditioning process. For example, might the CR of the prey involve a greater proportion of the components of the autonomic fear response, or be of greater magnitude (as measured energetically), than the CR of the predator?

30

K A R E N L. HOLLIS

A voluminous literature (see, e.g., Mackintosh, 1974; A. H. Black, 1971) attests to the reliability with which autonomic “fear” responses are classically conditioned. In these studies, however, psychologists have been more interested in the causal mechanism(s) of conditional behavior than in its function and, not surprisingly, an electric shock US has been substituted for a potential predator. Only occasionally have other noxious stimuli been used (Hilgard, 1931 ; Harris, 1943). Moreover, CR measurement typically involves a single target response. Multiple response observations are somewhat rarer and no study, to my knowledge, has investigated conditioning across the spectrum of behaviors which characterize the flight or fight response. Nonetheless, the studies in which multiple CRs have been observed demonstrate that the CS may elicit many of the autonomic fear responses elicited by potentially dangerous events. Simultaneous conditioning of heart rate and respiratory changes have been demonstrated in fish (Otis et al., 1957). dogs (Fitzgerald and Walloch, 1966), and pigeons (Cohen and Durkovic, 1966). Similarly, Dykman et af. (1965) measured heart rate, blood pressure, general activity, and leg flexion in dogs and found that all these responses became classically conditioned to a tone CS; heart rate and galvanic skin response changes were classically conditioned in rats (Holdstock and Schwartzbaum, 1965). Stebbins and Smith (1964; 0. A. Smith and Stebbins, 1965) found a positive correlation between blood flow through the terminal aorta and conditioned heart rate in monkeys while Yehle et af. (1967) observed a similar correlation of conditioned heart rate and blood pressure in the rabbit. But such a brief summary is to obscure the complicated patterns within these data. The conditional autonomic response is not merely an anticipatory manifestation of the emergency reaction discussed earlier. While the unconditional emergency reaction tends to be a general response, the CR involves fractional components of the total UR (0.A. Smith and Stebbins, 1965), which condition at different rates and evidence different patterns of change during the CS (Yehle et af., 1967). Moreover, for any given US, conditioning may result in a CR either similar to the UR or one which is opposite in direction (Santibanez et af., 1965; Otis et af., 1957). If these data present theoretical stumbling blocks for general process theory’s causal analysis (but see Mackintosh, 1974), they are the very foundation of a functional framework. The prefiguring hypothesis is that the function of Pavlovian defensive conditioning is a CR which optimizes the defense strategy; therefore, one would expect that situational variables would determine the timing, the patterning, and even the form of the CR (e.g.. similar or opposite to the UR). Indeed, from a functional viewpoint it would be surprising if this were not the case. For purposes of exposition, let us examine one particularly well-studied component of the complex fear response, heart rate, and, in particular, let us explore the relationship between ecological variables and the form of the CR. The heart rate response is a particularly interesting example because without

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

31

exception, the unconditional response elicited by electric shock in mammals and birds is a heart rate acceleration (A. H. Black, 1971), while heartrate CRs may be acceleratory (conditioned tachycardia) or deceleratory (conditioned bradycardia) depending upon the circumstances. 1 . Conditioned Bradycardia and the Fright Response

In a report entitled “Experimental search for the ‘conditioned response,’ Wamer (1932) complains of his futile attempts to find the real CR which is elicited by CSs signaling shock in rats. To the CS his subjects would merely hold their breath and wait tensely. However, it is now generally recognized that, often, freezing is the CR in rodents (Blanchard and Blanchard, 1969a; Shettleworth, 1978) and is accompanied by heart rate deceleration (deToledo and Black, 1966; Panish, 1967; deToledo, 1968). Conditioned heart rate deceleration, or bradycardia, has also been observed in the rabbit (Yehle et al., 1967). While Schwartz (1978) suggests that, generally, the function of conditioned bradycardia is to compensate anticipatorily for the physiological stress of an unconditioned heart rate increase, I am inclined to disagree. By a simple manipulation of the experimental situation, conditioned bradycardia can be reliably replaced, in the same animal, by conditioned tachycardia (heart rate acceleration). This response would, presumably, impose further strain upon the cardiovascular system. That the conditional response is highly dependent upon the experimental situation fits rather nicely, however, with what is known about the antipredator defense strategies of animals exhibiting both types of conditional response. The defense strategy of animals such as rodents and rabbits involves a systematic use of both flight and freezing behaviors, termed flash behavior (Edmunds, 1974). If a predator is spotted, the animal runs some distance and, if the predator is not in close pursuit, it turns to face the predator and rests motionless and cryptic. In this way a long chase is often avoided. The term “flash behavior” refers to a “flash of color” which accompanies this particular form of escape in many animals and which disappears when these animals come to rest. The white rumps of many rabbits and ungulates, for example, the white-tailed deer, the white tail feathers of many birds (e.g., juncos), and the bright color on the inside of the thighs in various species of frogs are some examples. The function of the color flash is disputed. Edmunds (1974) suggests that the predator may be startled by the flash and then be induced to track the color visually. The sudden disappearance of the color at a distance, accompanied by the animal’s cryptic immobility, may then deceive the predator into assuming that the prey has vanished. (Other possible functions of rump patch signaling are discussed by Harvey and Greenwood, 1978). The antipredator strategy of these animals is thus a composite of flight and freezing behaviors and, therefore, we might expect to find that both behaviors are conditionable, depending upon the circumstances. Blanchard and Blanchard ”

32

KAREN L . HOLLIS

were able to demonstrate just that. In rats a fearful situation, a black box in which they had been shocked, produced freezing behavior (1969a) while a discrete upprouching fear stimulus produced active avoidance and flight behavior ( 1969b). Moreover, Karpicke et ul. ( 1977) have shown that when freezing was elicited by CSs paired with shock, rats typically turned in the direction of the CS, facing the danger signal, as they would turn to face the predator in the wild. These data, juxtaposed with the conditioned bradycardia experiments, suggest that in rodents and rabbits, bradycardia is the autonomic correlate of freezing behavior. But what function does this autonomic response serve? Clearly, a pounding heart and general “nervousness” might increase the animal’s chances of being detected and thus be detrimental; but why should heart rate decrease’? Moreover, classically conditioned bradycardia in anticipation of a shock US has been observed in cats (Santibanez ef af.. 1965), dogs (Lang and Black, 1963), fish (Otis et ul., 1957), and humans (Notterman et al., 1952; Zeaman et al., 1954). species which do not possess a particularly well-defined freezing response. Of what special significance then is the anticipatory cardiac deceleration if it does not always accompany freezing? Data obtained by Webb and Obrist (1970), working with humans, may help to answer these questions. Their experimental paradigm was a simple reaction time (RT) task. Male subjects were seated in front of a panel with their hand poised over an unilluminated button. A “ready” signal was given, and, following a short interval-the preparatory interval-of a few seconds, the stimulus button was illuminated. The subject’s task was to depress the button as soon as possible after stimulus onset, simple reaction time being defined as the time between onset of the stimulus and the subject’s response to that stimulus. In addition to simple reaction time, heart rate was monitored together with measures of somatic activity, including eye blinks, eye movements, and electrical activity (EMG) in the chin muscles. (Chin EMG is an especially sensitive measure of general somatic activity.) For two groups of subjects, preparatory intervals of 2, 4, 8, and 16 sec were presented in either an ascending or descending series such that all trials of a given interval were presented together in one block. RT was tested 24 times per interval. In a third group the intervals were presented in a completely irregular and unpredictable manner. In all groups the inhibition of ongoing somatic events during the preparatory interval coincided with cardiac deceleration, the peak of these changes tending to occur where they just anticipated the subject’s response. This anticipatory cardiac deceleration in a simple RT task replicates the findings of earlier reports by Lacey (19671, Chase et a l . (1968). and Obrist et al. (1969). Of special interest in the Webb and Obrist (1970) study, however, is the manipulation of the foreperiod and its effects on RT performance. These data show clearly that short, irregularly presented preparatory intervals, which attenuated the cardiac and somatic changes, resulted in a marked decrement i n performance.

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

33

Other experiments by Obrist and his colleagues (see Obrist et al., 1972, for a review) demonstrate the relationship between heart rate deceleration, somatic inhibition, and performance on the RT task. Briefly, those subjects who show the greatest decreases in heart rate and somatic events tend to be those who respond more quickly (Obrist er al., 1969); and, within subjects, RT performance is best on those trials in which heart rate deceleration and somatic inhibition is greatest (Obrist et al., 1970). The authors conclude that “cardiac deceleration and the concomitant cessation of somatic events are a peripheral manifestation of a process initiated by the CNS which inhibits ongoing task-irrelevant activities and whose cessation facilitates attention and the ability to respond rapidly to the stimulus” (Obrist et al., 1972, p. 330). This position is very similar to one adopted earlier by Lacey and his colleagues (Lacey et al., 1963) who showed that, in humans, heart rate decreases during tasks which require subjects to attend to environmental stimulus changes. The significance of these data for aversively motivated bradycardia is to suggest that, in a number of species, the CR involves heart rate deceleration as part of a more general behavioral syndrome involving general inactivity (Belkin, 1968; E. N . Smith et al., 1974; Ruff, 1971; E. N . Smith, 1978) and which as a whole functions as a defensive alert strategy. Under these conditions an animal’s reaction time is quickest. In some of these species, for instance, rats and rabbits, this alert posture is accompanied, as well, by cryptic immobility. Thus, at the same time that freezing and fear bradycardia decreases the animal’s chances of being detected by a predator (E. N . Smith and Worth, 1980) the entire response syndrome, including the skeletal response of turning toward the predator (Karpicke et al., 1977) increases the animal’s chances of detecting that predator first. But would it not appear that the temporal parameters of laboratory investigations heavily strain an adaptive interpretation of the CR? After all, the “fright” reaction in rabbits and rats is elicited by CSs which immediately precede electric shock. Shouldn’t the rat and the rabbit attempt to run from the inevitable pain instead of freeze? Blanchard and Blanchard’s (1969a,b) studies, noted earlier, are instructive. Recall that they found that situational (i.e., static) CSs elicited freezing in rats but approaching CSs elicited flight behavior. It may be the case that rats (and rabbits) are predisposed-either as a result of their “wiring” or as a consequence of their tendency to generalize from previous experiences (Mackintosh, 1974)-to respond to static cues with fright behavior and moving cues with flight. Alternatively, flight is typically thwarted in the laboratory and animals may adjust their behavior accordingly. Blanchard et al. (1976) found that a brief familiarization with an inescapable testing environment changed rats’ initial response to a predator (a cat) from flight to freezing. Conditional stimuli may likewise determine the topography of the defense response and, in fact, may have contributed to Blanchard et al. ’s observations (cf. Bolles, 1975). Thus, the

34

KAREN L . HOLLIS

apparent paradox of the laboratory may be just that but, in any case, the question deserves further study. Indeed, one might expect that, in the rat, the conditional response to a discrete approaching fear stimulus under more natural conditions would be two-phased: While the initial phase should involve cardiac acceleration accompanied by active avoidance and flight, the second phase, commencing when the animal has reached a safe distance, should involve a cardiac deceleration accompanied by freezing. 2.

Conditioned Tachycardia and the Flight Response

Conditioned heart rate acceleration, or tachycardia, has been reported in the pigeon (Cohen and Durkovic, 1966), in the dog (Church er a l . , 1966; Overmier, 1966; Fitzgerald and Walloch, 1966; Obrist and Webb, 1967), and, occasionally, in the rat (R.W. Black and Black, 1967; Fehr and Stern, 1965) in response to CSs paired with a shock US. If, like the conditioned bradycardia response, the conditioned tachycardia response also optimizes an animal’s defense strategy, then one would expect that tachycardia would normally subserve an active flight response. Experiments in which the animal could avoid a shock by fleeing to a place of safety during a preshock warning signal (CS) have suggested that this is indeed the case (A. H. Black, 1959, 1965; A. H. Black and Dalton, 1965; Carlson, 1960; McCleary, 1954; Miller e t a l . , 1967; Perez-Cruet e t a l . , 1963; Soltysik, 1960; Soltysik and Kowalska, 1960; Stem and Word, 1962; Wenzel, 1961; Werboff et a l . , 1964; see A. H. Black, 1971, for a review). In all of these experiments, involving monkeys, dogs, and rats, the response to the CS was heart rate acceleration. Wenzel ( 196 1) found that the heart rate acceleration occurred just before and, as might be expected, during the skeletal avoidance response. That rats and dogs exhibit conditioned tachycardia in the avoidance task but conditioned bradycardia in other situations (e.g., deToledo and Black, 1966; Lang and Black, 1963, cited in A. H. Black, 1971) should not be surprising given that other components of the flight and fright responses are also differentially evoked in these situations. Thus it would appear that the form of the CR is highly dependent upon the experimental situation. If, for example, the CS is a discrete moving stimulus (Blanchard and Blanchard, 1969b), or if the experimental situation allows the animal to escape to a place of safety (e.g., A. H. Black, 1959; Overmier, 1966), the animal will exhibit the flight response, accompanied by conditioned tachycardia. If, on the other hand, the CS signals a dangerous environment (Blanchard and Blanchard, 1969a) or escape is not possible (Blanchard et d.,1976; Lang and Black, 1963), the animal will exhibit the fright response and conditioned brachycardia. The CR functions, in both cases, to maximize defense; conditioned bradycardia and fright behavior are part of a more general defensive alert strategy while conditioned tachycardia and flight behavior effectively remove the animal from a dangerous situation. Of course, not all interspecific defense behaviors fit the dichotomous classificatory scheme

BIOLOGlCAL FUNCTION OF PAVLOVIAN CONDITIONING

35

of heart rate changes developed here; anticipatory heart rate changes may serve other purposes in different species. The reader should recognize that, although the previous review of interspecific defense has largely been restricted to the strategies of but a few species, this has only been because the conditioning literature is similarly restricted. Several controversial issues surround the classically conditioned heart rate response. One of these is whether the heart rate changes are themselves classically conditioned or are merely manifestations of conditioned skeletal behavior, freezing and fleeing, for example. While some investigators (K. Smith, 1954; Shearn, 1961) claim that changes in heart rate are mediated by overt responses, more recent data (e.g., A. H. Black, 1971; A. H. Black and deToledo, 1972; Obrist el a l . , 1972) suggest that cardiac and somatic events are influenced by common CNS mechanisms. However, the mechanism by which heart rate changes, be it direct or indirect, does not bear on the functional arguments presented here. The conditioned fright and flight responses, including the heart rate changes, are “a reflection of biological processes concerned with the adaptive facilitation of organism-environmental interactions” (Obrist et a l . , 1972, p. 330). That is, the CR functions to maximize defense. A second issue, and one which is relevant to all classically conditioned behavior, is whether defensive CRs are the result of the Pavlovian (stimulusreinforcer) relationship between CS and US or are, instead, the result of an instrumental (response-reinforcer) relationship. Some writers (Prokasy, 1965; Perkins, 1968) have argued that the CR is established through an instrumental contingency; the conditional autonomic and skeletal defense response occurs, for example, because the CR succeeds in reducing the aversiveness of the US. According to this argument, the observation that rats, when offered a choice between signaled and unsignaled inescapable shock, prefer a signaled shock situation (Lockard, 1963) suggests that the signal allows the animal to respond in such a way that shock is made less aversive (Perkins, 1968). The reduction in aversiveness reinforces the “preparatory” response via an instrumental conditioning mechanism (see Fanselow, 1980, for a review). One must be careful here not to confuse cause (mechanism) with adaptive function. Pavlovian conditioning and instrumental conditioning posit different mechanisms for their operation (stimulus-reinforcer versus response-reinforcer associations, respectively). Yet, as noted earlier (Section III,D), the adaptivefuncrion of the CR may indeed be to optimize interaction with the US-here, to maximize the defense strategywithout requiring that an instrumental mechanism be responsible. The question, then, is not whether “preparation” in the form of a CR occurs, or even whether this CR has beneficial consequences, but whether an instrumental conditioning mechanism or a Pavlovian conditioning mechanism is responsible for that CR. A review of omission training experiments (Mackintosh, 1974), in which the shock US is omitted on trials in which a CR occurs, suggests that the Pavlovian relationship is sufficient for the development of the CR. Omission of the US

36

KAREN L . HOLLIS

typically weakens rather than strengthens the CR (Soltysik and Jaworska, 1962), just the reverse of what an instrumental learning analysis would predict. [Failure to obtain these results has frequently been reported with rats when the CR involves a running response (e.g., Brogden et a l . , 1938; Kamin, 1956). Omission training produces a more reliable running response than the Pavlovian conditioning schedule. However, this is because the Pavlovian schedule tends to elicit freezing instead of running. Thus, even in these experiments, omission training does not produce a more reliable response than Pavlovian conditioning; it merely produces a different one (Sheffield, 1948).] The conditional changes in heart rate, certainly one of the most intensively studied autonomic defensive responses in both laboratory and field, are but a tiny part of the picture of an animal’s defensive response strategy. Yet, together with skeletal behaviors involving conditional immobility and simple avoidance responses, these CRs constitute much of what is known about the role of learning-specifically Pavlovian conditioning-in antipredator defense. The picture is incomplete. Insufficient data preclude a satisfactory analysis of much of the subtle intricacies of the Pavlovian conditional response. Although one might hazard to guess, for example, that the fractional nature of the CR in defensive conditioning and the differential patterns of the individual response elements function to meet the animal’s immediate needs while at the same time reduce the overall energetic and/or physiological cost, this interpretation is highly speculative. To be sure, a similar function was suggested for food CRs. Conditioned salivation, for example, is accompanied by conditioned gastric secretion only if the CS is paired with a food US and not if the CS is paired with a substance to be expelled, like acid (Pavlov, 1927). However, much less is known about the responses involved in the defensive CR. If the prefiguring interpretation of defensive CRs is correct, then the anticipatory conditional response should be finely adjusted to the metabolic and physiological requirements of the animal’s species-specific defense strategy and, together with skeletal behaviors, should reflect preparation for optimal defense (cf. Kimmel and Bums, 1975).

B.

INTERSPECIFICDEFENSE AND CONDITION A L RESPONSE

THE

BACKWARD

Of particular relevance for functional considerations is the phenomenon of “backward conditioning,” wherein the pairings of CS and US are reversed and the US precedes presentation of the CS. The backward CS can hardly serve as a signal for an event which has already occurred, and yet, sometimes, a backward CS elicits a response. But is this really Pavlovian conditioning‘?That is, is the CS-elicited response a result of an association having been formed between that specific CS and the US? Following a very thorough treatment of the question, Mackintosh (1974) concludes that “all in all there is little reason to accept the

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

37

reality of backward conditioning” (p. 60). By this he means that US-CS pairings are not sufficient to establish conditioning of the response typically produced by forward pairings of CS and US. Nevertheless, for our purposes here, the vicissitudes of this disenfranchised conditioning phenomenon provide further indication of the adaptiveness of anticipatory CRs. Positive reports of backward conditioning have been the exclusive property of preparations employing an aversive US (e.g., Switzer, 1930, demonstrating eyelid contraction CRs with an air-puff US; Wolfle, 1930, demonstrating finger withdrawal anticipatory to shock US; and Champion and Jones, 1961, demonstrating the GSR also in anticipation of a shock US). The occurrence of a CR in such experiments has since been interpreted as the result of “pseudoconditioning” (Grether, 1938; Harlow, 1939; Mowrer, 1960; Kimble, 1961; Prokasy er a l . , 1962) and not the backward pairings per se. That is, merely as a result of repeated exposure to the US, any external stimulus elicits a response similar to the unconditional response; previous pairing with the US is not necessary. In other words, animals who have been repeatedly “threatened” by shock or some other such aversive stimulus are likely to react “defensively” to most anything. This interpretation is consistent with the report of Singh (1959) that the efficacy with which a backward CS elicits a fear response interacts in rats with such factors as emotional reactivity (hereditary strain) and shock intensity. Wells (1968) suggests that pseudoconditioning is likely to be an adaptive response in an environment where biologically important stimuli, like predators, are not single isolated events. Evans (1966a,b) similarly interprets the pseudoconditioned retraction response of Neresis worms following repeated tactile or electrical stimulation. In this context the paradox of backward conditioningthat an animal would evidence a “preparatory” response to a stimulus which follows the biologically important event, a preparation which would seem to be totally inappropriate-is hardly an embarrassment for the prefiguring hypothesis. Quite the contrary. Equally interesting a suggestion for a functional approach to backward conditioning is that made by Pavlov (1928), which is that with a sufSicient number of pairings, a backward CS might become “inhibitory.” A CS is termed inhibitory when its presentation significantly decreases the probability of a particular CR (here, the “fear” CR which would be established by the normal forward pairing of CS and US). Such decreases in conditional fear responses have been observed by Kamin ( 1963), Moscovitch and LoLordo ( 1968), James (1 97 I ) , and S. Siege1 and Domjan (1971). Segundo et al. (1961) demonstrated, for example, that when a tone CS was presented near the end of a shock US, the tone elicited a behavioral pattern which the authors describe as “relaxed.” This postural relaxation was accompanied by inhibition (i.e., reduction in amplitude and duration) of the cortical EEG elicited by shock. A backward CS signals a period of time, until the next US-CS trial, in which the threat of danger is absent. So, too, does a shock stimulus itself when it is

38

K A R E N L. HOLLIS

presented alone. An animal’s initial response to shock is followed by a compensatory reaction which functions to restore physiological homeostasis. The inhibitory response elicited by a backward CS might more easily be thought of as a “learned” manifestation of this compensatory CR. If pain does not always terminate with the termination of a painful stimulus-and our own experience should indicate that it may not-a backward CS might be thought of as providing a safety signal (Moscovitch and LoLordo, 1968). That is, the backward CS provides additional information that danger has passed. The compensatory response, whose function it is to reduce the physiological turmoil (and probably the energetic cost) induced by aversive stimulation may “move forward” in time. In so doing, the overall net cost of the physiological trauma may thereby be reduced. But under what conditions, adaptively speaking, should a backward CS elicit a defensive CR or serve to inhibit that defensive CR? Pavlov (1928) suggested that backward conditioning might produce CRs initially and, with extended training, the backward CS would become inhibitory. Razran (1956, 1971) and Asratyan (1965) cite a number of studies in support of this contention. While Mackintosh (1974) convincingly argues that the initial CRs are, in fact, pseudoconditioned (and, therefore, not the result of the backward pairings per se), it matters not for our purposes here which process controls the initial fear increase. Whether pseudoconditioning or “true” Pavlovian conditioning, the result of shock-CS pairings is an initial wariness. The animal reacts quickly to any stimulus, lest it be dangerous. Later, when further experience indicates that fearful reactivity is an unnecessary cost in this particular situation, the same CS elicits a compensatory response which functions to reinstate homeostatic balance. C. INTRASPECIFICDEFENSE Pavlovian conditioning of intraspecific aggression has been demonstrated in rats (Lyon and Ozolins, 1970; Vernon and Ulrich, 1966; Creer et al., 1966; Farris et al., 1968), in pigeons (Rachlin, 1969; Rackham, 1971, cited in Moore, 1973), and in fish (Adler and Hogan, 1963; Thompson and Sturm, 1965; Murray, 1973). In these experiments the CS was paired either with shock (e.g., Lyon and Ozolins, 1970) or with a conspecific (e.g., Thompson and Sturm, 1965) and the aggressive behavior was elicited by the CS in anticipation of the shock or rival “aggressor.” Other experiments have shown that, when an animal is required to perform a response to avoid shock, the shock avoidance task is frequently accompanied by what appears to be classically conditioned aggressive behavior (Azrin e? al., 1967; Foree and LoLordo, 1970; Pear et al., 1972). One interpretation of these experiments (Mackintosh, 1974) is that certain external stimuli, such as the response manipulandum or warning signal, become associated with the shock US through Pavlovian conditioning pairings. These CSs

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

39

then elicit responses similar to those elicited by shock. For example, when Graf and Bitterman (1963) required that pigeons merely move about in the apparatus to avoid regularly scheduled shocks, the pigeons’ movement soon evolved into a threat display: “The birds tended to adopt a characteristic crouching posture which involved a marked lowering of the anterior portion of the body. The response. . . took the form of hopping or of stamping the feet, often accompanied by fluttering of the wings” (p. 305). Although the above experiments would suggest that naturally occurring signals might also be capable of eliciting aggressive CRs, additional evidence is required to show that these anticipatory conditional aggressive responses increase the likelihood of successful attack and defense. Such evidence is available for the classically conditioned aggressive display of anabantid fish. Because in anabantids (e.g., Berm splendens) the ability to defend a territory from intruders determines in large part whether an individual will reproduce (Forselius, 1957), these species offer a unique opportunity to test the prefiguring interpretation of anticipatory conditional aggressive responses. In few other situations, except courtship itself, would the biological advantage of signal-elicited behavior be so closely tied to actual reproductive gains. The laboratory and ethological investigations of anabantids which are relevant to this article’s thesis are described in detail below. The reproductive season of anabantids commences with the monsoons when freshwater pools are created. Males migrate to suitable nesting grounds. There they establish territories by building a foam bubble nest at the surface which is anchored to floating vegetation. Later, when the females arrive in the nesting area, they are courted in the temtory. Spawning takes place at the site of the bubble nest. Reproduction is almost entirely dependent upon successful territory defense, since females rarely mate with nontemtory owners (Forselius, 1957). Intruders, be they later arriving males in search of a territory themselves or neighbors attempting to expand their own territory, are met with a rigorous aggressive defense. The agonistic display sequence is described by Simpson (1968) and Forselius (1957). It consists of an immediate fin erection, an unfolding and spreading of the fins, which is accompanied by gill-cover erection, an outward extension of the opercula which, in Betta splendens, exposes the bright red gill membranes. The territory owner rushes to the sight of the intrusion, meeting the rival in a head-on posture. This “charging” is termed the frontal approach. Typically an intruder will terminate the encounter at this point by assuming the submissive posture or by fleeing. If the encounter is not terminated, tailbeating behavior, side-to-side undulations of the body which direct water currents at the opponent, and biting follow. The fin erection component of the agonistic display also occurs in the context of “patrolling” wherein a territory owner spontaneously (i.e., in the absence of an intruder) erects his fins and swims about the territory, especially near the borders (Forselius, 1957).

40

KAREN L . HOLLIS

Thompson and Sturm (1965) demonstrated that the various components of the aggressive display could be classically conditioned in Betta splendens. They employed a delay conditioning procedure in which a 10-sec red light CS was followed by a 15-sec exposure to a mirror US (cf. Adler and Hogan, 1963, for a similar demonstration with a shock CS). Following repeated pairings of CS and US, anticipatory conditional fin erections, gill-cover erections, and tailbeatings (undulatory movements) occurred in the presence of the CS. Their data further suggest that the relative rates of acquisition of these individual display components are different. In both fish conditional fin erection appeared earlier than either gill-cover erection or tailbeating; frontal approach was the last behavior to condition. This latter observation might appear to be a puzzling one in that the frontal approach to a rival typically precedes fin erection. However, the red light CS in this experiment illuminated an entire side wall of the tank and flooded the tank interior with a diffuse red glow (T. Thompson, personal communication); not surprisingly, approach behaviors are notoriously difficult to condition with diffuse CSs (e.g., Hearst, 1975; Wasserman, 1973). Using an omission training procedure, Murray (1973) has since demonstrated that, for frontal approach, fin erection, and gill erection, the fin and gill erection display components are more greatly influenced by the stimulus-reinforcer relationship (CS-US) than is frontal approach. When a 15-sec mirror exposure was made contingent upon not performing a fin erection, a gill erection, or a frontal approach to a 10-sec diffuse red light CS, this response contingency did not greatly affect performance of the fin erection and gill erection components. The frequency of frontal approach, however, decreased more substantially, implying greater control by the instrumental response contingency. Thus, the Pavlovian CS-US pairings are, in themselves, sufficient to produce fin and gill erections and such associations are little influenced by relationships between performance of the response and the reinforcer. This is to say that the CS comes to elicit anticipatory aggressive display behavior similar to that elicited by the US and it does so by virtue of its association with that US, not because of the consequences which result from the CR. The prefiguring hypothesis would maintain that the function of these anticipatory conditional aggressive behaviors is to increase the likelihood of successful territory defense. Naturally occurring stimulus events which would reliably accompany an intruder would serve, like the CSs of the laboratory, as “learned releasers” of the agonistic display. The result of this conditioning process is that an intruder would be met at the territory boundary by an already aggressively displaying owner who is, therefore, better prepared to do battle. An experiment now underway with blue gouramies (Tricogasfertricopterous) tends, so far, to support this hypothesis. Matched pairs of males living in visually isolated halves of a tank received, independenlly of one another, either Pavlovian conditioning pairings of a red light CS and rival presentation US or rival presentations alone

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

41

(US-only control). Following this training, the barrier separating the fish was removed and they confronted each other for the first time, both defending their home territories. This confrontation was signaled for the classically conditioned member but unsignaled for the control fish who had been trained with unsignaled presentations of a rival. The frequency of biting and tailbeating behaviors was analyzed for the first 3 min of the aggressive encounter, a time period chosen so as to encompass all territory disputes. Biting and tailbeating were selected for analysis because these behaviors are known to inflict injury upon an opponent (Forselius, 1057) and, as such, should provide an easily quantifiable index of aggressive strength. So far, two pairs have been tested using these procedures. The resulting aggressive encounter between pair members was clearly dominated by the Pavlovian conditioned animals, as shown in Fig. 5. The Pavlovian conditioned fish of Pair A exhibited 40% more bites than the unsignaled control animal. In Pair B , 33% more bites were performed by the Pavlovian conditioned fish. Further, neither of the control animals exhibited any tailbeating behavior, in contrast to the Pavlovian conditioned animals. Of course, more data (and additional control groups) are required to evaluate properly the predictions of the prefiguring hypothesis in this context. Nonetheless, such data will provide a simple and direct test of the functional explanation of anticipatory aggressive CRs. Parenthetically, it may be interesting to note here that experiments which have explored lesion-induced deficits in the aggressive behavior of Berta splendens suggest that the impairment may result, at least in part, from an inability to prepare for aggressive interactions via Pavlovian conditioning (Hollis, 1979). Telencephalon-ablated Bettus are rarely observed to initiate aggressive behavior, although when eventually elicited, both the vigor of the aggressive display and the coordination of the individual display components are unaffected (see Hollis and Overmier, 1978, for a review). In an experiment in which swimming through a tunnel was followed by mirror-image presentation, ablates performed this response as well as normal and sham-operated subjects (Hollis, 1979). These and other data reviewed by Hollis and Overmier (1978) suggest that ablates are not merely “unaroused” fish. Indeed, the pattern of deficits and competences seen in a Pavlovian conditioning experiment suggests another interpretation.. When a red light CS was paired with a mirror-image US, telencephalon-ablated fish performed fewer conditional fin erections, a response which the findings of control groups suggested was conditioned to generalized cues of the apparatus. In normal animals this response occurred asymptotically on 80-95% of the CS presentations. Conditioning of gill erection was unimpaired in ablates; however, in ablates as well as in normals, the gill erection CR was performed, at asymptote, on only 45% of CS presentations. This pattern could account for ablates’ specific impairment in naturalistic settings. Ablated fish, who are less capable of utilizing generalized signals (static cues) in the environment, and must

42

KAREN L . HOLLIS

50

40

0

t30 3

0 w IT LL

20

10

P

C

P

C

50

10

40

8

> 0

>

TAILBEATING

530

0

6 t

3

2 E

% I I :

20

4 k

10

2

P

C

P

C

FIG.5 . The frequency of biting and tailbeating behaviors during the first 3 min of an aggressive encounter in two pairs of blue gouramies, Tricogasrer tricoprerus. The values for the Pavlovian conditioned animals (P) are shown by the stippled bars, while the values for the unpaired control animals (C) are shown by the open bars.

rely upon more specific cues of territorial aggression are capable of preparing at a significantly lower proportion of the time. That is, the ability to respond to a discrete CS with a conditional gill erection (and, of course, simultaneous fin erection) may not compensate for the inability to perform this fin erection on other occasions when only more general cues are available. Clearly, additional research is needed to test the validity of the prefiguring hypothesis ’ explanation of conditional intraspecific defense. The rich literature of aggressive behavior (see, e.g., Wilson, 1975; Parker, 1974; R. Dawkins and Krebs, 1978) suggests many interesting experiments, however: Do location var-

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

43

iables, such as whether the animal is in the home territory or in novel surroundings, determine the form of the CR? A frequent observation (e.g., N. Tinbergen, 1953), is that animals wage a vigorous aggressive defense when they are on their home territories and yet, when they intrude upon the territory of another, they evidence submissive display behavior instead. Conditional responses might show similar effects of location variables. Likewise, does the history of a particular animal ’s aggressive encounters affect its tendency to evidence conditional aggression versus conditional submission? Further, the autonomic responses of conditional aggressive behavior remain a fertile ground of investigation. In humans, for example, participation in violent aggressive encounters induces a larger release of the endocrine hormone norepinephrine, a blood pressure regulator, than it does of epinephrine, the hormone which triggers the massive general alarm response mentioned earlier (Martin, 1976). However, anticiparion of aggressive interaction induces the release of epinephrine alone. If prefiguring maximizes interaction with the US, as has been suggested of other conditioned behavior contexts, the CR would be expected to be highly tailored to these and other variables affecting optimal defense strategies.

v.

PAVLOVIAN CONDITIONING OF REPRODUCTIVE BEHAVIOR

Sevenster ( 1 968) reported that when male three-spined sticklebacks (Gastrrosteus aculeatus) were required, in an instrumental learning task, to bite a rod to obtain a view of a gravid female, they soon began to court the rod, circling around it and performing the typical zigzag dance of the male stickleback. This behavior interfered with the performance of the instrumental response and thus resulted in a lower reward rate than that obtained with other responses. Later, Sevenster (1973) noted that the behavior elicited by the rod was “clearly related to Pavlovian or classical conditioning” but he preferred to use the term “association, ” since Pavlovian conditioning “currently refers to the conditioned stimulus as a signal for the response, not as its target” (p. 282, italics added). Of course, demonstrations that Pavlovian CSs would elicit signal-centered responses-the autoshaping phenomenon-were only beginning to emerge then. Since that time, however, Pavlovian conditioning of courtship behavior, which is very clearly directed at the CS, has been observed in quail (Farris, 1967) and in pigeons (Rackham, 1971, cited in Moore, 1973; see also Gilbertson, 1975, for observations of classically conditioned courtship behavior intruding upon an instrumental key peck response in pigeons). Insofar as classically conditioned courtship behavior is concerned, the denouement of the functional story will be obvious by now. As we shall see in this next section, the classically conditioned male does get the female in the endand he does so more quickly than if the female were not signaled. But the story

44

KAREN L . HOLLIS

does not end here. The successful courtship will have produced young and Pavlovian conditioning of parental behavior emerges to serve as an intermediary between the resources of the parent and the needs of the young. A.

COURTSHIP

The male Japanese quail, Coturnix coturnix japonica, reaches sexual maturity in only 28 days posthatch; thereafter, it pursues the female with an intensity for which it is renowned (Farris, 1964). The males are the first to arrive on the breeding grounds in the spring. There, they set up a territory which they defend pugnaciously against intruders, and then call to attract the later arriving females for which they compete. A female is courted within seconds upon arrival. A receptive female may squat immediately or run a short distance and squat, allowing the male to mount. Cooperation by the female is not necessary for successful mating, however, and forced copulations may be very frequent. Both courtship and copulation are brief. Under these conditions-territoriality, competition for females, abbreviated courtship, and frequent forced copulation-males which are able to react quickly to the appearance and arrival of a female are less likely to lose her to nearby competitors. A male must be able to commence courtship and/or copulation immediately. Farris (1967) was able to classically condition this courtship behavior. When males were exposed to repeated presentations of a buzzer CS followed, 10 sec later, by the introduction of a female (US), the buzzer soon began to elicit courtship behavior CRs. In some of the animals, conditioning of the initial courtship components was evident by the fifth pairing and the complete response as early as the fourteenth trial. The mean number of pairings for the complete sequence to emerge during the 10-sec CS period was 27.6 trials. Control animals, which received unpaired presentations of the buzzer and female did not evidence any courtship behavior in the presence of the buzzer, though they did eventually court the female when she was presented. Of immediate relevance to the prefiguring hypothesis is Farris’ (1964) observation that not only did the latency to initiate courtship rapidly decrease over trials in the classically conditioned animals, but in three of the four animals mounting took place immediately upon introduction of the female. This decrease in latency to court and immediate mounting was not described for control animals. Similar findings have recently been obtained with male rats (Zamble and Hadad, 1980). For classically conditioned animals, access to a female (US)was signaled by placing the male rat in a holding cage (CS) for 10 min prior to contact. Three Pavlovian conditioning groups were created based upon the extent of US (female) contact. Males in one group could copulate to ejaculation (EJ). In a second group, males could copulate but the trial was terminated after three intromissions (INT). And, in a third group, only contact with the female was

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

45

permitted (CON). Three corresponding EJ, INT, and CON groups served as controls; they received placement in the holding cage and access to the female at random times, separated by at least I hr. After eight training trials all subjects were tested by placing them in the CS box for 10 min and then giving them the opportunity to copulate with a female to ejaculation. The results appear in Fig. 6. Overall, the latency to ejaculate appears to be lower in the classically conditioned animals than in the unpaired control groups; however, type of contact with the female during training had a significant effect on ejaculatory latency. When the Pavlovian conditioning groups were compared with their corresponding control groups, only the CON and INT conditioning animals were significantly faster than their respective controls. Ejaculatory latencies of the paired and unpaired EJ groups were not significantly different from one another. Although these data were interpreted to suggest that Pavlovian conditioning occurs only when ejaculation is prevented during training trials, a learning vs performance distinction should be made here. There is usually some improvement in copulatory performance with experience, an effect which may itself be Pavlovian (based upon the sight or odor of a particular female, the location, etc.) and which may have contributed to the decreased ejaculatory latencies of both EJ groups. Had this effect been attenuated by, say, testing with a novel female (or in a different location), the contribution of the CS signaling operation may have been better revealed. CON

FIG.6. Mean ejaculatory latencies (geometric mean values) for each of three groups of rats differing in extent of female (US) contact during training. The groups were an Ejaculation group (EJ), an Intromission group (INT), and a Contact group (CON). The values for the Pavlovian conditioning animals (P) are shown by the stippled bars while the values for the unpaired control animals (C) are shown by the open bars. (After Zamble and Hadad, 1980.)

46

KAREN L. HOLLIS

The findings of Zamble and Hadad, as well as Farris (1964), provide relatively conclusive evidence that the effect of a Pavlovian CS which signals the appearance of a female is to elicit behaviors which decrease the amount of time required for reproductive activities. Although Zamble and Hadad do not provide a record of the behaviors elicited by the CS, on the basis of Farris’ data and those of Sevenster (1968, 1973), Rackham (1971), and Gilbertson (1975). one would expect that the rats engaged in behaviors preliminary to copulation. These behaviors insure that when the female does appear the male is ready to mate. Generally, the function of this CR would be to provide the classically conditioned male with a competitive advantage andlor to insure that courtship and mating occurred with greater efficiency and speed. A similar conclusion is reached by Graham and Desjardins (1980) who have recently reported that the secretion of both luteinizing hormone, a pituitary gonadotropin, and testosterone could be classically conditioned to an olfactory CS in adult male rats. One would expect that an investigation of the role played by various ecological features (including type of mating system, presence of territoriality, etc.) would reveal interesting differences in the form of classically conditioned reproductive behavior. B.

PARENTAL BEHAVIOR

Lactating in mammals and brooding in fowl would appear to be very different behaviors. Nonetheless, they possess two commonalities of significance to the prefiguring hypothesis: Both depend upon dynamic interactions between the parent and the young where signaling would appear to be advantageous; and, both are classically conditionable. An analysis of these behaviors demonstrates the importance of Pavlovian conditioning operations in the success of the parent-young interaction.

I . Conditioning of the Mammalian Milk-Ejection Reflex Efficient nursing of the young mammal is, in large part, dependent upon the synchronization of neonatal and maternal reflexes (Jelliffe and Jelliffe, 1978). Lactogenesis, or milk synthesis, commences during the latter part of gestation; however, full lactogenesis is stimulated by the postdelivery hormonal change and by sucking of the neonate. Milk ingestion is dependent upon rooting, sucking, and swallowing reflexes. However, the maternal let-down, or milk-ejection reflex, plays a crucial role in this process. This is because in all but a few mammals (e.g., cows, goats) nearly all of the milk obtained during a feeding must be “let down” from the alveoli, where it is secreted, to the milk ducts and lactiferous sinuses just behind the nipple (Grosvenor and Mena, 1974; Jelliffe and Jelliffe, 1978). Yet, let-down, a neuroendocrine reflex whose afferent stimulus is sucking of the nipple, does not occur immediately upon stimulation. In humans, for

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

47

example, let-down initially takes approximately 2-3 min in undisturbed primiparae (M. Woolridge, unpublished results). Given the numerous feeding bouts of many mammals, the long latency of the let-down reflex appears inefficient and disruptive for lactating females which must forage to meet their increased energy requirements. However, the latency of let-down may be shortened through Pavlovian conditioning. Pavlovian conditioning of the let-down reflex, elicited by visual and auditory CSs, has been demonstrated in a number of animals and in humans (Grosvenor and Mena, 1974; Caldeyro-Barcia, 1969). In some women, for example, it may be triggered by the sound of the crying infant or by stimuli which typically occur just prior to nursing. More general observations have suggested that, with experience, the let-down reflex occurs immediately upon sucking by the infant. Most authors attribute this experimental decrease of let-down latency to “learning” (Isbister, 1954; Jelliffe and Jelliffe, 1978); it is most certainly the operation of a Pavlovian conditional response. The implications of conditioned milk ejection are that the female spends less time nursing the young. The time thus saved can then be spent on other, equally important activities such as feeding by the mother, grooming of the young, and taking care of the nest. Through conditioning the resources of the mother can be more closely synchronized with the needs of the young-to the clear advantage of both. 2.

Conditioning of Brooding Solicitation in Chicks

An experiment by Wasserman (1973) suggests that Pavlovian conditioning plays a similar intermediary role in the relationship between the chick, which is unable to thermoregulate endogenously, and the hen, which provides heat to the chick through brooding. Wasserman (1973) placed 3-day-old domestic chicks in a cooled chamber and activated a heat lamp at random intervals. For half of his subjects, the paired stimuli group, the heat irradiation was signaled by an 8-sec illumination of a green key light. For the remaining subjects, the random stimuli group, heat irradiation and green key light illumination were scheduled independently of one another. After an average of only eight presentations of the key light CS and heat US, subjects in the paired stimuli group began to peck the key when it was illuminated. This behavior was maintained for the duration of the experiment. Subjects in the random condition, however, rarely evidenced the key-pecking behavior. Moreover, in a separate omission training experiment with experimentally naive chicks, Wasserman established that the Pavlovian stimulus-reinforcer relationship was sufficient to produce the key-peck CR; chicks whose pecks on the key prevented US occurrence on that trial nonetheless continued to key-peck (see also Wasserman et al., 1975). An interesting feature of the key-peck CR was that, over the course of repeated CS-US pairings, the response showed a gradual change from a forceful peck to a

48

KAREN L. HOLLIS

more gentle contact accompanied by side to side movements of the chick’s head. Wasserman called this behavior “snuggling. Not surprisingly, perhaps, these CRs mimic behavior evoked by naturalistic signals; the peck and snuggling behaviors are the means by which the chick induces the hen to brood (Hogan, 1974; see also Sherry, 1981). The chick approaches the hen and begins to peck the breast feathers. The pecking behavior is frequently followed by snuggling, a side-to-side motion of the head in which the chick pushes its beak up into the hen’s feathers. The hen then squats, extends her wings, and vocalizes, presumably to call her other chicks. Wasserman suggests a Pavlovian interpretation of brooding soliciting behavior in the chick: ”

If we assume that the broody hen serves the dual functions of a localized visual stimulus [CS] and a heat source [US] and that the sight and warmth of the hen have been repeatedly paired during the chick’s first week of life, then approach and contact may come 10 be controlled by the former stimulus property (Wasseman, 1974, p. 157).

Of course, this interpretation assumes that the soliciting behavior of the chick is learned-r is, at least, strongly influenced by learned behavior. Why some stimulus events, such as the visual characteristics of the hen, might be learned signals instead of innate sign stimuli is discussed in the next section; however, in this specific case one very likely explanation may be proposed. Because the visual stimulus characteristics of the parent are likely to change over successive generations, learning merely provides a more efficient means of attaching the instinctive behavior, brooding solicitation, to the appropriate stimulus object. This explanation is analogous to one suggested for imprinting in which recognition of the parent is also dependent upon a form of learning (Bateson, 1979). Assuming that some learning is involved in brooding solicitation by the chick, pecking and snuggling are signal-elicited Pavlovian CRs which serve, in turn, as a signal for the hen. Although the hen herself may frequently initiate brooding, each chick acquires a means through which to communicate its own immediate thermoregulatory needs. And thus the Pavlovian CR functions to optimize the interaction between parent and young.

VI. ECOLOGICAL IMPLICATIONS OF PREFIGURING The foregoing discussion of Pavlovian conditioning has centered on the prefiguring interpretation of the anticipatory CR. Several other issues, no less germane to a biological approach to learning, have received insufficient attention. One of these concerns the relationship of Pavlovian conditioning to other classes of learned behavior. If, as most theorists argue, some instrumental learningand, perhaps, even habituation-is based upon Pavlovian conditioning, then a functional understanding of Pavlovian conditioning has even broader implications for the understanding of behavior than those envisaged here. Another issue

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

49

concerns naturally occurring CSs, in particular how the form of the CS influences the anticipatory CR and why these signals are learned instead of innate. This final section is addressed to these two important issues. A.

PREFIGURING A N D NON-PAVLOVIAN LEARNING

Prior to the 1970s, many theorists believed that the learning of skeletal behaviors, like key-pecking and lever pressing, was the association of a particular response with a particular reinforcing stimulus. On the other hand, the learning of autonomic behavior, like salivation, was of a different form and consisted of the association of two stimuli, the CS and the US. However, at about that time, some researchers of instrumental conditioning (e.g., Brown and Jenkins, 1968; Staddon and Simmelhag, 1971) demonstrated that what was formerly believed to be the prototypical operant response, pigeons pecking a key for food reinforcement, could be generated by Pavlovian conditioning contingencies. This demonstration had tremendous theoretical import: Could changes in behavior which had previously been attributed to the instrumental response-reinforcer relationship be attributed instead to the Pavlovian stimulus-reinforcer relationship? The answer to this question was, of course, “Yes.” In fact, Pavlovian contingencies between stimulus events were so effective in producing “operant” behavior that the phenomenon became known as autoshaping: The animal’s behavior was automatically shaped without the experimenter having to do so by reinforcing successive approximations to the desired response. So persuasive has been the force of the autoshaping literature that the explanatory pendulum has reversed its direction. Whereas instrumental learning theorists once claimed that all leaming was based upon response-reinforcer relationships, some researchers, most notably Bindra (1 978), have suggested that Pavlovian conditioning is sufficient to account for all of instrumental learning. And, Wagner (1976) has recently suggested that habituation-the decrement in responding to a stimulus brought about by repeated presentations of that same stimulus-is a result of Pavlovian conditioning. Although Wagner uses the language of information processing models of memory to explain the mechanism, the Pavlovian conditioning paradigm forms its basis. Environmental cues which are paired with the repeatedly presented stimulus come to inhibit respondingn6 Although a single-process approach to learning may be attractive to some theorists, nature appears to be very distrustful of single systems and rarely are important tasks entrusted to the care of but one mechanism. Research on behaviors ranging from hunger to homing has forced us to accept the fact that back-up systems, multiple mechanisms, and even redundancies are the rule 61f habituation is, in fact, Pavlovian (and some would argue that it is not, e . g . , Mackintosh, 1974). then the prefiguring hypothesis would suggest that the function of the CR-inhibition of responding elicited by contextual cues-is to prevent the occurrence of unnecessary, energy-wasting behaviors.

50

K A R E N L . HOLLIS

rather than the exception. So far as learning is concerned, many researchers (e.g., Jenkins, 1977; Boakes et al., 1978; Bolles et al., 1980) have adduced evidence in support of the mutual participation of both conditioning processes in learning. For example, whereas Pavlovian conditioning seems to be important in the acquisition of the learned behavior, instrumental learning may be important in its maintenance (Bolles et al., 1980) or, perhaps, in the gradual and selective adjustment of the CR to a specific US. Whatever the outcome of this causal mechanism controversy, it is clear that Pavlovian conditioning plays a much larger role in learned behavior than previously thought and, often perhaps, presently acknowledged. B.

NATURALLY OCCURRING CONDITIONAL STIMULI AS LEARNED RELEASERS

A biological approach requires that the Pavlovian conditional stimulus be slightly recast. Thus the CS has been interpreted as a “learned releaser” of species-specific behavior (Woodruff and Williams, 1976), a concept similar to that of the sign stimulus (N. Tinbergen, 195 I). Although, unlike sign stimuli, naturally occurring CSs are not innately capable of releasing a species-specific behavioral sequence, their reliable association with biologically important events eventually establishes their capacity to do so. Laboratory CSs merely mimic these naturally occurring learned releasers. Similar ideas have been expressed by Woodruff and Williams (1976) and by Jenkins et al. (1978). Recasting the CS in this way raises several questions about naturally occurring signals. The first of these questions is whether, for any given US, the CS itself has any influence on the CR. A limiting case of this question was discussed in the section on food aversion learning where evidence was presented which suggests that animals may be predisposed to associate particular CSs with internal malaise. LoLordo (1979) cites many other examples of similar predispositions. But, generally, might the form of the CS also impose qualitive limitations on the CR; that is, might the CR which is elicited by one CS be different from the CR which is elicited by a different CS? The prefiguring hypothesis would necessarily predict that it would. To cite but one example, the form of naturally occurring signals would vary with the proximity of the US.In animals, such as deer, which are capable of detecting olfactory signals at longer ranges than visual or auditory signals, olfactory CSs probably evoke a different form of evasive behavior than visual or auditory CSs. And, the finding that the conditional response magnitude increases with CS intensity (see Gray, 1965, for a review) may be similarly interpreted, functionally speaking, since more intense signals are likely to be nearer. Holland (1977, 1980) has shown in a series of experiments that, for a given US, the nature of the Pavlovian CS does determine the form of the conditional response. Visual CSs which signal food elicit a CR which is substantially dif-

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

51

ferent from that elicited by auditory CSs. In hungry rats, rearing forms a predominant part of the CR only when the CS is visual, whereas head-jerking appears only in the presence of auditory CSs. (These CRs are, of course, not evidenced by the appropriate conditioning control groups.) Further, diffuse CSs elicit very different conditional responses than those elicited by discretely localized CSs, the latter evoking more contact behavior (Hearst, 1975). Put another way, all of these findings might be said to demonstrate certain constraints imposed upon the CR by the choice of the CS. Whether these constraints are merely a reflection of learning about an environment which is similarly constrained, or whether they are wired into the animal’s nervous system, remains to be determined. Nonetheless, on a functional level, such constraints demonstrate the fine adjustment of the animal to its environment. Additional research on this topic would help us to appreciate better the adaptive advantage of the CR. The second question is “Why are the signal events learned releasers and not unconditional stimuli?” That is, if the CSs are reliable predictors of biologically important events, it is difficult to understand why natural selection has forced animals to depend upon learning. The answer is, probably, economy. The animal would have to possess a multitude of innate releasing mechanisms which would cover all possible signal situations potentially encountered by any individual member of the species. Although our present state of knowledge does not allow us to affix an exact price tag on either the physiological or genetic machinery required of this particular solution, it is bound to be more expensive than that required of a simple, yet flexible, learning mechanism. Pavlovian conditioning is tailored to each animal’s particular environment and life history. Not only does this enable an animal to associate many different signal events with the same natural releaser but it allows for the independent dissolution of particular associations (extinction) when a CS ceases to be a reliable predictor. Despite this flexibility, the Pavlovian mechanism is protected from associating spurious signals with biologically important events. Latent inhibition (see Section 111,E,2) and blocking are examples of phenomena wherein the formation of invalid associations is retarded. Blocking (Kamin, 1969) of a CS-US association results when one CS is first established as a reliable predictor of the US (is paired on several occasions) and then, later, is accompanied by another, novel, CS. This compound CS is paired with the US a number of times. Yet, in a subsequent test of the individual CSs, the novel CS fails to elicit a CR. Learning about the novel CS is said to have been blocked by presenting it with a stronger elicitor. In the wild, survival depends upon reliable signals; and, where reliable signals are available, associations between casually cooccumng events are dangerous ones. C.

CONCLUDING COMMENTS

I have attempted to show how Pavlovian conditioning might operate in a number of biological contexts; however, my purpose is not to advocate the

52

KAREN L . HOLLIS

universal importance of Pavlovian conditioning in all aspects of every individual’s daily life. Mine is the more modest aim to suggest that Pavlovian conditioning may transform some of the environment’s capriciousness, and that the Pavlovian conditional response plays an important role in optimizing interactions with biologically important events. Because the CR is often energetically demanding, however, performance of a CR would only guarantee a net benefit if it could somehow be restricted to reliable and valid signals. That is, the Pavlovian conditioning mechanism must be able to isolate, in some sense, the causes of events. As a mechanism with which to decipher cause and effect the simplicity and economy of Pavlovian conditioning belies its sophistication. Not only do latent inhibition and blocking tend to efface spurious signals, but the extinction process denigrates those whose future reliability is questionable. Further, the sorts of conditioning phenomena described by Rescorla (1 980a) suggest that animals can piece together separate associations in a two-stage process. Since underlying physiological states are known to raise or lower the threshold of both URs and CRs, the animal’s attention is focused on only those signals relevant to its immediate needs. In short, Pavlovian conditioning, one of, perhaps, several learning mechanisms, helps to insure that food and water, mates, predators, and rivals are all predictable events for which the animal can efficiently and optimally prepare. To borrow Niko Tinbergen’s apt phrase, Pavlovian conditional stimuli are truly “signals for survival.” But a simple description of these learning rules, indicating the conditions under which the CR does or does not o c c u r - o r what form it may take-could not, in itself, provide an adequate functional analysis of Pavlovian conditioning. The prefiguring hypothesis is, however, an attempt to do so. It is an explanation of why, for example, Pavlovian conditioning results in an anticipatory response, w’hy the CR is sometimes directed at the signals themselves and sometimes not, and why the form of the CR may vary from situation to situation or change with experience. That is, the prefiguring hypothesis provides an explanation, at the functional level, of the often bewildering variety of response rules which characterize the many phenomena of Pavlovian conditioning. Last, it is hoped that the predictions of the prefiguring hypothesis will provide a useful heuristic with which to approach the biology of learning. Acknowledgments This paper was written during my tenure as a Postdoctoral Research Student in the Animal Behaviour Research Group, Department of Zoology, Oxford. I would like to thank David McFarland for making research facilities available to me and for his most generous hospitality. Together with the members of the A.B.R.G. they provided an intellectually exciting and congenial atmosphere in which to work. I would also like to thank Richard Dawkins, Robert Hinde. John Krebs, Jay Rosenblatt, David Sherry, and Sara Shettleworth for providing many useful comments on the manuscript, and Alasdair Houston for helpful discussion.

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

53

References Adler, N., and Hogan, J . A. 1963. Classical conditioning and punishment of an instinctive response in Eetta splendens. Anitn. Eehav. 11, 351-354. Alcock. J. 1973. Cues used in searching for food by red-winged blackbirds (Agelaius phoeniceus). Eehaviour 46, 174- 188. Asratyan, E. A. 1965. “Conditioned Reflex and Compensatory Mechanism. ” Pergamon, Oxford. Assenmacher, I. 1973. The peripheral endocrine glands. In “Avian Biology” (D. S.Farner and J. R. King, eds.), Vol. 111, pp. 183-286. Academic Press, New York. Atnip. G. W. 1977. Stimulus- and response-reinforcer contingencies in autoshaping, operant. classical, and omission training procedures in rats. J . Exp. Anal. Eehuv. 28, 59-69. Azrin, N . H.. Hutchinson, R. R., and Hake, D. F. 1967. Attack, avoidance, and escape reactions to aversive shock. J . Exp. Anal. Eehav. 10, 131-148. Balagura, S. 1968. Conditioned glycemic responses in the control of food intake. J . Comp. Physiol. P v c h o l . 65, 30-32. Barelare, B., Jr., Holt. L. E., Jr., and Richter, C. P. 1938. Influence of vitamin deficiencies on appetite for particular foodstuffs. Am. J . Physiol. 123, 7-8. Bateson, P. 1979. How do sensitive periods arise and what are they for? h i m . Eehav. 27,470-486. Belkin. D. A. 1968. Bradycardia in response to threat. Am. Zool. 8, 775. Best, P. J . . Best. M. R.. and Mickley. G. A. 1973. Conditioned aversion to distinct environmental stimuli resulting from gastrointestinal distress. J . Comp. Physiol. Psychol. 85, 250-257. Bindra, D. 1978. How adaptive behavior is produced: A perceptual-motivational alternative to response-reinforcement. Eehav. Erain Sci. 1, 41-52. Black, A. H. 1959. Heart rate changes during avoidance learning in dogs. Can. J . Psychol. 13, 229-242. Black, A. H. 1965. Cardiac conditioning in curarized dogs: The relationship between heart rate and skeletal behavior. In “Classical Conditioning: A Symposium” (W.F. Prokasy, ed.), pp. 20-47. Appleton, New York. Black. A. H. 1971. Autonomic aversive conditioning in infrahuman subjects. I n “Aversive Conditioning and Learning” (F. R. Brush, ed.), pp. 3-104. Academic Press, New York. Black, A. H., and Dalton, A. J . 1965. The relationship between the avoidance response and subsequent changes in heart rate. Acta B i d . Exp. 25, 107-1 19. Black, A. H.. and deToledo. L. 1972. The relationship among classically conditioned responses: Heart rate and skeletal behavior. In “Classical Conditioning 11: Current Research and Theory” (A. H. Black and W . F. Prokasy, eds.), pp. 290-311. Appleton, New York. Black, R. W . , and Black, P. E. 1967. Heart rate conditioning as a function of interstimulus interval in rats. Psychon. Sci. 8, 219-220. Blanchard, R. 1 . . and Blanchard, D. C. 1969a. Crouching as an index of fear. J. Comp. Physiol. P q c h o l . 67, 370-375. Blanchard, R. J . , and Blanchard, D. C. 1969b. Passive and active reactions to fear-eliciting stimuli. J . Comp. Physiol. Psychol. 68, 129-135. Blanchard, R. J . , Fukunga. K . K . . and Blanchard D. C. 1976. Environmental control of defensive reactions to a cat. Bull. Psychon. Soc. 8, 179-181. Boakes, R. A., Poli. M., Lockwood, M. J . , and Goodall. G. 1978. A study of misbehavior: Token reinforcement in the rat. J . Exp. Anal. Eehav. 29, 115-134. Bolles. R. C. 1975. “Learning Theory.” Holt, New York. Bolles, R. C., Holtz, R., Dunn, T., and Hill, W. 1980. Comparisons of stimulus learning and response learning in a punishment situation. h a m . Motiv. 11, 78-96. Breland, K . , and Breland, M. 1961. The misbehavior of organisms. Am. Psychol. 16, 661-664. Breland. K.. and Breland, M. 1966. ”Animal Behavior.” Macmillan, New York.

54

KAREN L. HOLLIS

Brogden, W. J.. Lipman, E. A.. and Culler, E. 1938. The role of incentive in conditioning and extinction. Am. J . Psychol. 51, 109-117. Brower, T. P. 1969. Ecological chemistry. Sci. Am. 220, 22-29. Brown, P. L., and Jenkins, H. M. 1968. Auto-shaping of the pigeon’s key-peck. J. Exp. Anal. Behuv. 11, 1-8. Bykov, K . M. (Ed.) 1958. “Text-Book of Physiology.” Foreign Languages Publishing House, Moscow. Bykov, K . M. 1959. “The Cerebral Cortex and the Internal Organs.” Foreign Languages Publishing House,Moscow. Caldeyro-Barcia. R. 1969. Milk ejection in women. I n “Lactogenesis: The Initiation of Milk Secretion at Parturition” (M. Reynolds and S. J. Folley, eds.), pp. 229-243. Univ. of Pennsylvania Press, Philadelphia, Pennsylvania. Cannon, W. B. 1932. “The Wisdom of the Body.” Norton, New York. Cappell, H . , and LeBlanc, A. E. 1977. Gustatory avoidance conditioning by drugs of abuse: Relationships to general issues in research on drug dependence. In “Food Aversion Learning” (N. W. Milgram, L. Krames, and T. M. Alloway, eds.), pp. 133-167. Plenum, New York. Capretta, P. J., and Moore, M. J. 1970. Appropriateness of reinforcement to cue in the conditioning of food aversions in chickens (Gallus gallus). J . Comp. Physiol. Psychol. 7 2 , 85-89. Carlson, N. J . 1960. Primary and secondary reward in traumatic avoidance learning. J . Comp. Physiol. Psycho/. 53, 336-340. Champion, R. A,, and Jones, J . E. 1961. Forward, backward, and pseudo-conditioning of the GSR. J . EXP. Psychd. 6 2 , 58-61, Chase, W. G., Graham. F. K., and Graham, D. T. 1968. Components of HR response in anticipation of reaction time and exercise tasks. J. Exp. Psychol. 76, 642-648. Church, R. M., LoLordo, V., Overmier, J . B., Solomon, R. L., and Turner, L. H. 1966. Cardiac responses to shock in curarized dogs: Effects of shock intensity and duration, warning signal. and prior experience with shock. J . Cnmp. Physial. Psychol. 6 2 , 1-7. Cohen. D. H.. and Durkovic, R. C. 1966. Cardiac and respiratory conditioning, differentiation, and extinction in the pigeon. J . Exp. Anal. Behav. 9, 681-688. Craig, W. 1918. Appetites and aversions as constituents of instincts. B i d . Bull. Woods Hole 34, 9 1- 107. Creer, T. L., Hitzing, E. W., and Schaeffer, R. W. 1966. Classical conditioning of reflexive fighting. Psychon. Sci. 4, 89-90. Croze, H. 1970. Searching image in carrion crows. Z. Tierpsychol.. Suppl. 5 , 1-86. Culler. E. A. 1938. Recent advances in some concepts ofconditioning. Psychol. Rev. 45, 134-153. Curio, E. 1975. The functional organization of anti-predator behaviour in the pied flycatcher: A study of avian visual perception. Anim. Behav. 2 3 , 1-1 IS. Czaplicki, J. A., Borreback, D. E., and Wilcoxin, H. C. 1976. Stimulus generalization of an illness-induced aversion to different intensities of colored water in Japanese quail. h i m . Learn. Behav. 4, 45-48. Davies, N. B., and Krebs, J . R. 1978. Introduction: Ecology, natural selection and social behaviour. In “Behavioural Ecology: An Evolutionary Approach” (J. R. Krebs and N. B. Davies, eds.), pp. 1-18, Blackwell Scientific, Oxford. Dawkins, M. 1971. Perceptual changes in chicks: Another look at the ‘search image’ concept. h i m . Behav. 19, 566-574. Dawkins, R., and Krebs, J . R. 1978. Animal signals: Information or manipulation? In “Behavioural Ecology: An Evolutionary Approach” (J. R. Krebs and N. B. Davies, eds.), pp. 282-309. Blackwell Scientific, Oxford. Dawkins, R., and Krebs, J . R. 1979. Arms races between and within species. Proc. R. Soc. London. B 205, 489-51 1 . deToledo, L. 1968. “Changes in Heart Rate and Skeletal Activity during Conditioned Suppression in Rats.” Unpublished Ph.D. Dissertation, McMaster University, Hamilton. Ontario, Canada.

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

55

deToledo, L., and Black, A. H. 1966. Heart rate: Changes during conditioned suppression in rats. Science 152, 14041406 Deutsch, R. 1974. Conditioned hypoglycemia: A mechanism for saccharin-induced sensitivity to insulin in the rat. J. Comp. Physiol. Psychol. 86, 350-358. Domjan, M. 1973. Role of ingestion in odor-toxicosis learning in the rat. J. Comp. Physiol. Psychol. 84, 507-521. Domjan, M., and Wilson, N . E. 1972. Contribution of ingestive behaviors to taste-aversion learning in the rat. J. Comp. Physiol. Psychol. 80, 403-412. Durlach, P. J., and Rescorla, R. A. 1980. Potentiation rather than overshadowing in flavor-aversion learning: An analysis in terms of within-compound associations. J. Exp. Psychol., Anim. Behav. Proc. 6 , 175-187. Dykman, R. A,, Mack, R. L., and Ackerman, P. T. 1965. The evaluation of autonomic and motor components of the nonavoidance conditioned response in the dog. Psychophysiology 1, 209230. Edmunds, M. 1974. “Defence in Animals.” Longman, Harlow, Essex, England. Evans, S. M. 1966a. Non-associative avoidance learning in Nereid polychaetes. Anim. Behav. 14, 102- 106. Evans, S . M. 1966b. Non-associative behavioural modifications in the polychaete Nereis diversicolor. Anim. Behav. 14, 107-1 19. Ewer, R. F. 1963. The behaviour of the meerkat Suricara suricata (Schreber). Z . Tierpsychol. 20, 570-607. Ewer, R. F. 1969. The “instinct to teach. Nature (London) 222, 698. Fanselow. M. S . 1980. Signalled shock-free periods and preference for signalled shock. J. Exp. P q c h o l . . Anim. Behav. Proc. 6 , 65-80. Fanis, H. E. 1964. “Behavioral Development, Social Organization, and Conditioning of Courting Behavior in the Japanese Quail, Coturnix corurnixjaponica. ” Unpublished Ph.D. Dissertation, University of Michigan, Ann Arbor, Michigan. Fanis, H. E. 1967. Classical conditioning of courting behavior in the Japanese quail, Corurnix eorurnix japonica. J . Exp. Anal. Behav. 10, 213-217. Farris, H. E., Gideon, B. E., and Ulrich, R. E. 1968. Classical conditioning of aggression: A developmental study. Paper presented to the American Psychological Association, San Francisco. Fehr, F. S., and Stem, J . A. 1965. Heart rate conditioning in the rat. J. Psychosom. Res. 8, 44-453. Fitzgerald, R. D., and Walloch, R. A. 1966. Changes in respiration and the form of the heart-rate CR in dogs. Psychon. Sci. 5, 425-426. Foree, D. D., and LoLordo, V. M. 1970. Signalled and unsignalled free-operant avoidance in the pigeon. J. Exp. Anal. Behav. 13, 283-290. Forselius, S . 1957. Studies of Anabantid fishes. Zool. Bidr. Uppsala 32, 93-597. Friedman, M. I . , and Stricker, E. M. 1976. The physiological psychology of hunger: A physiological perspective. Psychol. Rev. 83, 409-431. Galef, B. G . , Jr. 1976. Social transmission of acquired behavior: A discussion of tradition and social learning in vertebrates. In “Advances in the Study of Behavior” (J. S. Rosenblatt, R. A. Hinde, E. Shaw, and C. Beer, eds.), Vol. 6. pp. 77-100. Academic Press, New York. Garcia, J . , and Koelling, R. A. 1966. Relation of cue to consequence in avoidance learning. Pqchon. Sci. 4, 123-124. Garcia, J.. Kimeldorf. D. J., and Koelling, R. A. 1955. Conditioned aversion to saccharin resulting from exposure to gamma radiation. Science 122, 157-158. Garcia, J.. Kimeldorf, D. J.. and Hunt. E. L. 1956. Conditioned responses to manipulative procedures resulting from exposure to gamma radiation. Radiat. Res. 5, 79-87. Garcia, J . , McGowan. B. K . . Ervin, F. R.. and Koelling, R. A. 1968. Cues: Theirrelativeeffectiveness as a function of the reinforcer. Science 160, 794-795. ”

56

KAREN L. HOLLIS

Garcia, J . . Kovner, R.. and Green, K . F. 1970. Cue properties versus palatability of flavors in avoidance learning. P s y h o n . Sci. 20, 313-3 14. Gardner, W. M. 1969. Auto-shaping in bobwhite quail. J . Exp. Anal. Eehav. 12, 279-281. Gilbertson, D. W. 1975. Courtship as a reinforcement for key pecking in the pigeon, Culumbu livia. Anini. Eehuv. 23, 735-744 Gillette, K., Martin, G. M.. and Bellingham, W. P. 1980. Differential use of food and watercues in the formation of conditioned aversions by domestic chicks (Callusgallus). J . Exp. P s w h o l . , Anitn. Behut,. Procases 6, 99- I I I . Goldstein. A., Aronow, L.. and Kalman, S. M. 1974. “Principles of Drug Action: The Basis of Pharmacology,” 2nd ed. Wiley, New York. Graf, V.. and Bittenan, M. E. 1963. General activity as instrumental: Application to avoidance training. J . Exp. Anal. Eehav. 6, 301-305. Graham. J . M., and Desjardins. C. 1980. Classical conditioning: Induction of luteinizing hormone and testosterone secretion in anticipation of sexual activity. Science 210, 1039-1041. Grastyin, E., and Vereczkei. L. 1974. Effects of spatial separation of the conditioned signal from the reinforcement. Eehuv. B i d . 10, I 2 I- 146. Gray. J . A. 1965. Stimulus intensity dynamism. Psychul. Bull. 63, 180-196. Gray. J . A. 1971. ”Psychology of Fear and Stress.” Weidenfeld and Nicolson, London. Grether, W. F. 1938. Pseudo-conditioning without paired stimulation encountered in attempted backward conditioning. J . Cumnp. P g c h o l . 25, 9 1-96, Grosvenor, C. E., and Mena, F. 1974. Neural and hormonal control of milk secretion and milk ejection. In “Lactation: A Comprehensive Treatise” (B. L. Larson and V . R. Smith, eds.). Vol. 1 , pp, 227-276. Academic Press. New York. Guha. D.. Dutta, S . N . and Pradhan, S. N. 1974. Conditioningof gastric secretion by epinephrine in rats. Pruc. Soc. Exp. Biol. Med. 147, 817-819. Harlow, H. F. 1939. Forward conditioning, backward conditioning and pseudoconditioning in the goldfish. J . Cen. Psychol. 55, 49-58. H.arris, J . D. 1943. Studies of non-associative factors inherent in conditioning. Conip. Psychol. Monogr. 18 ( I . Whole No. 33). Harvey, P. H., and Greenwood, P. J . 1978. Anti-predator defense strategies: Some evolutionary problems. 1n “Behavioural Ecology: An Evolutionary Approach” ( J . R. Krebs and N . B. Davies, eds.), pp. 129-151. Blackwell Scientific, Oxford. Hearst. E. 1975. Pavlovian conditioning and directed movements. In “The Psychology of Learning and Motivation: Advances in Research and Theory” (G. H. Bower, ed.), Vol. 9. pp. 215-262. Academic Press, New York. Hearst. E.. and Franklin, S. R. 1977. Positive and negative relations between a signal and food: Approach-withdrawal behavior to the signal. J . Exp. P s y h o l . . Anim. Eehav. Pruc. 3, 37-52. Hearst. E . . and Jenkins, H. M. 1974. Sign tracking: The stimulus-reinforcer relation and directed action. Munugr. Psvchonom Suc.., Austin, Texas. Hermstein, R. J . , and Loveland, D. H. 1972. Food-avoidance in hungry pigeons, and other perplexities. J . Exp. A n d . Eehav. 18, 369-383. Hilgard. E. R . 1931. Conditioned eyelid reactions to a light stimulus based on the reflex wink to sound. Psycho/. Monogr. 41 ( I , Whole No. 184). Hinde, R. A. 1954. Factors governing the changes in strength of a partially inborn response, as shown by the mobbing behaviour of the chaffinch (Fringilla corlebs) I. The nature of the response, and an examination of its course. Pruc. R. Suc. London Ser. E 142, 306-331. Hinde. R. A. 1975. The concept of function. In “Function and Evolution in Behaviour: Essays in Honour of Professor Niko Tinbergen, F.R.S.” (G.Baerends, C. Beer, and A. Manning, eds.), pp. 3-15. Clarendon, Oxford. Hinde. R. A , . and Stevenson-Hinde, J. (Eds.) 1973. “Constraints on Learning: Limitations and Predispositions.” Academic Press, New York.

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

57

Hinson, R. E.. and Siegel, S . 1980. The contribution of Pavlovian conditioning to ethanol tolerance and dependence. In “Alcohol Tolerance, Dependence, and Addiction“ (H. Rigter and 1. C. Crabbe. Jr.. eds), pp. 179-197. ElseviedNorth-Holland Biomedical Press. Amsterdam. Hogan, J . A . 1973a. Development of food recognition in young chicks: I. Maturation and nutrition. J . Comp. Physiol. Psycho/. 83, 355-366. Hogan, J . A. 1973b. Development of food recognition in young chicks: 11. Learned associations over long delays. J. Comp. Physicil. P q c h o l . 83, 367-373. Hogan, J . A. 1973~.How young chicks learn to recognize food. In “Constraints on Learning: Limitations and Predispositions” (R. A . Hinde and J . Stevenson-Hinde, eds.), pp. 119-139. Academic Press, New York. Hogan, 1. A. 1974. Responses in Pavlovian conditioning studies. Science 186, 156-157. Holdstock, T. L., and Schwartzbaum. J . S . 1965. Classical conditioning of heart rate and galvanic skin response in the rat. Psychophysiology 2, 25-38. Holland, P. C. 1977. Conditioned stimulus as a determinant of the form of the Pavlovian conditioned response. J . Exp. Psychol., Anim. Behav. Proc. 3, 77-104. Holland. P. C. 1980. CS-US interval as a determinant of the form of Pavlovian appetitive conditioned responses. J . Exp. Psychol., Anim. Behav. Proc. 6 , 155-174. Hollis. K . L. 1979. ”The Effect of Telencephalon Ablation Upon the Reinforcing and Eliciting Properties of Species-Specific Events in Berm splendens. Unpublished Ph.D. Dissertation, University of Minnesota, Minneapolis, Minnesota. Hollis. K . L.. and Overmier. J. B. 1978. The function of the teleost telencephalon in behavior: A reinforcement mediator. In “The Behavior of Fish and Other Aquatic Animals’’ (D. I. Mostofsky, ed.), pp. 137-195. Academic Press, New York. Humphries, D. A.. and Driver. P. M. 1971. Protean defense by prey animals. Oecolugia 5 , 285302. Isbister. C. 1954. A clinical study of the draught reflex in human lactation. Arch. Dis. Child. 29, 66-72. James. J . P. 1971. Acquisition. extinction, and spontaneous recovery of conditioned supression of licking. Psychon. Sci, 22, 156-158. Jelliffe. D. B., and Jelliffe. E. F. 1978. “Human Milk in the Modern World: Psychosocial, Nutritional. and Economic Significance.” Oxford Univ. Press, London and New York. Jenkins, H. M. 1977. Sensitivity of different response systems to stimulus-reinforcer and responsereinforcer relations. In “Operant-Pavlovian Interactions” (H. Davis and H.M.B. Hurwitz, eds.), pp. 47-66. Erlbaum. Hillsdale, New Jersey. Jenkins, H. M., and Moore, B. R. 1973. The form of the auto-shaped response with food or water reinforcers. J . Exp. Anal. Behav. 20, 163-181. Jenkins, H. M., Barrera, F. J . , Ireland, C., and Woodside, B. 1978. Signal-centered action patterns of dogs in appetitive classical conditioning. Learn. Motiv. 9, 272-296. Kalat, J . W. 1977. Biological significance of food aversion learning. In “Food Aversion Learning” (N. W. Milgram. L. Krames. and T. M. Alloway, eds.). pp. 73-103. Plenum, New York. Kamin, L. J . 1956. The effects of termination of the CS and avoidance of the US on avoidance learning. J . Comp. Physiol. Psychol. 49, 420-424. Kamin, L. J . 1963. Backward conditioning and the conditioned emotional response. J . Comp. Physiol. Psychol. 56, 517-519. Kamin, L. 1. 1969. Predictability, surprise, attention, and Conditioning. In “Punishment and Aversive Behavior” (B. A. Campbell and R. M. Church, eds.), pp. 279-296. Appleton, New York. Karpicke. J . , Christoph. G., Peterson, G . , and Hearst, E. 1977. Signal location and positive versus negative conditioned suppression in the rat. J. Exp. Psychol., h i m . Behav. Proc. 3, 105-1 18. Kimble. G . A. 1961. “Hilgard and Marquis’ Conditioning and Learning,” 2nd ed. Appleton, New York. Kimmel, H. D., and Burns, R. A . 1975. Adaptational aspects of conditioning. In “Handbook of ”

58

KAREN L. HOLLIS

Learning and Cognitive Processes, Vol. 2: Conditioning and Behavior Theory” (W.K . Estes, ed.), pp. 99-142. Erlbaum, Hillsdale, New Jersey. Krebs, J. R. 1978. Optimal foraging: Decision rules for predators. In “Behavioural Ecology: An Evolutionary Approach” (J. R. Krebs and N . B. Davies, eds.), pp. 23-63. Blackwell Scientific, Oxford. Lacey, J. I . 1967. Somatic response patterning and stress: Some revisions of activation theory. In “Psychological Stress: Issues in Research” (M. H. Appley and R. Trumbull, eds.), pp. 14-42. Appleton, New York. Lacey. J . 1.. Kagan. J.. Lacey. B. C., and Moss, H. A. 1963. The visceral level: Situational determinants and behavioral correlates of autonomic response patterns. In “Expression of the Emotions in Man“ (P. H. Knapp, ed.), pp. 161-196. International Universities Press, New York. Lang, W., and Black, A. H. 1963. Cardiac conditioning in dogs as a function of US intensity and difficulty of differentiation. Paper presented at The Eastern Psychological Association Meeting. Larue, C. 1975. Comparison des effets de I’anosmie pkripherique et de la bulbectomie sur la sequence alimentaire du rat. J. Physiol. (Paris) 70, 299-306. Li,A. D., Poulos, C. X., and Cappell, H. 1979. Conditioned tolerance to the hypothermic effect of ethyl alcohol. Science 206, 1109-1 110. Lehrman, D. S. 1955. The physiological basis of parental feeding behaviour in the ring dove (Streptopeliu risoria). Behaviour 7 , 241 -286. Lewontin. R. C. 1979. Sociobiology as an adaptationist program. Behav. Sci. 24, 5-14. Lockard, J . S. 1963. Choice of a warning signal or no warning signal in an unavoidable shock situation. J . Comp. Physiol. Psycho/. 56, 526-530. LoLordo. V. M. 1979. Constraints on learning. In “Animal Learning: Survey and Analysis” (M. E. Bitterman, V. M. LoLordo, J. B. Overmier, and M. E. Rashotte, eds.), pp. 473-504. Plenum, New York. Lorenz, K. Z. 1969. Innate bases of learning. In “On the Biology of Learning” (K. H. Pribaum. ed.), pp. 13-93. Harcourt, New York. Lyon, D. 0.. and Ozolins, D. 1970. Pavlovian conditioning of shock-elicited aggression: A discrimination procedure. J. Exp. Anal. Behav. 13, 325-331. MacArthur, R. H., and Pianka, E. R. 1966. On the optimal use of a patchy environment. Am. Nat. 100, 603-609.

McCleary, R. A. 1954. Measurement of experimental anxiety in the rat: An attempt. J . Genet. Pgchol. 64, 95-108. McCleery, R. H. 1978. Optimal behaviour sequences and decision making. In “Behavioural Ecology: An Evolutionary Approach” (J. R. Krebs and N. B. Davies, eds.), pp. 377-410. Blackwell Scientific, Oxford. Mackintosh, N. J. 1974. “The Psychology of Animal Learning.” Academic Press, New York. Mackintosh, N. J. 1979. Functions of conditioning. Paper presented to a conference on “Functional Aspects of Learning,” Association for the Study of Animal Behaviour, London. Martin, C. R. 1976. “Textbook of Endocrine Physiology.” Oxford Univ. Press, London and New York. Maynard Smith, J. 1978. Optimization theory in evolution. Annu. Rev. Ecol. Sysr. 9, 31-56. Miller, N . E., Banks, J. H., and Caul, W.F. 1967. Cardiac conditioned responses in avoidance and yoked-control rats. Psychon. Sci. 9, 581-582. Millner, J. R., and Palfai, T. 1975. Metrazol impairs conditioned aversion produced by LiCI: A time dependent effect. Pharmacol., Biochem. Behav. 3, 201-204. Moore, B. R. 1973. The role of directed Pavlovian reactions in simple instrumental learning in the pigeon. I n “Constraints on Learning: Limitations and Predispositions’’ (R.A. Hinde and J. Stevenson-Hinde, eds.), pp. 159-186. Academic Press, New York.

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

59

Moscovitch, A,, and LoLordo, V. M. 1968. Role of safety in the Pavlovian backward fearconditioning procedure. J. Comp. Physiol. Psychol. 66, 673-678. Mowrer, 0. H. 1960. “Learning Theory and Behavior.” Wiley, New York. Murray, C. S . 1973. “Conditioning in Betra splendens. Unpublished Ph.D. Dissertation, University of Pennsylvania, Philadelphia, Pennsylvania. Murton, R. K . 1971. The significance of a specific search image in the feeding behaviour of the wood-pigeon. Behaviour 40, 10-42. Nachman, M., Rauschenberger, J., and Ashe, J . H. 1977. Stimulus characteristics in food aversion learning. In “Food Aversion Learning” (N. W. Milgram, L. Krames, and T. M. Alloway. eds.), pp. 105-131. Plenum, New York. Neuringer, A., and Neuringer, M. 1974. Learning by following a food source. Science 184, 10051008. Nicolaidis, S. 1977. Sensory-neuro-endocrine reflexes and their anticipatory and optimizing role on metabolism. I n “The Chemical Senses and Nutrition” (M. R. Kare and 0. Maller, eds.), pp. 123-143. Academic Press, New York. Norton-Griffiths, M. 1969. The organisation, control and development of parental feeding in the oystercatcher (Haematopus ostralegus). Behaviour 34, 55-1 14. Notterman, J . , Schoenfeld, W., and Bersh, P. 1952. Conditioned heart rate responses in human beings during experimental anxiety. J. Comp. Physiol. Psychol. 45, 1-8. Obrist, P. A., and Webb, R. A. 1967. Heart rate during conditioning in dogs: Relationship to somatic-motor activity. Psychophysiology 4, 7-34. Obrist, P. A., Webb, R. A., and Sutterer, J. R. 1969. Heart rate and somatic changes during aversive conditioning and a simple reaction task. Psychophysiology 5 , 696-723. Obrist, P. A., Webb, R. A., Sutterer. J. R., and Howard, J. L. 1970. Cardiac deceleration and reaction time: An evaluation of two hypotheses. Psychophysiology 6, 695-706. Obrist, P. A., Sutterer, J . R., and Howard, J. L. 1972. Preparatory cardiac changes: A psychobiological approach. In “Classical Conditioning 11: Current Research and Theory” (A. H. Black and W. F. Prokasy, eds.), pp. 312-340. Appleton, New York. Otis, L. S . , Cerf, J. A,, and Thomas G. J. 1957. Conditioned inhibition of respiration and heart rate in the goldfish. Science 126, 263-264. Overmier, J . B. 1966. Instrumental and cardiac indices of Pavlovian fear conditioning as a function of US duration. J. Comp. Physiol. Psychol. 6 2 , 15-20. Palmerino, C. C., Rusiniak, K . W., and Garcia, J . 1980. Flavor-illness aversions: The peculiar roles of odor and taste in memory for poison. Science 208, 753-755. Parker, G . A. 1974. Assessment strategy and the evolution of fighting behaviour. J. Theor. Biol. 47, 223-243. Parrish, J . 1967. Classical discrimination conditioning of heart rate and barpress suppression in the rat. Psychon. Sci. 9, 267-268. Patten, R. L., and Rudy, J. W. 1967. Orienting during classical conditioning: Acquired VS.unconditioned responding. Psychon. Sci. 7, 27-28. Pavlov, I. P. 1927. “Conditioned Reflexes: An Investigation of the Physiological Activity of the Cerebral Cortex” (Trans. by G . V. Anrep). Oxford Univ. Press, London and New York. Pavlov, 1. P. 1928. “Lectures on Conditioned Reflexes,” Vol. 2. “Conditioned Reflexes and Psychiatry” (Trans. by W. H.Gantt). International Pub]., 1941, New York. Pavlov, 1. P. 1934. An attempt at a physiological interpretation of obsessional neurosis and paranoia. J. Men!. Sci. 80, 187-197. Pear, J . J . , Moody, J . E., and Persinger, M. A. 1972. Lever attacking by rats during free-operant avoidance. J . Exp. Anal. Behav. 18, 517-523. . Perez-Cruet, J . , Tolliver, C., Dunn, C.. Marvin, S., and Brady, J. V . 1963. Concurrent measure”

60

KAREN L. HOLLIS

ment of heart rate and instrumental avoidance behavior in the Rhesus monkey. J. Exp. Anal. Brhav. 6 , 61-64. Perkins, C. C., Jr. 1968. An analysis of the concept of reinforcement. Psychol. Rev. 7 5 , 155-172. Peterson, G. B., Ackil, J . E., Frommer, G. P.,and Hearst, E. S. 1972. Conditioned approach and contact behavior toward signals for food or brain stimulation reinforcement. Science 177, 1009-101 I . Pietrewicz, A. T.. and Kamil, A. C. 1979. Search image formation in the blue jay. (Cyanocirta cristata). Science 204, 1332- 1333. Prokasy, W. F. 1965. Classical eyelid conditioning: Experimenter operations, task demands, and response shaping. In “Classical Conditioning: A Symposium” ( W . F. Prokasy, ed.), pp. 208-225. Appleton, New York. Prokasy, W. F., Hall, J . F., and Fawcett, J . T. 1962. Adaptation, sensitization, forward and backward conditioning. and pseudo-conditioning of the GSR. Psychul. Rep. 10, 103-106. Pyke, G . H.. Pulliani, H. R., and Charnov, E. L. 1977. Optimal foraging: A selective review of theory and tests. Q. Rev. B i d . 52, 137-154. Rachlin, H. 1969. Autoshaping of key pecking in pigeons with negative reinforcement. J . Exp. Anal. Behav. 12, 52 1-53 I . Rackham, D. W . 1971. “Conditioning of the Pigeon’s Courtship and Aggressive Display.” Unpublished M.A. Thesis, Dalhousie University, Halifax, Nova Scoria. Canada. Razran. G . 1956. Backward conditioning. Psychul. Bull. 53, 55-69. Razran. G. 1971. “Mind in Evolution: An East-West Synthesis of Learned Behavior and Cognition.” Houghton Mifflin, Boston, Massachusetts. Rescorla, R. A. 1980a. “Pavlovian Second-Order Conditioning: Studies in Associative Learning. ’’ Erlbaum, Hillsdale, New Jersey. Rescorla, R. A. 1980b. Simultaneous and successive associations in sensory preconditioning. J . Exp. P s y h o l . , Anim. Behav. Proc. 6 , 207-216. Rescorla, R. A,, and Wagner, A. R. 1972. A theory of Pavlovian conditioning: Variations in the effectiveness of reinforcement and nonreinforcement. In “Classical Conditioning 11: Current Research and Theory” (A. H. Black and W. F. Prokasy, eds.), pp. 64-99. Appleton, New York. Revusky, S . . and Garcia, J . 1970. Learned associations over long delays. In “The Psychology of Learning and Motivation: Advances in Research and Theory” ( G . H. Bower. ed.), Vol. 4, pp. 1-84. Academic Press, New York. Richter, C. P., Holt, L. E., Jr., Barelare, B., Jr., and Hawkes, C. D. 1938. Changes in fat, carbohydrate and protein appetite in vitamin B deficiency. Am. J. Physiol. 124, 596-602. Roll, D. L.,and Smith, J . C. 1972. Conditioned taste aversion in anesthetized rats. In “Biological Boundaries of Learning” (M. E. P. Seligman and J. L. Hager, eds.), pp. 98-102. Appleton, New York. Royama, T. 1970. Factors governing the hunting behaviour and selection of food by the great tit (Parus major L.).J . Anim. Ecol. 39, 619-668. Rozin, P. 1967. Specific aversions as a component of specific hungers. J. Comp. Physiol. Psycho/. 64, 237-242. Rozin, P. 1969. Central or peripheral mediation of learning with long CS-US intervals in the feeding system. J. Comp. Physiol. P s y h o l . 67, 421-429. Rozin, P. 1976. The selection of foods by rats, humans, and other animals. In “Advances in the Study of Behavior” (J. S. Rosenblatt, R. A. Hinde, E. Shaw, and C. Beer, eds), Vol. 6, pp. 21-76. Academic Press, New York. Rozin. P., and Kalat, J . 1971. Specific hungers and poison avoidance as adaptive specializations of learning. Psychol. Rev. 7 8 , 459-486.

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

61

Rozin. P., Gruss. L., and Berk, G. 1979. Reversal of innate aversions: Attempts to induce a preference for chili peppers in rats. J. Cornp. Physiol. Psychol. 93, 1001-1014. Ruff, R. L. 1971. “Telemetered Heart Rates of Free-Ranging Uinta Ground Squirrels in Response to Social Interactions.” Unpublished Ph.D. Dissertation, Utah State University, Logan, Utah. Russek. M.. and Pina, S. 1962. Conditioning of adrenalin anorexia. Nature (London) 193, 12961297. Rzoska, J . 1953. Bait shyness, a study in rat behaviour. Br. J. Anim. Behav. 1, 128-135. Santibanez, G . , Saavedra, M. A,, and Tisler, S. 1965. Further studies on conditioned bradycardia in cats. Acra B i d . Exp. 25, 363-372. Schmidt-Nielsen, K . 1959. Salt glands. Sci. A m . 200, 109-116. Schwartz, B. 1978. “Psychology of Learning and Behavior.” Norton, New York. Schwartz, B., and Gamzu, E. 1977. Pavlovian control of operant behavior. In “Handbook of Operant Behavior” ( W . K . Honig and J . E. R. Staddon, eds.), pp. 53-97. Prentice-Hall, Englewood Cliffs, New Jersey. Segundo, J . P.. Galeano, C., Sommer-Smith, J. A,, and Roig, J. A. 1961. Behavioural and EEG effects of tones ‘reinforced’ by cessation of painful stimuli. In “Brain Mechanisms and Learning” (J. F. Delafresnaye, ed.), pp. 265-291. Blackwell Scientific, Oxford. Selye, H. 1952. “The Story of the Adaptation Syndrome.” Acta, Montreal, Canada. Sevenster, P. 1968. Motivation and learning in sticklebacks. In “The Central Nervous System and Fish Behavior” (D. Ingle, ed.), pp. 233-245. Univ. of Chicago Press, Chicago, Illinois. Sevenster, P. 1973. Incompatibility of response and reward. In “Constraints on Learning: Limitations and predispositions” (R. A. Hinde and J . Stevenson-Hinde, eds.), pp. 265-283. Academic Press. New York. Sheam, D. 1961. Does the heart learn? Psychol. Bull. 58, 452-458. Sheffield, F. D. 1948. Avoidance training and the contiguity principle. J. Comp. Physiol. Psychol. 41, 165-177. Sheffield, F. D. 1965. Relation between classical conditioning and instrumental learning. In “Classical Conditioning: A Symposium” ( W . F. Prokasy, ed.), pp. 302-322. Appleton, New York. Sheffield, F. D., and Campbell, B. A. 1954. The role of experience in the “spontaneous” activity of hungry rats. J. Comp. Physiol. Psychol. 47, 97-100. Sheny, D. F. 1977. Parental food-calling and the role of the young in the Burmese red junglefowl (Callus gallus spadireus). Anim. Behav. 25, 594-601. Sheny, D. F. 1981. Parental care and the development of thermoregulation in red junglefowl. Behaviour, 76, 250-279. Shettleworth, S . J. 1972. Constraints on learning. In “Advances in the Study of Behavior” (D. S. Lehrman, R. A. Hinde, and E. Shaw, eds.), Vol. 4, pp. 1-68. Academic Press, New York. Shettleworth, S. J . 1978. Reinforcement and the organization of behavior in golden hamsters: Punishment of three action patterns. Learn. Moriv. 9, 99-123. Shettleworth, S. J . Function and mechanism in learning. In “Advances in Analysis of Behavior, Vol. 3: Biological Factors in Learning” (M. Zeiler and P. Harzem, eds.). Wiley, New York, in press. Sidman, M., and Fletcher, F. G. 1968. A demonstration of auto-shaping with monkeys. J. Exp. Anal. Behav. 11, 307-309. Siegel, R. K. 1979. Natural animal addictions: An ethological perspective. In “Psychopathology in Animals: Research and Clinical Implications” (J. D. Keehn, ed.): pp. 29-60. Academic Press, New York. Siegel, S. 1972. Conditioning of insulin-induced glycemia. J. Comp. Physiol. Psychol. 78, 233241.

62

KAREN L . HOLLIS

Siegel, S . 1975. Evidence from rats that morphine tolerance is a learned response. J. Comp. Physiol. P q c h o l . 89, 498-506. Siegel, S. 1976. Morphine analgesic tolerance: Its situation specificity supports a Pavlovian conditioning model. Science 193, 323-325. /Siegel, S . 1977a. Morphine tolerance acquisition as an associative process. J. Exp. Psychol., Anim. Behav. Proc. 3, 1-13. Siegel. S. 1977b. Learning and psychopharmacology. In “Psychopharmacology in the Practice of Medicine” (M. E. Jarvik, ed.), pp. 61-70. Appleton, New York. Siegel, S. 1978. Tolerance to the hyperthermic effect of morphine in the rat is a learned response. J. Comp. Physiol. Psychol. 92, 1137-1 149. Siegel, S. 1979a. The role of conditioning in drug tolerance and addiction. In “Psychopathology in Animals: Research and Clinical Implications” (J. D. Keehn, ed.), pp. 143-168. Academic Press, New York. Siegel, S . 1979b. Pharmacological learning and drug dependence. In “Research in Psychology and Medicine” (D. J. Obome, M.M. Gruneberg, and J. R. Eiser, eds.), pp. 127-134. Academic Press, New York. Siegel, S., and Domjan. M. 1971. Backward conditioning as an inhibitory procedure. Learn. Motiv. 2, 1-11. Simmons, K. E. L. 1952. The nature of the predator-reactions of breeding birds. Behaviour 4, 161-171.

Simpson. M. J . A. 1968. The display of the Siamese fighting fish, Berfa splendens. Anim. Behav. Monogr. 1, 1-73. Singh, P. J . 1959. Conditioned emotional response in the rat: I. Constitutional and situational determinants. J . Comp. Physiol. Psychol. 52, 574-578. Sinnott. J. M., Sachs, M. B., and Hienz, R. D. 1980. Aspects of frequency discrimination in passerine birds and pigeons. J . Comp. Physiol. Psychol. 94, 401-415. Slivka, R. M., and Bitterman, M. E. 1966. Classical appetitive conditioning in the pigeon: Partial reinforcement. Psychon. Sci. 4, 181-182. Smith, E. N. 1978. Fear-evoked bradycardia of free ranging woodchuck. Physiologist 21. (Abstr.). Smith, E. N., and Worth, D. J . 1980. Atropine effect on fear bradycardia of the eastern cottontail rabbit, Sylvilagusfloridanus. In “A Handbook on Biotelemetry and Radio Tracking” (C. J. Amlaner, Jr. and D. W. Macdonald, eds.), pp. 549-555. Pergamon, Oxford. Smith, E. N . , Allison, R. D., and Crowder, W. E. 1974. Bradycardia in a free ranging American alligator. Copeia 1974, 770-772. Smith, J. C., and Roll, D. L. 1967. Trace conditioning with X-rays as an aversive stimulus. Psychon. Sci. 9, 11-12. Smith, K . 1954. Conditioning as an artifact. Psychol. Rev. 61, 217-225. Smith. 0. A., Jr., and Stebbins. W . C. 1965. Conditioned blood flow and heart rate in monkeys. J . Comp. Physiol. Psychol. 59. 432-436. Smith, S. G . . and Smith, W. M. 1971. A demonstration of auto-shaping with dogs. Psyehol. Rec. 21, 377-379. Soltysik, S. 1960. Studies on the avoidance conditioning: 11. Differentiation and extinction of avoidance responses. Acra Biol. Exp. 20, 171-182. Soltysik, S., and Jaworska, K . 1962. Studies on the aversive classical conditioning. 2. On the reinforcing role of shock in the classical leg flexion conditioning. Acra Biol. Exp. 22, 181-191. Soltysik, S . . and Kowalska, M. 1960. Studies on the avoidance conditioning. 1. Relations between cardiac (type I) and r ,tor (type 11) effects in the avoidance reflex. Acra Biol. Exp. 20, 157-170. Squier, L. H. 1969. Aut .haping key responses with fish. Psychon. Sci. 17, 177-178. Staddon, J. E. R., and SI ,nelhag, V. L. 1971. The “superstition” experiment: A reexamination of its implications for the principles of adaptive behavior. Psychol. Rev. 78, 3-43.

BIOLOGICAL FUNCTION OF PAVLOVIAN CONDITIONING

63

Stebbins, W. C., and Smith, 0. A. 1964. Cardiovascular concomitants of the conditioned emotional response in the monkey. Science 144, 881-882. Stem, J . A., and Word, T. J . 1962. Heart rate changes during avoidance conditioning in the male albino rat. J . Psyrhosom. Res. 6 , 167-175. Subkov, A. A., and Zilov, G. N . 1937. The role of conditioned reflex adaptation in the origin of hyperergic reactions. Bull. Biol. Med. Exp. 4, 294-296. Switzer, S. A. 1930. Backward conditioning of the lid reflex. J. Exp. Psychol. 13, 76-97. Thompson, T., and Sturm. T. 1965. Classical conditioning of aggressive display in Siamese fighting fish. J. Exp. Anal. Behav. 8, 397-403. Timberlake, W., and Grant, D. L. 1975. Auto-shaping in rats to the presentation of another rat predicting food. Science 190, 690-692. Tinbergen, L. 1960. The natural control of insects in pinewoods. I. Factors influencing the intensity of predation by song birds. Arch.N,Perl. Zool. 13, 265-343. Tinbergen, N. 1951. “The Study of Instinct.” Clarendon. Oxford. Tinbergen, N. 1953. “Social Behaviour in Animals. ” Methuen, London. Tinbergen, N . 1963. On aims and methods of ethology. Z. Tierpsychol. 20, 410-433. Vernon, W., and Ulrich, R. 1966. Classical conditioning of pain-elicited aggression. Science 152, 668-669. Wagner, A. R. 1976. Priming in STM: An information processing mechanism for self-generated or retrieval-generated depression in performance. In “Habituation: Perspectives for Child Development, Animal Behavior, and Neurophysiology” (T. J. Tighe and R. N. Leaton, eds.), pp. 95-128. Erlbaum, Hillsdale, New Jersey. Warner, L. H. 1932. Experimental search for the “conditioned response.” J. Genet. Psychol. 41, 91-115. Wasserman, E. A. 1973. Pavlovian conditioning with heat reinforcement produces stimulus-directed pecking in chicks. Science 181, 875-877. Wasserman, E. A. 1974. Responses in Pavlovian conditioning studies. Science 186, 157. Wasserman. E. A,, Franklin, S. R., and Hearst, E. 1974. Pavlovian appetitive contingencies and approach versus withdrawal to conditioned stimuli in pigeons. J. Comp. Physiol. Psychol. 86, 6 16-627. Wasserman, E. A., Hunter, N. B., Gutowski, K. A., and Bader, S. A. 1975. Autoshaping chicks with heat reinforcement: The role of stimulus-reinforcer and response-reinforcer relations. J. Exp. Psychol, Anim. Behav. Processes 1, 158-169. Waxman, H. M., and McCleave, J . D. 1978. Auto-shaping in the archer fish (Toxotes chatareus). Behav. Biol. 22, 541-544. Webb, R. A,, and Obrist, P. A. 1970. The physiological concomitants of reaction time performance as a function of preparatory interval and preparatory interval series. Psychophysiology 6 , 389403. Wells, M. 1968. “Lower Animals.” McGraw-Hill, New York. Wenzel, B. M. 1961. Change in heart rate associated with responses based on positive and negative reinforcement. J . Comp. Physiol. Psychol. 54, 638-644. Werboff, J . , Duane, D., and Cohen, B. D. 1964. Extinction of conditioned avoidance and heart rate responses in rats. J. Psychosom. Res. 8, 29-33. Whittaker, R. H., and Feeny, P. D. 1971. Allelochemics: Chemical interactions between species. Science 171, 757-770. Wilcoxin, H. C. 1977. Long-delay learning of ingestive aversions in quail. I n ‘‘Learning Mechanisms in Food Selection” (L. M. Barker, M. R. Best, and M. Domjan, eds.), pp. 419-453. Baylor Univ. Press, Waco, Texas. Wilcoxin, H. C., Dragoin, W. B., and Kral, P. A. 1971. Illness-induced aversions in rat and quail: Relative salience of visual and gustatory cues. Science 171, 826-828.

64

KAREN L . HOLLIS

Wilkie, D. M., MacLennan, A. J . , and Pinel, J. P. J . 1979. Rat defensive behavior: Burying noxious food. J . Exp. Anal. Behav. 31, 299-306. Williams, D. R., and Williams, H . 1969. Auto-maintenance in the pigeon: Sustained peckingdespite contingent non-reinforcement. J. Exp. Anal. Behav. 12, 5 I 1-520. Williams, G. C. 1966. “Adaptation and Natural Selection: A Critique of Some Current Evolutionary Thought. Princeton Univ. Press, Princeton, New Jersey. Williams, R. T. 1959. “Detoxification Mechanisms,” 2nd. ed. Wiley. New York. Wilson, E. 0. 1975. “Sociobiology: The New Synthesis.” Belknap, Cambridge, Massachusetts. Wolfe, H. M. 1930. Time factors in conditioning finger withdrawal. J . C e n . Psychol. 4, 372-378. Woodard, W. T. and Bitterman, M. E. 1974. Autoshaping in the goldfish. Behav. Res. Methods ftisrrument. 6, 409-410. Woodruff, G., and Williams, D. R. 1976. The associative relation underlying autoshaping in the pigeon. J . Exp. Anal. Behav. 26, 1-13. Wonis, R. P. 1969. The transition from dependent to independent feeding in the young ring dove. Anim. Behav. Monogr. 2 , 1-54. Yehle, A,, Dauth, G., and Schneiderman, N. 1967. Correlates of hear-rate classical conditioning in curarized rabbits. J . Comp. Physiol. Psycho/. 64, 98-104. Zamble, E. 1967. Classical conditioning of excitement anticipatory to food reward. J . Comp. Physiol. Psycho/. 63, 526-529. Zamble, E. 1969. Conditioned motivational patterns in instrumental responding of rats. J . Comp. Physiol. Psychol. 69, 536-543. Zamble, E. 1973. Augmentation of eating following a signal for feeding in rats. Learn. Motiv. 4, 138- 147. Zamble. E.. Baxter, D. J., and Baxter, L. 1980. Influences of conditioned incentive stimuli on water intake. Can. J . Psychol. 34, 82-85. Zeaman. D., Deane, G.. and Wegner, N. A. 1954. Amplitude and latency characteristics of the conditioned heart response. J . Psycho/. 38, 235-250. Zener, K . 1937. The significance of behavior accompanying conditioned salivary secretion for theories of the conditioned response. A m . J . Psychol. 50, 384-403. ”

ADVANCES IN THE STUDY OF BEHAVIOR. VOL. I 2

Selective Costs and Benefits in the Evolution of Learning TIMOTHY D. JOHNSTON RESEARCH BRANCH NORTH CAROLINA DIVISION OF MENTAL HEALTH RALEIGH, NORTH CAROLINA

I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11. Learning and Evolution-Historical Background and Current Concerns 111. An Ecological Conception of Learning . . . . . . .

. ...

IV. Cost-Benefit Analysis and the Evolution of Ad ................. V. The Selective Benefits of Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . , A. Adaptation to Environmental Variability.. B. Sexual Selection ......................................... C. Lack of Variation for Other Adaptive Solutions ........... VI. The Selective Costs of Learning . . . . . . . . . . . . . . . . . , . , . . . . . , . , . . , . A. Delayed Reproductive Effort and/or Success ............... B. Increased Juvenile Vulnerability . . . . . . . . . . . . . . . . . . . . . . . . . . . . , . . . . C. Increased Parental Investment in Each Offs D. Greater Complexity of the Central Nervous System. . . . E. Greater Complexity of the Genome.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . F. Developmental Fallibility . . . . . . . . . . . . . . . . . . . . . . . . . . VII. Learning and the Adaptive Complex.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Interactions of Selective Pressures and Adaptive Traits B. Limits on Adaptive Precision.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VIII. Implications for the Study of Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

I.

65 66 69 70 73 74 76 77 79 79 83 84 86 89 91 92 93

94 96 98

INTRODUCTION

To a large extent, the study of animal learning owes its existence to the theory of evolution, for it was the concern of Darwin (1871, 1872), Romanes (1884), and C. L. Morgan (1896) with demonstrating intellectual and, hence, evolutionary continuity among animal species that led to the rapid rise of comparative psychology after 1900 (see Gottlieb, 1979). The comparative study of learning received from its founders an emphasis on uncovering similarities in learning ability among different species, and has retained that focus in large measure for 65

Cupynght 0 1982 by Academic Press. Inc All rights of reproduction in any fuan reserved ISBN 0-12-004512-5

66

TIMOTHY D. JOHNSTON

almost 100 years. While that emphasis may be appropriate enough for an emerging comparative science, it should be noted that the theory of evolution, which provides the theoretical justification for comparative study, is equally receptive to an alternative emphasis; namely, that different animals possess adaptive specializations that equip them for survival in the particular environments in which they live. These two issues, continuity of descent and specialization of adaptation, are familiar sides of the evolutionary coin. The psychology of learning has, throughout its history, tended to emphasize the former to the virtual exclusion of the latter. In this article I wish to consider one issue that arises when learning is considered from the viewpoint of evolutionary adaptation: the problem of the selective costs and benefits attendant on its evolution. Looked at from a teleological point of view, learning is manifestly a very useful ability. Almost any individual animal would appear to gain some advantage from the possession of more varied or more efficient learning abilities. However, from an evolutionary point of view, what is important is not “usefulness” in the colloquial sense, but “benefit” in the technical, reproductive sense (Williams, 1966; see Section IV). I shall argue that, although learning might well be considered quite generally useful, it does not confer a universal selective benefit. Indeed, it will be seen that learning incurs a number of important selective costs that tend to oppose its evolution. An appreciation of the relative costs and benefits of learning is essential to a proper perspective on the problem of its evolution.

A N D EVOLUTION-HISTORICAL BACKGROUND AND 11. LEARNING CURRENT CONCERNS

The comparative study of learning has drawn repeatedly on evolutionary thinking throughout its history, albeit with different emphases and different degrees of success. At first, the primary concern was to establish the fact of mental evolution by demonstrating intellectual continuity among animal species, including humans. The largely anecdotal accounts compiled by Romanes ( 1884) were too uncritically accepted by him as evidence of feats of learning and insight rivaling those of humans, but they provided the basis for Lloyd Morgan’s (1896) more critical assessment and paved the way for later experimental studies of learning. Thorndike (191 1, p. 22) declared that the main purpose of the experimental study of animal learning was “to learn the development of mental life down through the phylum.” By that he meant that its aim should be to establish continuities in learning mechanisms among species, showing how the capabilities of more advanced forms might have been derived in evolution from those of more primitive animals. Thorndike realized that controlled experimentation offered the only hope of any real insight into the mechanisms of learning, and his work established the field of animal learning as a major branch of experimental psychology.

LEARNING AND NATURAL SELECTION

67

The elucidation of phylogenetic relationships among different forms of learning has been a major concern for comparative psychology, as represented by modem workers such as Harlow (1958), Warren (1965), and Bitterman (1965, 1975). This brand of comparative study has been criticized for evolutionary naivete (Hodos and Campbell, 1969; Capitanio and Leger, 1979), and it is less prominent now than it once was. Although phylogenetic questions of continuity and discontinuity among learning mechanisms have certainly been preeminent in the comparative study of learning, in the early history of the field there was at least passing interest in the question of the adaptive significance of learning. That question was the primary motivation for the functionalist school of psychology (James, 1890; Angell, 1907; see Boring, 1950; Cravens and Burnham, 1971), which, drawing on the pragmatic philosophy of Charles Peirce and William James, attempted to relate the operations of the mind (such as learning) to the everyday requirements of the environment. The major learning theorist most clearly influenced by functionalism was Clark Hull, whose early writings include analyses of the adaptive significance of conditioning mechanisms (Hull, 1929, 1937). As late as 1943, Hull wrote, “Since the publication by Charles Darwin of the Origin of Species, it has been necessary to think of organisms against a background of organic evolution and to consider both organismic structure and function in terms of survival” (Hull, 1943, p. 17). Despite his early evolutionary interests, the main body of Hull’s theory of learning lacks any evolutionary content and the same is true of other learning theorists who have, on occasion, expressed their sympathy with evolutionary thinking (see, e.g., J. B. Watson, 1924; Skinner, 1974). The emphasis on phylogenetic continuity among the learning abilities of different species was bolstered by the widespread acceptance of a clear-cut dichotomy between learning and instinct, particularly during the early development of psychological learning theory (see Marquis, 1930; McGraw, 1946; Anastasi, 1954). Instinct was viewed as an endogenous provision of “nature” (evolution) that permits an animal to deal with the particular set of problems posed by its environment. Learning, on the other hand, was a process whereby “nurture” (experience) could supplement the provisions of instinct, permitting an animal to deal with unusual or unexpected circumstances by acquiring the necessary behavioral skills. Despite the cautions of some contemporary writers (e.g., Kuo, 1922; Carmichael, 1925; Gesell, 1933; McGraw, 1946), the tacit acceptance of a dichotomy between learning and instinct had profound effects, both methodological and conceptual, on the subsequent development of psychological learning theory. Methodologically, it sanctioned and even demanded the use of biologically arbitrary tasks for the study of learning, for only in that way could the psychologist study “the association process, free from the helping hand of instinct” (Thorndike, 1911, p. 30). The most cursory glance at the methodological approaches currently used for the study of animal learning will reveal the con-

68

TIMOTHY D. JOHNSTON

tinuing effects of that bias. Conceptually, the learninghstinct dichotomy cleared the way for the acceptance of learning as a general process (Seligman, 1970), free from any species-specific “contaminations. ” Since learning, in this view, is a “supraspecific ” characteristic, the adaptive demands placed on particular species by their environments cannot be relevant to its understanding. Comparative study soon revealed that species differ in their performance on learning tasks. Such findings prompted a distinction between performance on a task and the learning processes that underlie performance. That distinction allowed the idea of a general process to be retained, by arguing that species differences reflect differences in performance variables such as perceptual and motor capabilities. Species-specific adaptations related to performance are thus overlaid on a general learning process, but are not the primary concern of learning theory (see LoLordo, 1979). Variations in the learning process itself have also been admitted, but were limited to quantitative differences in rates of conditioning, and to the presence or absence of a limited number of kinds of learning, such as habituation, operant conditioning, or discrimination learning (Tolman, 1949; Thorpe, 1963; Bitterman, 1965, 1975; Lorenz, 1969; Razran, 1971). In the latter case, each kind of learning has been seen as a general process in that its characteristics transcend species boundaries. The psychology of learning thus stands in a curiously ambiguous relation to evolutionary theory. Although, as a comparative science, it draws its main justification from the theory of evolution, many important evolutionary implications, in particular those regarding adaptive specialization, have been overlooked. Ethologists have frequently taken psychology to task for this omission (e.g., N . Tinbergen, 1951; Lorenz, 1965, 1969), but, since ethology has not shown the same broad concern with problems of learning, it has not produced theories of learning that might challenge those proposed by psychologists. As a result, its criticisms have been somewhat blunted. In recent years there has been a number of criticisms of the study of animal learning from within the field itself, criticisms that indicate a resurgence of functionalism in psychology (Shettleworth, in press; see also Petrinovich, 1979). Those criticisms originated in the findings of Garcia and Koelling (1966) that rats will readily learn to avoid a sweet-tasting solution if its ingestion is paired with toxicosis but not if ingestion is paired with foot-shock. Subsequent experiments demonstrated that such learning occurs even when the delay between ingestion and toxicosis is as long as 2 hr (Garcia et d.,1966). These findings appear to conflict with certain central assumptions of general process learning theory (Seligman, 1970) and the implications of this conflict have been explored in detail by a number of theorists (Rozin and Kalat, 1971; Revusky, 1971, 1977; Shettleworth, 1972, in press; Hinde, 1973; Logue, 1979; LoLordo, 1979; Johnston, 1981). The similar implications of data from other areas of animal learning have been discussed by Bolles (1970, 1971), Seligman and Hager (1972), and Hinde and Stevenson-Hinde (1973).

LEARNING AND NATURAL SELECTION

69

One implication that has been drawn from the research just cited is that there are “biological constraints” on an animal’s learning abilities (cf. N . Tinbergen, 1951, p. 145), constraints that are drawn in large part by the evolutionary history of the species concerned. It has been proposed, notably by Rozin and Kalat (1971) and Hinde (19731, that a proper understanding of an animal’s learning abilities requires that they be analyzed in terms of the animal’s overall biological adaptation to its environment. There are two related senses in which we might seek to understand learning in relation to the concept of adaptation: first, as a producr of evolutionary adaptation, that is, as the outcome of a history of selection pressures acting on the gene pool of the population to which the individual belongs; second, as a process of ontogenetic adaptation by which the individual adjusts to certain characteristics of the environment over the course of its own lifetime. It is in regard to learning as a product of evolution that I shall discuss the question of its selective costs and benefits. Elsewhere (Johnston and Turvey, 1980; Johnston, 198 I , in press) a theoretical approach to the study of learning as a process of ontogenetic adaptation has been disucssed in some detail.

111. A N ECOLOGICAL CONCEPTION OF LEARNING

If we are to be able to discuss the selective costs and benefits of learning, it is important to begin with an understanding of what phenomena are to count as instances of learning. Providing a detailed account of a conception of learning that is an adequate basis for evolutionary analysis would be a substantial theoretical endeavour in its own right and I do not propose to attempt it within the scope of this article (see Johnston and Turvey, 1980). It is clear, however, that the conception must be an ecological one; that is, it must be given in terms of the relationship between the animal and its environment, rather than in terms of the animal alone. The selective history of a population is a history of interactions between an evolving gene pool and a particular changing environment, and if learning is to be understood in the context of that history it must be understood in terms of both the animal and its environment. The requirement for an ecological conception of learning means that traditional conceptions of learning as conditioning (e.g., Kimble, 1967), which have been articulated within the nonecological, general process tradition, are likely to be of very limited usefulness. The general process tradition (Seligman, 1970) is one in which demonstrating the relevance of an experimental learning task to the ecological demands that an animal normally faces in its environment has been of no concern whatsoever. It is entirely possible that the kinds of learning abilities revealed by those tasks (and embodied in the conceptions of learning to which they have given rise) are never employed by the animal under normal circumstances. If this is so, then selection has never acted on them and any inquiry into their selective history is largely meaningless.

70

TIMOTHY D. JOHNSTON

Defining learning in terms of conditioning (Bolles, 1975; Rescorla and Holland, 1976) is therefore likely to be quite inappropriate and, in particular, too restrictive for the purposes of this article. Some forms of conditioning may well be involved in an animal’s ontogenetic adaptation to its environment, but there are other relevant developmental phenomena that are excluded by definitions of conditioning (see Gottlieb, 1976a; Johnston and Turvey, 1980). For my present purposes, I shall use the term “learning” to refer to any process in which, during normal, species-typical ontogeny , the organization of an animal’s behavior is in part determined by some specific prior experience. In the absence of the requisite experience, either some behavioral ability will be altogether lacking, or its organization will be different from that of similar individuals for whom the experience was available. The requirement that the requisite experience be specific is designed to exclude such phenomena as the effects of inadequate nutrition on behavioral development (Leathwood, 1978). Such effects are highly nonspecific and I do not see that anything is gained by admitting them as instances of learning. Specificity is, of course, a matter of degree (Bateson, 1976), so that the preceding definition permits the identification of developmental phenomena as more or less typical instances of learning, rather than as either learning or not learning (see Johnston and Turvey, 1980, and Johnston, in press, for arguments in support of such a conception of learning).

Iv. COST-BENEFIT ANALYSIS A N D THE EVOLUTION OF ADAPTATIONS The purpose of this section is to review, very briefly, some of the evolutionary theory that motivates the analysis of learning in terms of selective costs and benefits. Nothing in this account is either original or controversial, but it may help to avoid misinterpretation of subsequent arguments if their premises are stated explicitly here. The aim of a cost-benefit analysis is to allow us to make statements about the conditions under which particular kinds of adaptations are likely to evolve and, hence, explain why these adaptations are observed in some populations and not in others. Such statements may be either qualitative or quantitative and they may account to a greater or lesser degree for the details of an adaptive trait. The present analysis is limited to qualitative considerations and will not attempt to account for any of the details of particular learning abilities. My aim will be to elucidate the selective costs and benefits associated with the evolution of any developmental system in which the organization of a behavioral characteristic is in part determined by prior experience. The term “evolution” in the above formulation refers to a change in the population-typical phenotype over some period of time, and I shall not attempt to

LEARNING AND NATURAL SELECTION

71

provide an account in genetic terms. The genetic basis of any learning ability is likely to be very complex, since learning is a function of entire epigenetic systems, not of single genes. Selection for the evolution of a learning ability cannot be between competing alleles within a population (the usual model of population genetics) but must be between competing genetic systems, either whole genotypes or parts of genotypes.’ The study of such evolutionary problems is at present in a very early stage (DeBenedictis, 1978; Lloyd, 1977) and the application of rigorous genetic analyses to the evolution of learning is likely to remain a very difficult task for some time to come (see Frazzetta, 1975, for further comments on the problems encountered in analyzing the evolution of developmental adaptations). Phenotypic models of evolution assume that individuals in the population vary with regard to the phenotypic trait of interest, in this case in the manner in which development of a behavioral ability is influenced by prior experience, and that there is a correlation between phenotypic and genetic variation in this regard. If individuals with a high degree of developmental sensitivity to certain prior experiences have greater reproductive success than those with a lower degree of sensitivity, then this phenotypic trait is said to confer a net selective benefit on its possessor. If this state of affairs persists, then possession of the trait will gradually come to be typical of individuals in the population. If the reverse is true, then the trait incurs a net selective cost and the possession of a low degree of developmental sensitivity will become, or remain, typical. The possession of any phenotypic trait may be expected to have a number of consequences, each of which, taken separately, may be associated with a certain selective cost or benefit. Only when the total benefits outweigh the total costs is there a net benefit to the individual. By considering various costs and benefits of learning in isolation, a cost-benefit analysis can indicate the situations under which learning may carry a net selective benefit and so is expected to evolve. The discussion of individual selective costs and benefits in this article will assume that “other things are equal. Consider the proposition that a given selective cost is incurred by an individual who has a greater dependence on learning in the development of some behavioral skill than do other individuals in the population. That does not mean that the learning ability confers no selective advantage. Rather, it means that if all individuals develop equivalent behavioral skills, regardless of the extent to which those skills are learned, then the cost in question will oppose the evolution of that learning ability. Obviously, learning must cany a net selective benefit under some conditions, for many animals do in fact learn. However, in order to understand what those conditions are, it is necessary first to consider the various costs and benefits in isolation, “other things being equal.” ”

‘This is not to say that single allele substitutions do not affect learning; there is abundant evidence that they may do so (e.g., Aceves-Pifiaand Quinn, 1979; Dudai, 1979). However, models thatassume selection between the hypothetical alleles learning and no-learning are unlikely to provide an adequate account of the evolution of learning.

72

TIMOTHY D. JOHNSTON

Some of the complex interactions among individual costs and benefits that are certain to occur as learning evolves are discussed later in the article (Section VII ,A). It is important to emphasize that the costs and benefits associated with learning are to be understood in purely reproductive terms. Although a particular learning ability may enable an animal to perform certain tasks more efficiently or skillfully, and to that extent may be seen as beneficial to the individual, a change in the population-typical phenotype can only occur if there is also a reproductive benefit (Darwin, 1859; Williams, 1966; Ghiselin, 1974, Lewontin, 1974, 1978). There are a number of problems of interpretation involved in the analysis of the selective costs and benefits of learning. Some of these are common to all attempts to infer the selective history of an adaptive trait, since direct evidence is usually unavailable and we must rely on indirect evidence from present-day populations. We therefore search among extant species for correlations between the existence of learning abilities, for example, and the existence of putative selective costs and benefits. Assuming that the same principles of natural selection hold today as in the past, we may then infer that these costs and benefits for which the expected correlations hold were involved in the evolution of learning. This approach has some well-known drawbacks and limitations (see J. L. Brown, 1975; Hinde, 1975), but it is usually the only course open to us. In some cases it may be possible to make direct tests of hypotheses concerning the cost of benefit of a trait, by making experimental alterations and observing their reproductive effects (e.g., N. Tinbergen eral.. 1963). Unfortunately, where learning is the trait in question, we cannot make the necessary modifications to the phenotype. However, the method of adaptive correlation has proved very useful in the study of behavioral evolution (see Brown, 1975; Alcock, 1979), not least as a means of stimulating the investigation of questions whose importance was not previously recognized, and I make no apology for employing it here. There is a second problem, peculiar to the evolutionary study of learning; namely, that only a few studies of learning have investigated the development of ecologically relevant behavioral skills. Thus, the data we possess on the learning abilities of different species frequently do not allow us to make any statements about the role that learning may play in the species-typical development of behavior under normal circumstances. Such data are of very limited relevance to any evolutionary inquiry: A learning ability that is never manifest under natural conditions cannot reveal anything about the selection pressures that may be involved in the evolution of learning, since it cannot affect the reproductive success of individuals that possess it. While it is possible that such “laboratory” learning abilities may once have been of adaptive significance to their possessors, the complete lack of information about the nature of past environments in which this may have been the case prevents us from using such abilities as the basis for evolutionary analysis. The question of why those abilities should have evolved at all will be considered later in this article (Section VI1.B).

LEARNING AND NATURAL SELECTION

73

The evidence that we do possess about ecologically relevant learning abilities is therefore most often indirect. Where it is more direct, complementary data on possible selective pressures in extant populations are usually lacking. I hope that by drawing attention to these lacunae in our knowledge, and by pointing out their importance to the evolutionary study of learning, this article can help to stimulate further study of ecologically relevant learning abilities and, thus, permit the questions that are raised here to be posed in a more rigorous manner, and eventually answered.

v.

THE SELECTIVE BENEFITS OF LEARNING

It might seem that the selective benefits of learning are so obvious as hardly to merit extended discussion. Surely an organism that can learn can obtain far more information about its environment, and, hence, can adapt more successfully to it, than one that is unable to learn. On this view, selection ought always to reinforce and extend current learning abilities and to favor the evolution of new ones. The idea that evolutionary insights can be obtained by ordering organisms on a scale of learning ability (e.g., Bitterman, 1965, 1975; Yarczower and Hazlett, 1977; see Hodos and Campbell, 1969) appears to reflect allegiance to some such view. In a recent paper, Mayr provides a clear statement of this point of view: The great selective advantage of a capacity for learning is, of course, that it permits storing far more experiences, far more detailed information about the environment, than can be transmitted in the DNA of the fertilized zygote. Considering this great advantage of learning, it is rather curious in how relatively few phyletic lines genetically fixed behavior patterns have been replaced by the capacity for the storage of individual acquired information (Mayr, 1974, p. 652).

This is a curious position for an evolutionary theorist to adopt, because it draws no distinction between the colloquial and technical senses of the “advantage” of a characteristic to its possessor. The potential usefulness of a learning ability to some organism is not the same thing as its selective benefit and it has nothing to do with the likelihood that the ability will evolve (Williams, 1966). In addition, Mayr’s formulation implies that the selective advantage of a learning ability is independent of the environment in which evolution is occurring. As many authors have pointed out, adaptation specifies a relation that may or may not exist between an organism and some environment (Sommerhoff, 1950; Williams, 1966; Ghiselin, 1966, 1974; Slobodkin and Rapoport, 1974). Adaptation is not an environment-neutral property of either an organism or a particular characteristic of an organism, such as its learning abilities. The adaptive significance and selective benefits of learning can only be assessed with respect to particular sets of environmental factors, not in an ecological vacuum.

74

A. 1.

TIMOTHY D. JOHNSTON

ADAPTATION TO ENVIRONMENTAL VARIABILITY Nature of the Benefit

The environment of an organism may be characterized by a list of ecological factors (including abiotic, biotic, and social ones) that affect its well-being and survival (Mason and Langenheim, 1957). Insofar as relatively invariant (i.e., constant or only slowly changing) ecological factors are important to the relationship between an organism and its environment, adaptation to them may be effected by natural selection, regardless of the existence of developmental sensitivity on the part of individuals within the population. Such factors produce sustained selection pressures that constrain the range of genetic variation to just those genotypes that give rise to well-adapted phenotypes (Lewontin, 1974). Other factors that are more variable may still permit natural selection to effect the necessary adaptation if some constant pattern of variation is manifest within individual life spans. Light intensity, for example, varies on both a diurnal and an annual cycle, but since the patterns of variation remain invariant over long periods of time, natural selection can effect adaptation to those patterns in animals with life spans of more than 1 year. Many organisms interact with their environments in ways that require them to adapt to ecological factors that do not exhibit such invariant properties. The nature and distribution of food sources, location of a burrow or nest, routes of travel, individual identities of parents or mate, and social relationships with conspecific neighbors are all ecological factors that may vary importantly between the life spans of successive generations and/or within that of a single generation. Clearly, natural selection cannot effect adaptation to such factors in the absence of some degree of developmental sensitivity to the environment, since the direction of selection varies stochastically from one generation to the next and there can be no net adaptive change in the gene pool of the population. Under such circumstances, a selective benefit will accrue to those individuals that possess the requisite developmental sensitivity. The ability to learn, as one form of developmental sensitivity to the environment, has as its primary selective benefit that it permits adaptation to ecological factors that vary over periods of time that are short in comparison with the lifetime of an individual (G. Bateson, 1963; Williams, 1966; Slobodkin, 1968; Slobodkin and Rapoport, 1974; Plotkin and Odling-Smee, 1979; Johnston and Turvey, 1980). 2 . Available Evidence Because of the paucity of ecologically motivated studies of learning, and the extreme difficulty of testing adaptive hypotheses about learning skills, evidence that adaptation to environmental variability confers a selective advantage on individuals able to learn is hard to come by. Generally speaking, the most compelling evidence would be to find closely related species, or populations of

LEARNING A N D NATURAL SELECTION

75

the same species, that exhibit differences in learning ability clearly related to differences in environmental variability. For example, Sasvhri (1979) demonstrated that great tits (Parus major) are more adept at learning feeding techniques by observation than either blue tits ( P . cueruleus) or marsh tits ( P . palustris). The great tit occupies a wider range of habitats than either of the other two species (Perrins, 1979), and so individuals may be expected to encounter different kinds of prey, depending on the location in which they are reared. If different feeding techniques are required for efficient exploitation of different prey types, as suggested by the work of Kear (1962) and Partridge (1976), then observational learning may permit great tits to exploit the wider range of prey that they typically encounter. Gray ( 1979, 1981; Gray and Tardif, 1979) has studied the feeding diversity of several species of deermice (Peromyscus spp.). The results from two species, P . maniculatus sonoriensis and P . leucopus, are of special interest. P . m. sonoriensis inhabits southwestern desert environments and P . leucopus inhabits more northerly woodlands. Gray and Tardif (1979, Experiment 2) found that laboratory-reared P . leucopus had lower adult feeding diversity than wild-caught adults, whereas laboratory rearing had no effect on adult feeding diversity in P. m. sonoriensis. These results imply that experience may affect the development of feeding diversity to a greater extent in P . leucopus than in P . m . sonoriensis, an implication that was borne out by further experiments (Gray and Tardif, 1979, Experiment 3). Gray (1981) suggests that the different susceptibilities of the two species to the effects of laboratory rearing is related to the fact that P . leucopus inhabits an environment that may be more variable than that inhabited by P . m . sonoriensis, so that selection has favored the evolution of developmental sensitivity in the former species but not in the latter. More detailed study of the two species' environments is needed, however, before this hypothesis can be properly evaluated. Emlen (1969, 1970) has shown that early experience is important for the development of stellar navigation skills in the indigo bunting (Passerina c y m e s ) , a nocturnal migrant. Birds that are exposed to the night sky prior to their first migration season subsequently select an appropriate migratory orientation by reference to star patterns, whereas birds lacking this experience do not. The precession of the earth's axis changes the relation between stellar and geographical directions at the rate of about 3" every 1000 years. Emlen (1975) suggests that this rate of change may be too rapid to effect adaptive changes in the gene pool of the population, and that individual learning has been selected to permit adaptation to this aspect of the environment. 3.

Evaluation

Only as more investigators turn to the study of learning from an ecological perspective will it become possible to accumulate data to address the hypothesis

76

TIMOTHY D. JOHNSTON

that learning confers the selective benefit of adaptation to environmental variation. Although it is undeniably true that learning can in principle confer such a benefit, further data are needed before it can be concluded that it does do so in evolving populations. Studies such as Gray’s, which relate differences in learning to differences in environmental variation among related species, are likely to offer the most valuable insights in this regard. One way in which the search for relevant data might be stimulated is by the development of more refined theoretical models that predict the evolution of particular kinds of learning in particular environments. Treatment of this issue lies outside the scope of the present article, but the reader is referred to the work of Estabrook and Jesperson (1974). Bobisud and Potratz (1976), and Arnold (1978) for examples of such models. The ability to learn is likely to be of greatest selective advantage in resourcelimited populations in which individuals must compete for food, nest sites, and other resources. In such relatively K-selected species (MacArthur and Wilson, I967), the most successful individuals tend to be those that utilize the available resources most efficiently. Learning is likely to contribute substantially to such efficiency when the nature and distribution of limited resources is variable. Under those conditions, an important selective benefit will be realized by individuals that possess the requisite learning skills. In more r-selected species (MacArthur and Wilson, 1967), by contrast, reproductive output rather than efficiency of resource exploitation is the primary determinant of reproductive success, and learning is unlikely to be of great selective advantage in such species (see also Section VI,C and D). B.

SEXUAL SELECTION

1.

Nature of the Benefit

Sexual selection has been proposed by Nottebohm (1972) as the agent responsible for the evolution of vocal learning in many species of songbirds. According to this argument, females select males with the most elaborate vocal repertoires; therefore, developmental mechanisms that increase the number of song types in the repertoire will tend to evolve. Proponents of the argument claim that the genome is limited in the amount of information it can encode for elaborate species-typical song, and that sexual selection favors the evolution of learning as a mechanism for increasing the vocal repertoire. 2.

Available Evidence

Nottebohm presented no evidence in support of his hypothesis and, so far as I know, none has yet been forthcoming. J . L. Brown (1975) criticized Nottebohm’s suggestion on the grounds that large repertoires occur primarily in

LEARNING A N D NATURAL SELECTION

77

monogamous species, rather than in polygamous ones in which sexual selection would presumably be stronger. He suggests that a more important function of vocal learning may be to maintain pair bonds. There is some evidence that selection may favor large repertoires, but not on the basis of female choice. Kroodsma ( 1 976) found that female canaries (Serinus canarius) exposed to recordings of large repertoires (35 syllable types) built nests faster and laid more eggs than did females exposed to small repertoires (5 syllable types). Whether or not this is an example of sexual selection may be open to question.

3 . Evaluation Even if repertoire size does influence female choice, we would only expect selection for large repertoires to result in the evolution of learning if there are limits on the capacity of the genome to encode the relevant information. The finding that male song sparrows (Melospiza rnelodia) develop larger repertoires when given the opportunity to learn from adult song models than when reared in isolation (Kroodsma, 1977) lends some indirect support to that suggestion, but arguments based on the information encoding capability of the genome tread on very insecure ground. Our present state of knowledge about the way in which genetic information is translated into the structure of behavior in the course of development is almost nonexistent, and in Section VI,E,I shall present some arguments that learning may actually involve more, rather than less, genetic information. At present, the issue probably cannot be decided. There are in any case only rather few kinds of learning skills that might evolve by sexual selection, namely, those that involve the development of courtship behavior.

c.

LACKOF

VARIATION FOR OTHER

ADAPTIVE SOLUTIONS

1 . Nature of the Benefit

Natural selection is an opportunistic process-it must work with whatever variation is present in the population and can only effect adaptation to the extent that appropriate variation is available (see further Section VII1,B). Even where a style of development that is wholly or largely independent of experience could well effect adaptation to some aspect of the environment, the ability to learn the requisite skills may still evolve, if variation in regard to learning is the only appropriate kind of variability available in the population. In such a situation, the learning ability confers a selective benefit on its possessor, even though learning is not more adaptive than an alternative style of development (Lewontin, 1978, 1979; Gould and Lewontin, 1979). Styles of development that are either more or less dependent on experience might occupy equivalent adaptive peaks (Wright, 1932) and the style that evolves would then depend upon the available variation in the population.

78

TIMOTHY D. JOHNSTON

2 . Available Evidence Direct evidence in favor of this selective benefit is unattainable, since it depends upon knowing the nature of phenotypic and genetic variation in extinct populations and such information is forever lost to us. Indirect evidence may be obtained by comparing species in which similar kinds of behavioral skills develop with greatly differing degrees of dependence on the nature of experience. If the skill in each species effect adaptation to equally invariant aspects of the environment, then we may argue that the different styles of development represent equivalent adaptive peaks. Consider the contrasts that are found among a number of species of songbirds, in which normal song development shows quite different degress of dependence on auditory experience (reviews by Marler and Mundinger, 1971; Nottebohm, 1970; Marler, 1975). For example, song sparrows (Melospiza melodia) reared without exposure to conspecific adult song models develop song that is indistinguishable from, although less variable than, that of normally raised birds (Mulligan, 1966; Kroodsma, 1977). In another species, the white-crowned sparrow (Zonotrichia leucophrys; Marler, 1970), exposure to adult song is an absolute requirement of normal song development. Both species develop elaborate song repertoires under normal circumstances and it is not clear what, if any, adaptive significance is to be assigned to the developmental differences between them (Marler, 1967; Nottebohm, 1972). A second line of indirect evidence comes from studies of the developmental effects of self-stimulation. To give but one example, Gottlieb (1978) has shown that surgically devocalizing Peking ducklings (Anas plutyrhynchos) just prior to hatching prevents the development of a highly specific approach response to the maternal assembly call, which is shown by normal (embryonically vocal) ducklings after hatching. Other aspects of the prenatal auditory environment, such as exposure to maternal calls, are not required for normal development of this behavior (Gottlieb, 1971). The self-produced auditory environment of the embryo is quite invariant between generations and, indeed, artificially varying that environment, by playing altered recordings of the embryonic call to devocalized embryos, does not produce a corresponding change in the postnatal approach preference (Gottlieb, 1980). It is difficult to see what the adaptive significance of an experientially dependent style of development might be in this situation, since complete independence from the nature of individual experience would seem able, in principle, to produce the same behavioral end-point. 3 . Evaluation

Since the preceding lines of evidence rely on the lack of any apparent adaptive interpretation for experience-dependent development in the examples discussed, they are obviously somewhat unsatisfactory. The lack of adaptive interpretations

LEARNING A N D NATURAL SELECTION

79

may simply reflect our incomplete understanding of the ecological context surrounding these instances of development and such explanations may be forthcoming in the future. That learning may confer a selective advantage when other adaptive solutions are unavailable cannot be seriously doubted in principle, but demonstrating the involvement of this benefit in particular instances of the evolution of learning is likely to remain extremely problematic. One value of pointing out the probable existence of this selective benefit is to stress that adaptive explanations for the possession of learning abilities may not always be available (Lewontin, 1979; Could and Lewontin, 1979; see Section VII1,B).

VI. THESELECTIVE COSTSOF LEARNING The selective benefits of learning have been discussed before, by a number of authors, but the selective costs involved in the evolution of learning have received much less attention. No complete understanding of learning as a product of evolutionary adaptation can be achieved until we gain an appreciation of those costs and the extent to which they are incurred in evolving populations. In this section I shall identify six potential selective costs of learning and review evidence that these costs are in fact incurred in the course of evolution. A.

DELAYED REPRODUCTIVE EFFORTAND/OR SUCCESS

1. Nature of the Cost

In general, the earlier an organism can begin reproduction, the more offspring it can produce at one time, and the more frequently it can reproduce, the greater will be its selective advantage. However, any reproductive effort is a heavy drain on an organism’s resources of time and energy (Ricklefs, 1974) and so there will tend to be selection against investment in reproduction at those times during an individual’s life when such investment would unduly compromise the survival of either the parent@) or offspring. As the development of those behavioral skills that are involved in individual survival and parental care becomes more dependent on individual experience, there is a period in early life during which the developing organism’s behavioral competence is relatively limited in certain respects? namely, in those that involve the skills being learned. During this time, diversion of resources into reproductive effort is likely to be a poor investment, since the *I do not wish to imply that the young animal is merely an inadequate adult; there are certainly many characteristicsof juvenile behavior and morphology that are adaptations to the specific ecological niche occupied by the young animal (see Oppenheim, 1980, 1981; Galef, 1981). My attention in this discussion is limited to those skills involved in reproductive activities, rather than to the entire behavioral repertoire.

80

TIMOTHY D. JOHNSTON

individual is in a particularly precarious adaptive position. Its lack of competence not only compromises its own survival as a result of reproductive effort but also makes it less likely that adequate preparation can be made for the birth and early development of its offspring. There are two selective consequences of this situation. If an organism’s dependence on learning is high, the number of offspring that it can expect to rear during this period of relative behavioral incompetence will be relatively low. In comparison with other individuals in the population whose behavioral skills develop in a manner that, while equally adaptive, is less dependent on learning, it is therefore at a selective disadvantage. In addition, there may be selection in favor of delaying the age of first reproduction until full behavioral competence is reached. This is likely to occur if early reproductive effort places such a strain on the individual’s resources that either its own survival is threatened or the success of subsequent reproduction is compromised. 2.

Available Evidence

Evidence that the selective cost of delayed reproduction is incurred by the evolution of learning requires two types of studies: demonstrations of reduced breeding success in young animals in comparison with conspecific adults; and demonstrations of a role for experience in the development of certain crucial behavioral skills. a . Juvenile Breeding Success. Lack (1954) proposed that the delayed reproduction found in many animals, especially birds and mammals, might be an adaptation to the difficulties to be expected by an unskilled juvenile in supporting both itself and its offspring. Later (Lack, 1966) he supported this argument with data from a number of studies that demonstrate lower breeding success in young birds of several species. Particularly complete data concerning the relationship between age and breeding success are available for the great tit (Parus major; Perrins, 1965), the kittiwake (Rissa triductyla; Coulson and White, 1961; Wooler and Coulson, 1977), the yellow-eyed penguin (Megadyres antipodes; Richdale, 1957), and the European blackbird (htrdus merula; Snow, 1958). In these species, clutch size, hatching rate, and fledging rate are all lower for birds breeding in their first year of life than for those breeding in their second or subsequent years. In the penguin and kittiwake this juvenile disadvantage extends into the second and perhaps third years. b. Development of Adult Behavioral Competence. Among the behavioral skills that would be expected to contribute most directly to an individual’s ability to raise offspring are foraging and nest-site selection. A number of studies have shown that the foraging efficiency of young birds is lower than that of adults, but, unfortunately, relevant data are not available for the species mentioned above, whose breeding biology has been studied particularly thoroughly. However, it has been noted, e.g., by Ashmole (1963), that breeding tends to be

LEARNING A N D NATURAL SELECTION

81

especially long delayed (up to 5 or 6 years in some cases) in species of predatory birds that appear to require a considerable degree of skill in catching their prey. Marked disparities between juvenile and adult foraging efficiencies have been reported in several such species, including the little blue heron (Florida caerulu; Recher and Recher, 1969), the royal tern (Thalusseus muximus; Buckley and Buckley , 1974), the sharp-shinned hawk (Accipiter striatus; Mueller and Berger, 1970), the herring gull (Larus argentatus; Verbeek, 1977), the sandwich tern (Sternu sandivicensis; Dunn, 1972), and the brown pelican (Pelecanus occidentalis; Orians, 1969). Bene (1945) observed that young black-chinned hummingbirds (Archilochus alexandri) often attempt to feed from leaves and twigs as well as flowers but later concentrate their feeding attempts more appropriately. Although he made no direct tests of the involvement of learning in this increase in foraging efficiency, he did find that adults could learn to adjust their foraging behavior in response to changes in the distribution and concentration of nectar sources, suggesting that learning may also be involved in the earlier development of efficient foraging. In the case of the brown pelican, Blus and Keahey (1978) have found that immature birds ( 1 -3 years old) lay smaller clutches, hatch fewer eggs per clutch, and fledge fewer young than do adult birds (> 3 years old). This study also found that immature birds construct more nests in low-lying areas and, as a result, lose many more clutches to flooding than do adult birds. Nest placement does not seem to be a result of competition for nest sites, since there is frequently unoccupied high ground near a flooded nest. Blus and Keahey suggest that experience may play a role in nest-site selection in this species. No data are available on the breeding biology of the other species whose age-related foraging efficiency has been studied, but in the arctic tern (S. paradisaea), a close relative to the sandwich tern, Coulson and Horobin (1976) found that hatching and fledging success increase between 3 and 8 years of age and that most birds breed for the first time at 3 or 4 years.

3 . Evaluation These data on foraging efficiency and breeding success suggest that, especially where foraging involves the capture of mobile, elusive prey, young birds may be unable to catch enough food to breed as successfully as older birds. The question arises whether their lower foraging efficiency reflects the learning of this skill or is due to some other factor. So far as I know, in none of the species mentioned has the development of foraging been studied to elucidate the role of experience. It might be, as Groves (1978) has suggested, that the difference between adults and juveniles simply reflects the death of the least successful juveniles. This is an unconvincing explanation, because juvenile mortality is certainly high enough in almost all species of birds (see Lack, 1966) to provide strong selection pressure for high foraging efficiency early in life. That such efficiency has not evolved in

82

TIMOTHY D. JOHNSTON

many species suggests quite strongly that selection among alternative behavioral phenotypes is not sufficient to ensure high efficiency and that individual experience is required for its development. An alternative explanation for these data is that adult efficiency must await the maturation of some physiological system independently of experience, such as the growth of a sufficiently powerful musculature to permit certain necessary manipulations of the prey. Deciding between these alternatives in particular cases must await further experimental analysis of the factors involved in the development of these skills. It is of interest that in Groves’ (1978) study of foraging in the ruddy turnstone (Arenaria interpes), the difference between adult and juvenile success lay mainly in prey handling and search times. One of the few studies in which the role of experience in the development of species-typical foraging skills has been studied directly is that of Davies and Green (1976) on the reed warbler (Acrocephalus scirpaceus). They found that whereas success in catching flies (i.e., captures per attempt) was not affected by prior experience with flies, handling time per fly caught was significantly lower in experienced than in nonexperienced birds. Thus, in this one instance, experience does appear to affect an aspect of foraging behavior that has been shown to improve with age. Further research along the lines of Davies and Green’s study is needed to establish the generality of such experiential effects in the development of foraging behavior. In general, there is some circumstantial evidence in favor of the proposition that learning some behavioral skills involves a reduction in early reproductive success and/or a postponement of the onset of breeding. This selective cost will not be incurred to an equal extent by the evolution of all learning abilities. Evolving the ability to learn those skills, such as foraging and nest-site selection, that contribute most directly to breeding success will incur this selective cost most heavily. Skills that can be learned completely before the individual is physiologically or anatomically capable of breeding, or before environmental conditions are suitable for breeding, will in general be more or less exempt from the selective cost of delayed reproduction. The crucial data that are lacking concern the role of experience in the development of skills such as foraging and nest-site selection. There is considerable scope for research on these problems, which would make a substantial contribution to our understanding of the factors involved in the evolution of learning. The work of behavioral ecologists has provided sophisticated descriptive accounts of the strategies employed by adult animals in behavior such as foraging (e.g., Pyke rt a l . , 1977; Krebs, 1978; Pyke, 1978), and some students of learning have begun to turn their attention to the development of such behavior (e.g., Staddon, 1980). If it can be shown that the role of learning is greatest in those species in which breeding is longest delayed, or that have the most reduced breeding success as juveniles, then the argument for a relationship between learning and this selective cost would be further strengthened.

LEARNING A N D NATURAL SELECTION

B.

INCREASEDJUVENILE VULNERABILITY

I.

Nature of the Cost

83

The more an individual must rely on learning for the development of adaptive behavioral skills, the more vulnerable it is during the period before the necessary experience has been obtained, or while it is being obtained. If one individual’s recognition of its food supply, for example, depends more on learning than does that of another individual, then the former is more likely to select an occasional nonnutritious or even poisonous item. An individual whose social behavior is less genetically constrained than another’s runs the risk of responding inappropriately as a juvenile in situations where such a response may be dangerous. Other things being equal then, individuals whose development is more dependent on learning are at higher risk as juveniles than those whose development is less so and, therefore, they will tend to be selected against. Although a good prima facie case can thus be made for the existence of juvenile vulnerability as a selective cost of learning, evidence in its favor is hard to come by. In fact, at first sight it would seem that the evidence tends to support the contrary hypothesis, since juvenile mortality is typically very much higher in animals such as fish, in which behavior appears to owe rather little to individual experience, than in mammals, in which learning is an important component of behavioral development. The reason for this may be that prenatal care is much more prevalent in the latter (see Section V I , C ) and this tends to offset juvenile mortality.

2 . Available Evidence The studies of foraging efficiency discussed previously suggest that in species with complex foraging skills an inexperienced juvenile animal expends more time and energy to obtain each gram of food than does a more experienced adult. Juveniles will thus be more liable to nutritional deficiencies, especially at times when food is scarce, which increases the risk of disease, predation, and accidental death. Bolles (1970) has suggested that the costs in terms ofjuvenile and even adult mortality incurred by learning to avoid predators are so high that such learning is highly unlikely to evolve in any population. He argues that instead animals have evolved species-specific defense reactions (SSDRs) commensurate with their ability to hide, escape, or fight and that these are elicited by any dangerous situation, independently of any specific learning. One way to demonstrate the cost of juvenile vulnerability is to compare juvenile mortality in species having different degrees of dependence on learning a particular skill, when parental care is removed. For example, if two related species differ in the extent to which foraging skills are learned, then that species with the greater dependence on learning should have higher mortality from nutritional deficiencies (and related causes) when parental care is removed. Al-

84

TIMOTHY D. JOHNSTON

though I know of no such comparative studies, Stirling and Latour (1978) found that polar bear (Thalarctos maritimus) cubs less than 2.5 years old were less successful hunters than adults and concluded, on the basis of an admittedly small number of tagging returns, that their mortality is very high if they are deprived of parental care. They suggest that hunting skill in this species depends on experience but concede that there is no direct evidence that this is the case. 3 . Evaluation Direct evidence that dependence on learning incurs the cost of increased juvenile mortality is virtually nonexistent, partly because it is hard to design the appropriate experiments. In principle, examining the increase in juvenile mortality following removal of parental care in two species that differ in their dependence on learning would provide the relevant data, but some important difficulties of interpretation with such a study should be noted. If, as will be argued in Section VI,C, parental care evolves to offset juvenile mortality due to learning, it may also offset mortality from other causes that have nothing to do with learning, such as inadequate thermoregulation or disease transmission by ectoparasites that are removed by grooming. It is important, therefore, that the causes of mortality be determined in any such study and this may be very difficult, especially under field conditions. C. 1.

I N EACHOFFSPRING INCREASED PARENTAL INVESTMENT

Nature of the Cost

In all natural populations, some mortality between birth and maturity is to be expected regardless of the presence of any learning ability. Two parental strategies are available, in principle, to counter the effects of this selection: An individual may produce large numbers of offspring of which at least a few will survive, and invest very little in each one, or it may produce small numbers of offspring and invest heavily in each one. In general, organisms are expected to adopt some compromise between these two alternatives, and the question is, in which direction from the median will the strategy lie? Each end of the continuum has its own costs associated with it. As I have argued, organisms whose behavioral development is heavily dependent on learning may tend to be at higher risk as juveniles than organisms whose development is less dependent on learning and so the requirements to offset expected mortality tend to be higher. In general, increasing the dependence of behavioral development on learning also requires greater structural complexity (see Section VI,D) and so the energetic investment in the production of each offspring tends to be relatively high. A parental strategy in the direction of overproduction of offspring is thus unlikely to be selected for; instead, it is to be expected that an increase in learning capacity will carry with it the requirement for an increased parental investment in each

LEARNING A N D NATURAL SELECTION

85

offspring. This strategy reduces the number of offspring that any individual can rear at one time and, if the period of dependency extends beyond the next breeding season, may reduce the number of times an individual can breed during its lifetime, and, hence, reduce its reproductive success in comparison with other members of the population.

2 . Available Evidence The low reproductive rates of many seabirds and large predatory birds such as condors and eagles have been attributed to the necessity for a long period of postfledging care (Amadon, 1964; Ashmole, 1963). These birds must forage over wide areas to obtain food, expending a large amount of energy to capture each food item, which places a premium on efficient flight and optimum techniques of prey capture. The study of Mueller and Berger (1970) already referred to showed that juvenile sharp-shinned hawks tended to select inappropriate prey more often than did adults. Many seabirds that range over wide ocean areas in search of prey have very long periods of parental care that extend well beyond the time when the young are fully capable of sustained flight (Ashmole and Tovar, 1968). W . Y. Brown (1976) found that parental care in the brown noddy (Anous srolidus) extends for over 100 days postfledging and Simmons (1970) noted a similar period of dependency in the brown booby (Sulu leucogaster). Both authors comment on the skilled fishing techniques of these species and suggest that long experience is required before an individual becomes sufficiently skilled to feed itself. Houston (1976) made determinations of the energy budget of Ruppell’s griffon vulture (Gyps rueppellii) in East Africa and concluded that during at least part of the rearing period the adults are unable to obtain enough food to meet both their own and their young’s energy requirements. A similar conclusion was reached by Hainsworth (1977) in his analysis of the energy budget of breeding sparkling violetears (Colibri coruscans, a species of hummingbird). Since both species breed when food is maximally abundant, these findings suggest that foraging efficiency may be an important limiting factor in their respective energy budgets and that experience may well be necessary for adequate foraging efficiency to develop. Houston (1 976) notes that young vultures spend a considerable amount of time after fledging practicing gliding. These birds must cover large areas by gliding flight in search of food and Houston’s observations suggest the involvement of learning in at least this component of foraging behavior.

3 . Evaluation The finding that seabirds with complex foraging skills show very long periods of parental care suggests that parental care may be a selective cost of learning but, once again, the crucial data that are lacking concern the role of experience in the development of these skills. It seems hard to question the view that increased

86

TIMOTHY D. JOHNSTON

dependence on learning requires increased parental care but evidence in its support remains somewhat scanty. A corollary of this selective cost is that where mortality factors that cannot be offset by any conceivable behavioral skill (such as disease or extreme environmental fluctuations) exert strong selection on the population, a high degree of parental investment in each offspring is unlikely to evolve. In such populations, characteristic of r-selected species (MacArthur and Wilson, 1967). the best reproductive strategy is to produce large numbers of offspring, since even an individual that can attain a high level of behavioral competence as a result of learning has no greater chance of survival, and demands a greater investment in its production, than an individual that cannot learn as much.

D. GREATERCOMPLEXITY OF THE

CENTRAL

NERVOUS SYSTEM

I . Nature of the Cost As the development of behavior becomes more dependent on the n,ature of experience, the individual must process increasing amounts of information during its development. The nature of this processing depends on the particular conception of learning that is being entertained. It may involve the detection of complex relationships among objects and events in the environment (Humphrey, 1933; Johnston, 1978), the formation of associative connections between stimuli and responses (Rescorla and Holland, 1976), or the storage and retrieval of information in memory (Estes, 1973). In any event, as the information detection or processing requirements associated with learning increase, the complexity of the physiological basis for learning in the central nervous system (CNS) also increases. Since the capacity of a single neuron to encode information is largely fixed, increasing the capacity of the CNS as a whole requires more nerve cells and a greater degree of connectivity among them. There are two components to the selective cost of an increase in the size and/or complexity of the nervous system: the high energetic cost of maintaining nerve tissue, and the requirement of shielding the functioning brain from even minor physiological fluctuations. In adult humans, the brain, comprising about 2% of the total body weight, accounts for 20% of the total basal oxygen metabolism (Sokoloff, 1960). In the first decade of life, the consumption of the brain may exceed 50% of the total basal metabolism (Kennedy and Sokoloff, 1957). This high respiratory rate (3.5 mg 02/100g/min in adults) is maintained even during severe metabolic disorders such as hypoglycemia and diabetic acidosis, indicating that during metabolic stress, the cost of maintaining the brain may be proportionately much higher. The brain has an obligatory dependence on glucose metabolism and, since glucose cannot be stored, the brain requires a continuous supply, either by food intake or from other metabolic processes.

LEARNING AND NATURAL SELECTION

87

Recent work (e.g., Oleson, 1971; Raichle, 1975; Lassen et al., 1978) has demonstrated that the local metabolism of individual brain structures is dependent upon their functional activity. A large brain not only incurs a high biological maintenance cost, but also a high operating cost. Such costs are incurred by the diversion of available food resources away from direct investment in reproduction, putting their bearer at a selective disadvantage with respect to other, smaller-brained individuals in the p ~ p u l a t i o n . ~ In addition to this direct energetic cost, there are indirect costs to be borne by the possessor of a large, complex nervous system. Nerve impulse transmission, the significant event of CNS functioning, involves small changes in ion balance, weak bioelectric phenomena, and the diffusion of small amounts of transmitter substances across synaptic gaps. All of these events are sensitive to even minor changes in the chemical milieu that invests the CNS (Tschirgi, 1960). Biochemical noise will be introduced into CNS functioning by changes in the organism’s physiological state unless it is shielded from such fluctuations by mechanisms capable of precise homeostatic control. Such mechanisms, of which the bloodbrain barrier is the best known, have indeed evolved and their development and maintenance is a further cost incurred by the large brain that an increased learning capacity requires. Several experiments have shown that CNS functioning is readily disturbed by minor fluctuations in body temperature (Roots and Prosser, 1962; Tebecis and Phillips, 1968; Peterson and Prosser, 1972) and diversion of resources, whether of time (Heath, 1965) or materials (Wheeler, 1978), into thermoregulatory strategies to control brain temperature represent another indirect cost to the possessor of a large, complex nervous system. Finally, it may be pointed out that the very complexity of the nervous system makes it increasingly vulnerable to accidental mechanical and biochemical disturbances that may render its possessor incapable of successful or prolific reproduction.

2 . Available Evidence Granted that central nervous tissue is a “biologically expensive ” commodity, this expense will only be incurred as learning evolves if an increase in learning ability entails a more elaborate CNS. Evidence concerning the relationship between learning and CNS complexity comes from two sources: comparative studies of learning ability in animals that differ in brain size, and artificial selection experiments involving learning and brain indices. ’It may be worth repeating here that this selective cost is incurred presuming “other things equal.” That is, a large-brained individual will be at a disadvantage with respect to a small-brained one if they are both equally competent in those abilities that affect reproductive success, whether directly or indirectly. Of course there will be situations in which larger-brained individuals have a reproductive advantage (vide the brain size of Homo sopiens), but only when the selective costs discussed here are outweighed by the benefits of possessing such a larger brain.

88

TIMOTHY D. JOHNSTON

u . Comparative Studies. Rensch (1956, 1959) presented data from a variety of species that, he argues, indicate a positive correlation between absolute brain size and learning ability. Unfortunately, his conclusion is hard to evaluate properly because almost none of the comparisons he makes are based on data from the same study and may well reflect differences in task difficulty and/or training procedures. More recent studies of this issue have been equivocal. Riddell and Corl (1977) reviewed data from a number of primate studies supporting Rensch’s conclusions, whereas Miller and Tallarico (1974) found no relation between brain size and detour-learning ability in two species of birds. None of the studies cited above used learning tasks that are natural ones for the species involved and so their relevance to the question of selection for larger brains may be questioned. Eisenberg and Wilson (1978), on the other hand, found that among bats, those species that normally feed on patchily distributed food that is both spatially and temporally unpredictable tend to have larger relative brain sizes than those that feed on more uniformly distributed food. They suggest that the foraging strategies of the former group of species require more learning than those of the latter and offer this as a possible selection pressure for larger relative brain size, but direct experimental data on the role of learning in the development of feeding behavior in bats are not available. Jerison (1970, 1973) presents evidence from fossil studies that the relative brain size (i.e., brain/body ratio) of both carnivores and herbivores showed a progressive increase until recent times, with carnivores having relatively larger brains than the herbivores on which they preyed. This he attributes to reciprocal selection pressures for more elaborate behavioral adaptations for offense and defense by carnivores and herbivores respectively. If such adaptations included the ability to modify behavior on the basis of prior experience, then these data might constitute evidence for the requirement of a larger brain to support learning. Recently, however, Radinsky (1978) has questioned the basis of Jerison’s conclusions and the proper interpretation of these fossil data remains in doubt. 6 . Selectiori Experiments. Since the early experiments of Tryon (1 940) it has been known that selection can produce both increases and decreases in learning abilities of different sorts. More recent studies have examined the correlation between learning ability and various brain indices in differently selected lines, mainly of mice. The data from these studies have been reviewed by Wahlsten (l972), who finds no compelling evidence for a correlation between learning and brain size as a result of selection (see also Jensen, 1977).

3 . Evaluation

Both comparative studies and selection experiments pose serious difficulties of interpretation. As noted above, the majority of comparative studies have used highly artificial learning tasks, e.g., reversals of discrimination and delayed alternation (Riddell and Corl, 1977). No evidence is presented that proficiency in such tasks is of any adaptive significance in the species tested and, if it is not,

LEARNING A N D NATURAL SELECTION

89

then the relevance of such tasks to the evolution of learning is very limited. The comparative approach is also limited by the impossibility of controlling for differences other than learning ability among the species tested. Jerison (1973) discusses a number of other selection pressures besides learning ability that might be responsible for differences in brain indices. He suggests that increasing demands for sophisticated perceptuomotor skills are likely to have been of particular importance in the evolution of the brain, and such skills will usually be very hard to dissociate from learning ability in comparative studies (see also Welker, 1976). Intraspecies selection experiments offer a more direct test of relationships between learning ability and brain size in evolution. However, most experiments have involved selecting for smaller or larger brains and examining the effects of this selection on learning ability (Wahlsten, 1972). A more appropriate design would be to select for learning ability (preferably of some ecologically relevant kind) and examine the effects of this selection on brain size or other indices, since under natural conditions it is learning ability, not brain size, that is the determinant of reproductive success. While such an experiment is certainly possible it does not appear to have been done. In summary, the evidence supporting the popular notion of a relation between brain size and learning ability in the course of evolution is very scanty and mostly equivocal. Perceptual requirements may have played a more important role than learning in brain evolution (Jerison, 1973; Welker, 1976).

E.

GREATER COMPLEXITY OF T H E GENOME

1 . Nature of the Cost

Williams (1966) has suggested that a plastic, or facultative, developmental system, such as is implied by the possession of an enhanced learning capacity, requires a more elaborate genetic specification than one that is less plastic. If this suggestion is correct, then it identifies a further selective cost of learning, since a greater amount of genetic material is more susceptible to the randomizing effects of mutation. In order to offset these effects, either selection must supply mechanisms for the recognition and repair of damaged sequences of DNA (Bessman t'r ul., 1974; Richardson et a l . , 1964; see J . D. Watson, 1970, p. 292, for a discussion of repair mechanisms and their evolutionary significance) or reproductive effort must be increased to compensate for losses due to deleterious mutations (for some examples of neurological mutations and their effects see Sidman et a l . , 1965). 2.

Available Evidence

It is hard to establish the existence of such a selective cost since our present knowledge of the involvement of the genome in developmental processes such as

90

TlMOTHY D. JOHNSTON

learning is virtually nonexistent. While it seems reasonable to suppose that developmental plasticity requires more elaborate genetic specification, it seems equally plausible, in principle, to argue that if the structure of behavior is more closely specified by the animal’s experience (as a result of learning) less of the specification need be supplied by the genome. Indeed, limits on the capacity of the genome to encode information have been offered as one source of selection pressure in favor of learning (Nottebohm, 1972; Mayr, 1974) and, on this view, learning would require less rather than more genetic material. Two lines of evidence suggest that the first of these two arguments may be the correct one, although both lines are indirect. In the first place, any facultative developmental process should, at the molecular level, require more genetic material than a nonfacultative one, since more regulatory genes must exist to control the underlying changes in protein synthesis. Consider the lac operon in the bacterium Escherichia coli, which is concerned with the production of enzymes for lactose metabolism. Here, three structural genes are responsible for the production of the enzyme p-galactosidase and of two other enzymes involved in lactose metabolism. In the absence of lactose in the environment, these genes are repressed by a repressor substance produced by a neighboring regulatory gene. In the presence of lactose, the repressor is inactivated and transcription of the operon is induced. The structural gene in this case comprises some 5100 DNA base pairs. the regulatory gene adds another 1020 base pairs (Watson, 1970). If the lac operon were not facultative but instead produced p-galactosidase continuously, only 83% of the total DNA involved in its functioning would be necessary. This example is not intended to suggest that the behavioral events of learning correspond in any simple manner to the regulated activity of single genes. Ultimately, however, all developmental events are determined by an interaction between the coordinated activity of the genome and the environment and the preceding example suggests the need for a more complex genetic base for such determination when it is facultative rather than obligatory. The second line of evidence is provided by studies of the proportion of transcribable DNA that is present in different body tissues. Grouse et al. (1972), using RNA hybridization techniques, found that 1 1 % of the DNA of brain neurons is transcribable (i.e., actively involved in protein synthesis and regulation), compared with 4-5% in liver, kidney, and spleen. That means that a greater proportion of the total DNA is involved in CNS function than in the function of other tissues, suggesting that if the evolution of a more complex CNS is required to support an increased learning capacity, then this may involve the acquisition of more DNA sequences. A set of related findings of interest is that rats reared in an enriched environment have larger brains with higher connectivity (Rosenzweig, 1966), show an enhancement of learning ability on at least some learning tasks (Krech el a l . ,

LEARNING A N D NATURAL SELECTION

91

1962), and have brains with a higher proportion of transcribable DNA (Grouse er a l . , 1978), than do rats reared under impoverished conditions. The proper interpretation of these findings is not entirely clear, but they point to some interesting relationships among CNS complexity, genome complexity, and learning ability that invite further investigation, preferably involving ecologically relevant learning tasks.

3 . Evaluation Any claims that learning requires a more complex genetic basis are, at best, highly speculative. The indirect evidence cited above is intriguing but it is offered more in the hope that it will stimulate further thought on this issue than as evidence in favor of the proposition.

F.

DEVELOPMENTAL FALLIBILITY

I.

Nature of the Cost

The very flexibility of behavior that is the primary selective benefit of learning (Section V,A) implies that by not being exposed to, or not attending to, the appropriate kinds of environmental influences, an animal may develop maladaptive behavior patterns. Thus there will tend to be selection against forms of learning that do not incorporate some form of “protection” against such maladaptive developmental responses. We therefore expect to find extant examples of learning associated with some form of protection.

2.

Evidence

Protection against maladaptive development may be of two kinds: social or developmental. Social protection includes parental care which both ameliorates the adverse consequences of behavioral errors during early learning and helps to increase the uniformity of the early environment, ensuring that appropriate experiences for learning are available. The possibility of an association between increased parental care and learning has already been discussed (Section VI,C). Developmental protection includes the provision of constraints on development that specifically preclude maladaptive developmental responses to the environment. There is extensive evidence that such constraints exist (e.g., Shettleworth, 1972) and their possible role as protection against developmental error has recently been discussed by Marler (1977) in regard to song learning. Many species of songbirds require exposure to an adult song model for normal vocal development (Marler and Mundinger, 197 1). Under natural conditions, the young bird is surrounded by a complex acoustic environment, yet it selectively learns the conspecific song rather than others that it may hear. This selectivity reflects developmental constraints on the stimuli that are accepted as song models

92

TIMOTHY D. JOHNSTON

(Nottebohm, 1970; Marler and Peters, 1977; Marler, 1977), a phenomenon sometimes referred to as an “auditory template” (e.g., Marler, 1976). In some cases, social bonds may also play a role in ensuring normal song development (Immelmann, 1969; Dietrich, 1980). Maternally naive ducklings, and other precocial birds, show a selective approach response to the maternal call of their species shortly after hatching (Gottlieb, 1971). This response may serve to prevent the formation of inappropriate social attachments (Gottlieb, 1965; Johnston and Gottlieb, 198 I ) , despite the very wide range of artificial objects to which attachments may be formed under laboratory conditions (Sluckin, 1973). 3 . Evaluation

The existence of adaptive constraints in development that prevent learning from going astray is probably very widespread. Learning can only be a successful component of an animal’s adaptive strategy to the extent that such constraints are provided and, in their absence, learning may be expected to incur a heavy selective cost for its possessor.

VII.

LEARNING AND

THE

ADAPTIVE COMPLEX

In many cases, the existence of selective costs and benefits associated with the evolution of learning rests largely on prima facie arguments that are supported by only limited amounts of empirical evidence. It is clear that more evidence is needed before the selective pressures associated with the evolution of learning can be identified with any confidence and I hope that by indicating the kinds of evidence that are required, the preceding discussion may contribute to the eventual solution of this problem. In order to simplify the discussion, I have presented each of the selective costs and benefits independently of one another. However, it would be a mistake to leave the impression that in any real instance of evolution, the selective picture will be so simple. Interactions among selective pressures are likely to be the rule rather than the exception. It is, furthermore, a biological truism that organisms evolve as adapted wholes rather than as assemblages of adaptive traits. In order to gain some insight into the evolution of particular traits, such as the various learning abilities that an animal may possess, it is necessary to abstract the trait(s) of interest from this adaptive complex and consider them in isolation, but it is important not to lose sight of the numerous interactions that occur in evolution, or of the consequences of those interactions. In this section, I wish to consider some of the more obvious problems that arise when one attempts to consider the evolution of learning within the context of the overall adaptive complex.

LEARNING A N D NATURAL SELECTION

A.

93

INTERACTIONS O F SELECTIVE PRESSURES A N D ADAPTIVE TRAITS

Interactions among the various costs and benefits of learning are unlikely to be linearly additive, since there will most likely be reciprocal interactions among the various selection pressures that act on an evolving population. The evolution of different learning abilities will not involve the same selective costs and benefits, nor will the same cost or benefit always be involved to the same extent. Different stages of the life cycle will be exposed to different selection pressures and will involve different kinds and degrees of learning so that the selective context vis-a-vis learning will change, probably in complex ways, as the organism develops. Given the current lack of data on ecologically relevant learning abilities that might permit an evaluation of selective costs in the context of life history traits, breeding biology, and other relevant characteristics of particular organisms, it is probably futile to attempt a detailed integration of the overall selective picture attendant on the evolution of learning. To avoid leaving an overly simplified impression, however, it is worthwhile to indicate, albeit briefly and in general terms, some of the interactions that probably do exist. At the very least, such a discussion may help to stimulate further thought on the problems involved. Those selective costs most directly concerned with breeding success (i.e., delayed reproduction, juvenile vulnerability, and parental care) are perhaps most likely to show strong nonadditive interactions during the evolution of learning. Juvenile mortality and parental care are in many respects complementary costs: Animals that incur the latter are less likely to have to bear the former as well. If the evolution of learning incurs the cost of added CNS complexity, on the other hand, then that of parental care may become obligatory, because of the energetic investment represented by each offspring. As more parental care is provided, the cost implied by the fallibility of learning may become less important, although this will depend on the nature of the care provided as well as on its extent (in terms of days postpartum). The costs of learning interact not only with each other but also with other elements of the organism’s overall adaptation to the environment. Numerous examples of such interactions may easily be thought of. If selection favors the evolution of complex perceptuomotor skills for reasons unrelated to learning, requiring an increase in CNS complexity, then the subsequent or concurrent evolution of a learning ability may be partially exempt from that cost. The evolution of a complex social system, especially one that involves a relatively close-knit family unit, may relieve some of the costs of parental care, perhaps by kin selection acting through nonreproducing relatives (e.g., the “aunt” role in some primate species: Hrdy, 1976; McKenna, 1979). On the other hand, the evolution of such a social system will probably require the ability to identify individual group members (among other requirements) and such identities must

94

TIMOTHY

D.

JOHNSTON

of course be learned. The evolution of a learning ability may not incur the cost of delayed reproduction if selection acts to favor such a delay because of other physiological or ecological requirements of breeding, such as attaining a certain body size or obtaining a breeding territory. Not all selective pressures act equally throughout the life cycle. Delayed reproduction is a cost that is incurred only early in life and may be offset if an increase in learning ability enables the reproductive period to be extended into later life, or permits a greater reproductive success when reproduction does begin. Parental care, on the other hand, cannot be a selective factor until after the onset of reproduction. The interaction of these costs with other elements of the selective context will therefore change during the life cycle and they may tend to be offset by appropriate scheduling of other developmental events. These considerations suggest that theories of life-history evolution (Steams, 1977), which are primarily concerned with reproductive parameters such as age at maturity and clutch size, might include the evolution of learning as part of their analyses. Finally, it should be noted that the evolution of one kind of learning ability may make the evolution of others less expensive. If selection favors, say, a delay in reproduction to permit the evolution of one learning ability, then the evolution of others will not incur this selective cost and there may be a “snowball” effect, permitting the relatively rapid evolution of several learning abilities, even if their selective benefits, taken individually, are relatively slight. Of course, if the selective picture changes, so that the first ability becomes less advantageous, then the other abilities may rapidly incur a heavy selective cost and individuals in which are less well developed may be at a sudden selective disadvantage. As in any discussion of evolutionary phenomena, such complexities might be multiplied almost indefinitely and no useful purpose would be served by extending this discussion further. Frazzetta (1975) provides a most illuminating discussion of the problems inherent in explaining the evolution of complex adaptations, of which learning is certainly one. A mature evolutionary account of learning will have to cope with these problems and in this preliminary discussion I can do no more than draw attention to their existence.

B. LIMITSON ADAPTIVE PRECISION It has sometimes been argued, e.g., by Cody (1974), that natural selection produces an optimally adapted phenotype, that is, a phenotype that is precisely tailored to the requirements of the organism in its environment. This view has been justly criticized by Ghiselin (1966, 1974), Darlington (1977), Lewontin (1978, 1979), and Gould and Lewontin (1979)-indeed, Darwin (1859) was well aware of the limits of natural selection in effecting adaptation. Several factors limit the possibilities for adaptive precision, including insufficiently strong selection pressure, competing adaptive demands on the organism, pleiotropy, genetic

LEARNING A N D NATURAL SELECTION

95

linkage, correlated growth effects, and lack of appropriate genetic or phenotypic variation. An evolutionary account of learning, to which this article is a contribution, must avoid what Lewontin (1979) calls the “adaptationist fallacy” of assuming that all characteristics of an organism are tailored to subserve some adaptive function. Let us suppose that in some population, selection favors the evolution of a particular learning ability, L, say the ability to adopt new foraging patterns on the basis of changes in prey distribution or abundance (e.g., Bent, 1945; L. Tinbergen, 1960). Does this mean that the outcome of selection (assuming selection to be successful) will be only those characteristics necessary to effect L, and that individuals in the selected population will exhibit no other new learning abilities? Natural selection theory strongly suggests that this will not be the case and to argue otherwise is to commit the adaptationist fallacy. While selection will produce at least those characteristics sufficient to effect L, there is no reason to suppose that this will be the only outcome of selection, a fact that is clearly shown by studies of artificial selection on other traits. For example, Pyle (1976, 1978) has shown that selection for positive or negative geotaxis in Drosophila rnelanogaster can have correlated effects on other characters that were not directly selected. These characters include female oviposition site preferences (Pyle, 1976), mating speed, courtship duration, locomotor activity, and aristal morphology (Pyle, 1978). Similar sorts of effects are very likely to occur in the course of selection for learning abilities. There are two kinds of account that may be given of this outcome (discussed by Pyle, 1978). If the genes controlling geotaxis are pleiotropic or are closely linked to other genes with different functions, then some correlated characters may be incidental by-products of selection. Natural selection can establish such characters in a population if they are adaptively neutral, as is particularly the case for what may be called “hidden phenotypic characters.” Characters may be hidden either.physically (e.g., the color of internal organs; Lewontin, 1978) or ecologically, if they are not revealed under the conditions in which the animal lives and in which selection is operating. I shall return to the latter possibility. On the other hand, some correlated characters may, unknown to the investigator, be essential to the performance of the task being selected-Pyle (1978) suggests that the aristae may be involved in geotactic behavior and that changes in aristal morphology may subserve the changes in geotactic behavior that were the direct outcome of selection in his experiment. The probable widespread existence of such correlated effects of selection has some important consequences for the evolution of learning. As one learning ability is selected for in a population, others may evolve as correlated effects of selection. This may be the result of pleiotropism or linkage, or it may be that the physiological mechanisms involved in the selected learning ability incidentally support other abilities of no adaptive relevance. It is likely that such abilities will

96

TIMOTHY D. JOHNSTON

remain hidden in individuals that possess them, because the experiences necessary for them to be revealed in development will not be provided by the environment in which the population lives. We might call such abilities “ecologically surplus” learning abilities because they make no contribution to the animal’s ontogenetic adaptation to its environment (see Boice, 1977). If circumstances change, such surplus abilities may be available as preadaptations (Bock, 1959; Bock and von Wahlert, 1965; Gans, 1979). Surplus abilities may also be revealed in experiments on animal learning, in which a nonnatural environment is provided for the animal, and in the concluding section of this article I shall consider some of the implications of this point for the study of learning.

VIII.

IMPLICATIONS FOR THE STUDY OF

LEARNING

In this article I have provided a preliminary account of some of the selective costs and benefits that are involved in the evolution of learning by natural selection. Although in some cases rather little evidence for the existence of a particular cost or benefit is available, overall I feel that we are justified in concluding that despite its selective benefits, the evolution of learning almost always incurs some selective costs and that these costs may sometimes be very heavy. On this analysis, it does not seem particularly surprising that learning is not more widespread in the animal kingdom (cf. Mayr, 1974). The situations in which learning confers a net selective benefit on its possessors may indeed be few and in attributing “obvious” advantages to learning, we may merely be reflecting an unwarranted anthropocentric bias (Nottebohm, 1972). If this is an accurate assessment of the situation attendant on the evolution of learning, then we are led to a position that may have important implications for the way we should go about the study of learning. It seems rather unlikely that a “general learning ability” exists that has been increased and refined by natural selection in the course of evolution, since this view implies a more or less universal advantage to the ability to learn (Rozin, 1976). A more appropriate view might be that selection acts to favor those individuals that possess the least costly learning abilities permitting successful reproduction in the environment in which the population is evolving. Nothing is gained, and much may be lost, by an individual that possesses ecologically surplus learning abilities. Animals may thus be expected to possess close to the minimum effective learning ability that is adaptive for them, although, based on the arguments in Section VII,B, we expect to find at least some ecologically surplus abilities in most animals that are able to learn. The distinction between adaptive and surplus abilities is not a trivial one. The former abilities reflect a particular ecological relationship between an animal and its environment that is the result of many generations of selection. The later, by

LEARNING A N D NATURAL SELECTION

97

contrast, reflect what we might call the “operating characteristics” of the underlying physiological and/or genetic support for learning. The possible interpretations that may be given to these two kinds of ability, i.e., the roles that they play in theories of learning, are clearly quite different and it is very important that our theories offer some way to distinguish between them (see Johnston, 1981). Current theories of learning provide no basis for such a distinction because they assign no theoretical significance to descriptions of the animal’s natural environment; and it is description of the environment that is precisely what is required to distinguish adaptive from surplus learning abilities. An ecological theory of learning, incorporating a description of the learner’s natural environment as an integral part, would provide the basis for distinguishing between adaptive and surplus abilities and would stand in sharp contrast to current theories of learning, which are thoroughly nonecological. A possible approach to such a theory has been discussed in detail elsewhere (Johnston and Turvey, 1980; Johnston, 1981); at present, let me simply remark on some implications of accepting the surplus/adaptive distinction for the conduct of research on animal learning. Many, perhaps most, of the tasks used to investigate the learning abilities of animals have no discernable ecological relevance. To the extent that animals are able to perform them, they presumably either mimic tasks that are ecologically relevant or tap ecologically surplus learning abilities. In either case, the use of such tasks may make a contribution to ecological theories of learning, but only if we know which of these two possibilities is the case. If the former, then the task can serve as a simplified experimental situation for the study of an ecologically relevant learning ability. In the case of a surplus ability, the contribution is of a different kind, since the nature of such an ability cannot be understood in terms of its contribution to ecological adaptation. Its analysis may however contribute to an understanding of the mechanisms underlying ecologically relevant learning abilities, in the same way that an analysis of developmental pathologies (i.e., responses to stimuli for which the developing system is evolutionarily unprepared) can contribute to our understanding of normal development. It is obviously important that we be able to distinguish in our theories between normal and pathological developmental processes and this is just as true of theories of learning as it is of theories of other developmental phenomena. Analysis of an animal’s learning abilities may be likened to the process of analyzing the operation of a complex system such as a computer program. When a human designer produces such a system, he or she will usually try to anticipate the kinds of inappropriate inputs it may receive when in use and include some error-catching facilities in the design. If a program is intended to operate on numerical data, alphabetic input may produce a warning message, rather than a program crash or, worse, uninterpretable output. Since natural selection cannot anticipate, the use of biologically inappropriate stimuli in learning experiments

98

TIMOTHY D. JOHNSTON

may well produce behavior that is uninterpretable from an ecological point of view. Unfortunately, human ingenuity is such that some interpretation of the behavior will usually be found, but such interpretations only contribute to our understanding of contrived and possibly artifactual phenomena. These various considerations suggest a change in emphasis in the kinds of problems that are studied by investigators of animal learning. Very few of the problems currently under investigation (see Bitterman et al., 1979) have any clear relevance to the problems faced by animals outside the artificial environment of the laboratory4; at least, if they do, their relevance remains to be demonstrated by those engaged in such investigation. In an ecological approach to learning, the primary focus of experimental analysis would be on the role that experience plays in the development of behavioral skills that the animal normally employs in its dealings with the environment (Marler, 1975; Gottlieb, 1976b). Ethologists have been making this suggestion for many years (e.g., N . Tinbergen, 1951; Lorenz, 1965), and the arguments in this article lend it additional support. The literature of ethology and behavioral ecology abounds with detailed descriptions of such skills and there is enormous scope for the application of laboratory techniques of developmental analysis to the study of their acquisition (for examples of such application, see Marler, 1970; Emlen, 1972; Hall et a l ., 1977; Freeman and Rosenblatt, 1978; Gottlieb, 1979; Gray and Tardif, 1979). Studies of this kind will lay the basis for any future ecological theory of learning, which will be necessary to an understanding of the evolution of this important mode of ontogenetic adaptation.

Acknowledgments I am grateful to Gilbert Gottlieb, David B. Miller, Ronald W . Oppenheim, and the editors of this volume for many helpful comments on earlier versions of the manuscript. The paper was completed while I was a Visiting Scientist at the Institute of Animal Behavior, Rutgers University, and I am grateful to Jay S . Rosenblatt and Colin Beer for providing that opportunity. Discussions with Gregory Ball and Sarah Lenington clarified my thinking on a number of points. Preparation of this paper was supported in part by Grant No. HD-00878 from the National Institute of Child Health and Human Development.

References Aceves-Pina, E. 0.. and Quinn, W . G. 1979. Leaming in normal and mutant Drosophilu larvae. Science 206, 93-96. Alcock, J. 1979. “Animal Behavior: An Evolutionary Approach. ” 2nd. ed. Sinauer, Sunderland, Massachusetts. 4Possible exceptions to this general statement are studies of taste aversion learning (Rozin and Kalat, 1971; Kalat, 1977) and of avoidance leaming (Bolles, 1970, 1971; see Shettleworth, in press).

LEARNING AND NATURAL SELECTION

99

Amadon, D. 1964. The evolution of low reproductive rates in birds. Evolution 18, 105-110. Anastasi, A. 1954. The inherited and acquired components of behavior. I n “Genetics and the Inheritance of Integrated Neurological and Psychiatric Patterns’’ (D. Hooker and C. C. Hare, eds.), pp. 67-75. Williams & Wilkins, Baltimore, Maryland. Angell, J . R. 1907. The province of functional psychology. Psychol. Rev. 14, 61-91. Arnold, S. J . 1978. The evolution of a special class of modifiable behaviors in relation to environmental pattern. Am. Nar. 112, 415-427. Ashmole, N . P. 1963. The regulation of numbers of tropical oceanic birds. Ibis 103b, 458-473. Ashmole, N . P., and Tovar, S. H.1968. Prolonged parental care in Royal Terns and other birds. Auk 85, 90-100. Bateson, G. 1963. The role of somatic change in evolution. Evolurion 17, 529-539. Bateson, P. P. G. 1976. Specificity and the origins of behavior. I n “Advances in the Study of Behavior” (J. S. Rosenblatt. R. A. Hinde, E. Shaw, and C. Beer, eds.), Vol. 6, pp. 1-20. Academic Press, New York. Benk, F. 1945. The role of learning in the feeding behaviour of Black-chinned Hummingbirds. Condor 47, 3-22. Bessman, M. J . , Muzycka, N., Goodman, M. F., and Schnaar, R. L. 1974. Studies on the biochemical basis of spontaneous mutation. 11. The incorporation of a base and its analogue into DNA by wild-type, mutator, and anti-mutator DNA polymerases. J. Mol. Biol. 88, 409-421. Bitterman, M. E. 1965. Phyletic differences in learning. Am. Psychol. 20, 396-410. Bitterman, M . E. 1975. The comparative analysis of learning. Science 188, 699-709. Bitterman, M. E . , LoLordo, V. M., Overmier, J. B., and Rashotte, M. E. (eds.). 1979. “Animal Learning: Survey and Analysis.” Plenum, New York. Blus. L. J . , and Keahey, J . A. 1978. Variation in reproductivity with age in the brown pelican. Auk 95, 128-134. Bobisud, L. E . , and Potratz, C. J . 1976. One-trial learning versus multi-trial learning for a predator encountering a model-mimic system. Am. Nar. 110, 121-128. Bock, W. J . 1959. Preadaptation and multiple evolutionary pathways. Evolurion 13, 194-21 I . Bock, W . J . , and von Wahlert, G. 1965. Adaptation and the form-function complex. Evolurion 19, 269-299. Boice, R. 1977. Surplusage. Bull. Psychon. Sci. 9, 452-454. Bolles, R. C. 1970. Species-specific defense reactions and avoidance learning. Psychol. Rev. 77, 32-48. Bolles, R. C. 1971. Species-specific defense reations. I n “Aversive conditioning and learning’’ (F. R. Brush, ed.), pp. 183-233. Academic Press, New York. Bolles, R. C. 1975. “Learning Theory.” Holt, New York. Boring, E. G. 1950. The influence of evolutionary theory upon American psychological thought. I n “Evolutionary Thought in America” (S. Persons, ed.), pp. 268-298. Yale Univ. Press, New Haven, Connecticut. Brown, J . L. 1975. ”The Evolution of Behavior.” Norton, New York. Brown, W. Y. 1976. Prolonged parental care in the sooty tern and brown noddy. Condor 78, 128-129. Buckley, F. G., and Buckley, P. A. 1974. Comparative feeding ecology of wintering adult and juvenile Royal Terns. Ecology 55, 1053-1063. Capitanio, S. J . , and Leger, D. W. 1979. Evolutionary scales lack utility: A reply to Yarczower and Hazlett. Psychol. Bull. 86, 876-879. Carmichael, L. 1925. Heredity and environment: Are they antithetical? J . Abnorm. Soc. Psychol. 20, 245-260. Cody, M. L. 1974. Optimization in ecology. Science 183, 1156-1 164.

100

TIMOTHY D. JOHNSTON

Coulson, 1. C . , and Horobin, J. 1976. The influence of age on the breeding biology and survival of the Arctic tern, Sterna paradisaea. J . Zool. 178, 247-260. Coulson, J. C., and White, E. 1961. An analysis of the factors influencing the clutch size of the Kittiwake. Proc. Zool. SOC. Lond. 136, 207-217. Cravens, H., and Burnham, J. C. 1971. Psychology and evolutionary naturalism in American thought, 1890-1940. Am. Quart. 23, 635-657. Darlington, P. J . 1977. The cost of evolution and the imprecision of adaptation. Proc. Nutl. Acrid. Sci. U.S.A. 74, 1647-1651. Darwin, C. R. 1859. “On the Origin of Species by Means of Natural Selection, or the Preservation of Favoured Races in the Struggle for Life.” John Murray, London. Darwin, C. R. 1871. “The Descent of Man, and Selection in Relation to Sex.” John Murray, London. Darwin, C. R. 1872. “The Expression of the Emotions in Man and Animals.’’ John Murray, London. Davies, N. B., and Green, R. E. 1976. The development and ecological significance of feeding techniques in the reed warbler (Acrocephalus scirpaceus). Anim. Eehav. 24, 2 13-229. DeBenedictis, P. A. 1978. Are populations characterized by their genes or by their genotypes’? Am. Nut. 112, 155-175. Dietrich, K. 1980. Vorbildwahl in der Gesangsentwicklung beim Japanischen movchen (Lunchuru striata var. domestica, Estrildidae). Z . Tierpsychol. 52, 57-76. Dudai, Y. 1979. Behavioral plasticity in a Drosophila mutant, dunceDB27fi. J . Comp. Physiol., A 130, 271-275. Dunn, E. K. 1972. Effect of age on the fishing ability of Sandwich Terns, Sterna sandvicensis. Ibis 114, 360-366. Eisenberg, J. F., and Wilson, D. E. 1978. Relative brain size and feeding strategies in the Chiroptera. Evolution 32, 740-75 1. Emlen, S. T. 1969. The development of migratory orientation in young indigo buntings. Living Bird 8, 113-126. Emlen, S . T. 1970. Celestial rotation: Its importance in the development of migratory orientation. Science 170, 1198-1201. Emlen, S. T. 1972. The ontogenetic development of orientation capabilities. In “Animal Orientation and Navigation” (S. R. Galler. K . Schmidt-Koenig, G. J. Jacobs, and R. F. Belleville. eds.), pp. 191-210. NASA, Washington, D.C. (SP-262). Emlen, S. T. 1975. The stellar-orientation system of a migratory bird. Sci. Am. 233(2), 102-1 1 1 . Estabrook, G. A., and Jesperson, D. C. 1974. Strategy for a predator encountering a model-mimic system. Am. Nat. 108, 443-457. Estes, W. K . 1973. Memory and conditioning. In “Contemporary Approaches to Conditioning and Learning’’ (F. J . McGuigan and D. B. Lumsden, eds.), pp. 265-286. Wiley, New York. Frazzetta, T. H. 1975. “Complex Adaptations in Evolving Populations. Sinauer, Sunderland, Massachusetts. Freeman, N. C. G., and Rosenblatt, J. S. 1978. The interrelationship between thermal and olfactory stimulation in the development of home orientation in newborn kittens. Dev. Psychobiol. 11, 437-457. Galef, B. G . 1981. The ecology of weaning: Parasitism and the achievement of independence by altricial mammals. In “Parental Care” (D. J. Gubernick and P. H. Klopfer, eds.). Plenum, New York. Gans, C. 1979. Momentarily excessive construction as the basis for protoadaptation. Evolution 33, 227-233. Garcia, J., and Koelling, R. A. 1966. Relation of cue to consequence in avoidance learning. Psychon. Sci. 4, 123-124. ”

LEARNING AND NATURAL SELECTION

101

Garcia, J . , Ervin, F. R., and Koelling, R. A. 1966. Learning with prolonged delay of reinforcement. Psyhon. Sci. 5, 121-122. Gesell. A. 1933. Maturation and the patterning of behavior. In “A Handbook of Child Psychology” (C. Murchison. ed), 2nd. ed., pp. 209-235. Clark Univ. Press, Worcester, Massachusetts. Ghiselin, M. T. 1966. On semanatic pitfalls of biological adaptation. Phifos. Sci. 33, 147-153. Ghiselin, M T. 1974. “The Economy of Nature and the Evolution of Sex.” Univ. of California Press, Berkeley. Gottlieb, G. 1965. Imprinting in relation to parental and species identification by avian neonates. J. Comp. Phvsiol. Pqchol. 59, 345-356. Gottlieb, G. 1971. “Development of Species Identification in Birds.” Univ. of Chicago Press, Chicago, Illinois. Gottlieb, G. 1976a. Conceptions of prenatal development: Behavioral embryology. Psychol. Rev. 83, 215-234. Gottlieb, G. 1976b. The roles of experience in the development of behavior and the nervous system. In “Studies on the Development of Behavior and the Nervous System. Vol. 3. Neural and Behavioral Specificity” (G. Gottlieb, ed.), pp. 25-54. Academic Press, New York. Gottlieb, G. 1978. Development of species identification in ducklings: IV. Change in speciesspecific perception caused by auditory deprivation. J. Comp. Physiol. Psychol. 92, 375-387. Gottlieb, G. 1979. Comparative psychology and ethology. In “The First Century of Experimental Psychology” (E. Hearst, ed.), pp. 147-176. Erlbaum, Hillsdale, New Jersey. Gottlieb, G. 1980. Development of species identification in ducklings: VI. Specific embryonic experience required to maintain species-specific perception in Peking ducklings. J. Comp. Phvsiol. Psychol. 94, 579-587. Could, S. J., and Lewontin, R. C. 1979. The spandrels of San Marco and the Panglossian paradigm: A critique of the adaptationist program. Proc. R. SOC. h n d . B 205, 581-598. Gray, L. 1979. Feeding diversity in deermice. J. Comp. Physiol. Psychol. 93, I 118-1 126. Gray, L. 1981, Genetic and experiential differences affect foraging behavior. In “Foraging Behavior: Ecological, Ethological, and Psychological Approaches” (A. Kamil and T. D. Sargent, pp. 455-473. Garland STPM, New York. Gray, L., and Tardif, R. R. 1979. Development of feeding diversity in deermice. J. Comp. Physiol. Pqchol. 93, 1127-1 135. Grouse, L., Chilton, M.-D., and McCarthy, D. J. 1972. Hybridization of ribonucleic acid with unique sequences of mouse deoxyribonucleic acid. Biochemisrry 11, 798-805. Grouse, L. D., Schrier, B. K., Bennett, E. L.,Rosenzweig, M. R., and Nelson, P. G. 1978. Sequence diversity studies of rat brain RNA: Effects of environmental complexity on rat brain RNA diversity. J. Neurochem. 30, 191-203. Groves, S. 1978. Age-related differences in ruddy turnstone foraging and aggressive behavior. Auk 95, 95-103. Hainsworth, F. R. 1977. Foraging efficiency and parental care in Colibri coruscans. Condor 79, 69-75. Hall, W. G . , Cramer, C. P., and Blass, E. M. 1977. The ontogeny of suckling in rats: Transitions toward adult ingestion. J. Comp. Physiol. Psychol. 91, 1141-1 155. Harlow, H. F. 1958. The evolution of learning. In “Behavior and Evolution” (A. Roe and G. G. Simpson, eds.), pp. 269-290. Yale Univ. Press, New Haven, Connecticut. Heath, J . E. 1965. Temperature regulation and diurnal activity in homed lizards. Univ. Calif. Publ. Zool. 64(3), 97-136. Hinde, R. A. 1973. Constraints on learning-an introduction to the problems. In “Constraints on Learning: Limitations and Predispositions” (R. A. Hinde and J. Stevenson-Hinde, 4 s . ) . pp. 1-19. Academic Press, New York. Hinde, R. A. 1975. The concept of function. In “Function and Evolution in Behaviour: Essays in

102

TIMOTHY D. JOHNSTON

Honour of Professor Niko Tinbergen, FRS” (G.Baerends, C. G.Beer, and A. Manning, eds.). pp. 3-15. Clarendon, Oxford. Hinde, R. A,, and Stevenson-Hinde, J. (eds.) 1973. “Constraints on Learning.” Academic Press, New York. Hodos, W., and Campbell, C. B. G . 1969. Sculu nuturue: Why there is no theory in comparative psychology. Psychol. Rev. 76, 337-350. Houston, D. C. 1976. Breeding of the White-backed and Riippell’s Griffon Vultures, Gyps ufricunus and G. rueppellii. Ibis 118, 14-40. Hrdy, S. B. 1976. Care and exploitation of nonhuman primate infants by conspecifics other than the mother. In “Advances in the Study of Behavior” (J. S. Rosenblatt, R. A. Hinde, E. Shaw, and C. Beer, eds.), Vol. 6, pp. 101-158. Academic Press, New York. Hull, C. L. 1929. A functional interpretation of the conditioned reflex. Psychol. Rev. 36, 498-51 I . Hull, C. L. 1937. Mind, mechanism, and adaptive behavior. Psyehol. Rev. 44, 1-32. Hull, C. L. 1943. “Principles of Behavior.” Appleton, New York. Humphrey, G . 1933. “The Nature of Learning in Its Relation to the Living System.” Kegan Paul, London. Immelmann, K. 1969. Song development in the zebra finch and other estrildid finches. In “Bird Vocalizations” (R. A. Hinde, ed.), pp. 61-74. Cambridge Univ. Press, London and New York. James, W. 1890. “Principles of Psychology.” Holt, New York. Jensen, C. 1977. Generality of learning differences in brain-weight selected mice. J. Comp. Physiol. Pvchol. 91, 629-641. Jerison, H. J. 1970. Brain evolution: New light on old principles. Science 170, 1224-1225. Jerison, H. J. 1973. “Evolution of the Brain and Intelligence.” Academic Press, New York. Johnston, T. D. 1978. “Ecological dimensions of learning: A study of motor skill acquisition in infant primates. Unpublished doctoral dissertation, University of Connecticut. Johnston, T.D. 1981. Contrasting approaches to a theory of learning. Behuv. Bruin Sci. 4, 125-173. (with commentary). Johnston, T. D. Learning and the evolution of developmental systems. In “Readings and Essays in Evolutionary Epistemology” (H. C. Plotkin, ed.). Wiley, New York. (In press.) Johnston, T. D., and Gottlieb, G. 1981. Development of visual species identification in ducklings: What is the role of imprinting? Anim. Behav. 29, 1082-1099. Johnston, T. D., and Turvey, M. T. 1980. A sketch of an ecological metatheory for theories of learning. In “The Psychology of Learning and Memory” ( G . H.Bower, ed.), Vol. 14, pp. 147-205. Academic Press, New York. Kalat, J. W. 1977. Biological significance of food aversion leaming. In “Food Aversion Learning” (N. W. Milgram, L. Krames, and T. M. Alloway, eds.), pp. 73-103. Plenum, New York. Kear, J. 1962. Food selection in finches with special reference to interspecific differences. Proc. Zool. SOC.London, 138, 163-204. Kennedy, C., and Sokoloff, L. 1959. An adaptation of the nitrous oxide method to the study of the cerebral circulation in children: Normal values for cerebral blood flow and cerebral metabolic rate in childhood. Clin. Invest. 36, 1130-1 137. Kimble, G . A. 1967. “Foundations of Conditioning and Learning.” Appleton, New York. Krebs, J. 1978. Optimal foraging: Decision rules for predators. In “Behavioural Ecology” (J. R. Krebs and N. 9. Davies, eds.), pp. 23-63. Blackwell’s, Oxford. Krech, D., Rosenzweig, M. R.. and Bennett, E. L. 1962. Relations between brain chemistry and problem-solving among rats raised in enriched and impoverished environments. J . Comp. Physiol. Psychnl. 55, 801-807. Kroodsma, D. E. 1976. Reproductive development in a female songbird: Differential stimulation by quality of male song. Science 192, 574-575. Kroodsma, D. E. 1977. A re-evaluation of song development in the song sparrow. h i m . Behuv. 25, 390-399. ”

LEARNING A N D NATURAL SELECTION

103

Kuo. 2 . Y. 1922. How are our instincts acquired? Psychol. Rev. 29, 344365. Lack, D. 1954. “The Natural Regulation of Animal Numbers.” Oxford Univ. Press, London and New York. Lack, D. 1966. “Population Studies of Birds.” Oxford Univ. Press, London and New York. Lassen, N. A., Ingvar, D. H., and Skinhoj, E. 1978. Brain function and blood flow. Sri. Am. 239(4), 62-71. Leathwood, P. 1978. Influence of early undernutrition on behavioral development and learning in rodents. In “Studies on the Development of Behavior and the Nervous System. Vol. 4. Early Influences” ( G . Gottlieb, ed.), pp. 187-209. Academic Press, New York. Lewontin, R. C. 1974. “The Genetic Basis of Evolutionary Change.” Columbia Univ. Press, New York. Lewontin, R . C. 1978. Adaptation, Sci. Am. 239(3), 212-230. Lewontin, R. C. 1979. Sociobiology as an adaptationist program. Behuv. Sci. 24, 5-14. Lloyd, D. G . 1977. Genetic and phenotypic models of natural selection. J . Theor. Biol. 69, 543560. Logue, A. W. 1979. Taste aversion learning and the generality ofthe laws of learning. Psychol. Bull. 86, 276-296. LoLordo, V. M. 1979. Constraints on learning. In “Animal Learning: Survey and Analysis” (M. E. Bitterman, V; M. LoLordo, J. B. Overmier, and M. E. Rashotte, eds.), pp. 473-504. Plenum, New York. Lorenz, K . Z. 1965. “Evolution and Modification of Behavior.” Univ. of Chicago Press, Chicago, Illinois. Lorenz, K. 2. 1969. Innate bases of learning. In “On the Biology of Learning” (K. H. Pribram, ed.), pp. 13-93. Harcourt, New York. MacArthur, R.H., and Wilson, E. 0. 1967. “The Theory of Island Biogeography.” Princeton Univ. Press, Princeton, New Jersey. McGraw, M. 1946. Maturation of behavior. In “Manual of Child Psychology” (L. Carmichael. ed.), 1st ed., pp. 332-369. Wiley, New York. McKenna, J. J. 1979. The evolution of allomothering behavior among colobine monkeys: Function and opportunism in evolution. Am. Anrhropol. 81, 818-840. Marler, P. R . 1967. Comparative study of song development in sparrows. Proc. XIVth. Int. Ornithol. Congr. 1967, 23 1-244. Marler, P. 1970. A comparative approach to vocal development: Song learning in the white-crowned sparrow. J. Comp. Physiol. Psychol. Monogr. 71 (No. 2, Pt. 2). 1-25. Marler, P. 1975. On strategies of behavioral development. In “Function and Evolution in Behaviour: Essays in Honour of Professor Niko Tinbergen, FRS” (G. Baerends, C. Beer, and A. Manning, eds.), pp. 254-275. Clarendon, Oxford. Marler, P. 1976. Sensory templates in species-specific behavior. In “Simpler Networks: An Approach to Patterned Behavior” (J. Fentress, ed.), pp. 314-329. Sinauer, New York. Marler, P. 1977. Perception and innate knowledge. In “The Nature of Life” (W. H. Heidcamp, ed.), pp. 1 1 1-139. University Park Press, Baltimore, Maryland. Marler, P., and Mundinger, P. 1971. Vocal learning in birds. I n “The Ontogeny of Vertebrate Behavior” (H. Moltz, ed.), pp. 389-450. Academic Press, New York. Marler, P., and Peters, S . 1977. Selective vocal learning in a sparrow. Science 198, 519-521. Marquis, D. G. 1930. The criterion of innate behavior. Psychol. Rev. 37, 334-349. Mason, H. L.. and Langenheim, J. H. 1957. Language analysis and the concept environment. Ecology 38, 325-340. Mayr, E. 1974. Behavior programs and evolutionary strategies. Am. Scient. 62, 650-659. Miller, D. B., and Tallarico, R. 1974. On the correlation of brain size and problem-solving behavior of ring doves and pigeons. Brain Behav. Evol. 10, 265-273. Morgan, C. L. 1896. “Habit and Instinct.” Arnold, London.

104

TIMOTHY D. JOHNSTON

Mueller, H. C., and Berger, D. D. 1970. Prey preferences in the sharp-shinned hawk: The roles of sex, age, experience, and motivation. Auk 87, 452-457. Mulligan, J . A. 1966. Singing behavior and its development in the song sparrow. Melospizu melodia. Univ. Culv. Publ. 2001.81, 1-76. Nottebohm, F. 1970. Ontogeny of bird song. Science 167, 950-956. Nottebohm. F. 1972. The origins of vocal learning. A m . Nut. 106, 116-140. Oleson, J. 1971. Contralateral focal increase of cerebral blood flow in man during arm work. Bruin 94, 635-646. Oppenheim. R. W. 1980. Metamorphosis and adaptation in the behavior of developing organisms. Dev. Psychobiol. 13, 353-356. Oppenheim, R. W. 1981. Ontogenetic adaptations and retrogressive processes in the development of the nervous system and behaviour: A neuroembryological perspective. In "Maturation and Development: Biological and Psychological Perspectives" (K.J. Connolly and H. F. R. Prechtl. eds.), pp. 73-109. Lippincott, Philadelphia, Pennsylvania. Orians. G . H. 1969. Age and hunting success in the brown pelican (Pelecunus oc,cidentu/is). Anirn. Behuv. 17, 316-319. Partridge, L. 1976. Field and laboratory observations on the foraging and feeding techniques of blue tits (furus cueruleus) and coal tits (Purus uter) in relation to their habitats. Anim. Behav. 24, 534-544. Perrins, C. M. 1965. Population fluctuations and clutch size in the Great Tit, Purus major. J. Anim. Go/. 34, 601-647. Perrins, C. 1979. "British Tits." Collins, London. Peterson, R. H., and Prosser, C. L. 1972. The effects of cooling on electrical responses of goldfish (Curussinus uurutus) central nervous system. J . Comp. Biochem. Physiol. 42A, 1019-1037. Petrinovich, L. 1979. Probabilistic functionalism: A conception of research method. Am. Psycho/. 34, 373-390. Plotkin, H. C., and Odling-Smee, F. J . 1979. Learning, change, and evolution: An enquiry into the teleonomy of learning. In "Advances in the Study of Behavior" (J. S. Rosenblatt, R. A. Hinde, C. G. Beer, and M. C. Busnel, eds.), Vol. 10, pp. 1-41. Academic Press, New York. Pyke, G . H. 1978. Are animals efficient harvesters'? Anim. Behuv. 26, 241-250. Pyke. G. H.. Pulliam, H. R., and Charnov, E. L. 1977. Optimal foraging: A selective review of theories and facts. Q.Rev. Biol. 52, 137-154. Pyle, D. W . 1976. Oviposition site differences in strains of Drosophilu selected for divergent geotactic maze behavior. A m . Nut. 110, 181-184. Pyle. D. W. 1978. Correlated responses to selection for a behavioral trait in Drosophilu melunogusIer. Behuv. Genet. 8 , 333-340. Radinsky, L. 1978. Evolution of brain size in carnivores and ungulates. Am. Nut. 112, 815-831. Raichle, M. 1975. Senson-motor area increase of oxygen uptake and blood flow in the human brain during contralateral hand exercise: Preliminary observations by the 0-15 method. In "Brain Work. The Coupling of Function, Metabolism and Blood Flow in the Brain" (D. H. lngvar and N. A. Lassen, eds.), pp. 372-376. Munksgaard, Copenhagen. Razran, G. 1971. "Mind in Evolution." Houghton Mifflin, Boston, Massachusetts. Recher. H. F., and Recher, J . A. 1969. Comparative foraging efficiency of adult and immature little blue herons (Floridu cuerulu). Anim. Behuv.17, 320-322. Rensch, B. 1956. Increase in leaming capactiy with increase in brain size. Am. Nut. 90, 81-95. Rensch, B. 1959. Trends towards progress of brains and sense organs. Cold Spring Harbor Symp. Quunt. Biof.24, 291-303. Rescorla, R. A , , and Holland, P. C. 1976. Some behavioral approaches to the study of learning. In "Neural Mechanisms of Learning and Memory" (M. R. Rosenzweig and M . L. Bennett, eds.), pp. 165-192. MIT Press, Cambridge, Massachusetts.

LEARNING AND NATURAL SELECTION

105

Revusky. S. 1971. The role of interference in association over a delay. I n “Animal Memory” (W. K. Honig and P. H. R. James, eds.). pp. 155-213. Academic Press, New York. Revusky, S. 1977. Learning as a general process with an emphasis on data from feeding experiments. In “Food Aversion Learning” (N. W. Milgram, L. Krames. and T. M. Alloway, eds.). pp. 1-51. Plenum, New York. Richardson, C. C., Inman, R. B., and Kornberg, A. 1964. Enzymatic synthesis of deoxyribonucleic acid. XVIII. The repair of partially single-stranded DNA templates by DNA polymerase. J. Mol. B i d . 9, 46-69. Richdale. L. F. 1957. ”A Population Study of Penguins.” Oxford Univ. Press, Oxford. Ricklefs, R. E. 1974. Energetics of reproduction in birds. In “Avian Energetics” (R. A. Paynter, ed.). pp. 152-292. Nuttall Ornithological Club, Cambridge, Massachusetts. Riddell. W . I . , and Cod, K . G. 1977. Comparative investigation of the relationship between cerebral indices and learning abilities. Brain Behav. Evol. 14, 385-398. Romanes, G. J . 1884. “Mental Evolution in Animals.” Appleton, New York. Roots, B. I . , and Prosser, C. L. 1962. Temperature acclimation and the nervous system in fish. J. EXP.Bi01. 39, 617-629. Rosenzweig. M. R. 1966. Environmental complexity. cerebral change. and behavior. Am. Psvchol. 21, 321-332. Rozin, P. 1976. The evolution of intelligence and access to the cognitive unconscious. Progr. Psyrhobiol. Physiol. P s y h o l . 6 , 245-280. Rozin, P., and Kalat, J . W. I97 I . Specific hungers and poison avoidance as adaptive specializations of learning. P s y h o l . Rev. 78, 459-486. Sasviri. L. 1979. Observational learning in great, blue and marsh tits. Anim. Behav. 27, 767-771, Seligman, M . E. P. 1970. On the generality of the laws of learning. Psychol. Rev. 77, 406-418. Seligrnan, M. E. P., and Hager, J . L. (eds.) 1972. “Biological Boundaries of Learning.” PrenticeHall, Englewood Cliffs, New Jersey. Shettleworth. S. J . 1972. Constraints on learning. I n “Advances in the Study of Behavior” (D. S. Lehrman, R. A. Hinde, and E. Shaw, eds.), Vol. 4.. pp. 1-68. Academic Press, New York. Shettleworth. S. J . Function and mechanism in learning. In “Advances in Analysis of Behavior” ( M . Zeiler and P. Harzem. eds.), Vol. 3. Academic Press, New York. (In press.) Sidman, R. L., Green, M. C., and Appel, S. H. 1965. “Catalog of Neurological Mutants of the Mouse.” Harvard Univ. Press, Cambridge, Massachusetts. Simmons, K . E. L. 1970. Ecological determinants of breeding adaptations and social behaviour in two fish-eating birds. In “Social Behavior in Birds and Mammals” (J. H. Crook, ed.), pp. 37-77. Academic Press, London. Skinner, B. F. 1974. “About Behaviorism.” Knopf, New York. Slobodkin, L. B. 1968. Towards a predictive theory of evolution. In “Population Biology and Evolution” (R. C. Lewontin, ed.), pp. 187-205. Syracuse Univ. Press, Syracuse, New York. Slobodkin, L. B., and Rapoport, A. 1974. An optimum strategy of evolution. Q . Rev. Biol. 49, I8 1-200. Sluckin, W. 1973. “Imprinting and Early Learning. Aldine. Chicago, Illinois. Snow, D. W . 1958. The breeding of the Blackbird, Turclus merula, at Oxford. Ibis 100, 1-30. Sokoloff. L. 1960. Metabolism of the central nervous system in vivo. I n “Handbook of Physiology, Section I : Neurophysiology” (J. Field, H. W. Magoun, and V. E. Hall, eds.), Vol. 111, pp. 1843-1864. American Physiological Society. Washington, D. C. Somrnerhoff, G. 1950. “Analytical Biology.” Oxford Univ. Press, London and New York. Staddon, J . E. R. (ed.) 1980. “Limits to Action: The Allocation of Individual Behavior.” Academic Press, New York. Steams, S. C. 1977. The evolution of life history traits: A critique of the theory and a review of the data. Annu. Rev. Erol. S W . 8, 145-171. ”

106

TIMOTHY D. JOHNSTON

Stirling, I., and Latour, P. B. 1978. Comparative hunting abilities of polar bear cubs of different ages. Can. J . Zool. 56, 1768-1772. Tebecis, A. K . , and Phillips, J . W. 1968. Reflex response changes of the toad spinal cord to variations in temperature and pH. Comp. Biochem. Physiol. 25, 1035-1047. Thorndike, E. L. 191 I . “Animal Intelligence.” Hafner, New York. Thorpe, W. H. 1963. “Learning and Instinct in Animals.” Methuen, London. Tinbergen, L. 1960. The natural control of insects in pinewoods: 1. Factors influencing the intensity of predation by songbirds. Arch. Need. Zool. 13, 265-343. Tinbergen, N. 1951. “The Study of Instinct.” Oxford Univ. Press, London and New York. Tinbergen, N. 1963. On aims and methods of ethology. Z. Tierpsychol. 20, 410-429. Tinbergen, N., Broekhuysen, F.. Feekes, F., Houghton, C. W., Kruuk, H.. and Szulc, E. 1963. Egg shell removal by the blackheaded gull, Lurus ridibundus L: A behaviour component of camouflage. Behaviour 19, 74-1 17. Tolman, E. C. 1949. There is more than one kind of learning. Psychol. Rev. 56, 144-155. Tryon, R. C. 1940. Genetic differences in maze-learning ability in rats. In “Thirty-ninth Yearbook of the National Society for the Study of Education,” pp. I 1 1-1 19. Public School Publishing, Bloomington, Illinois. Tschirgi, R. D. 1960. Chemical environment of the central nervous system. In “Handbook of Physiology, Section I : Neurophysiology” (J. Field, H. W. Magoun, and V. E. Hall, eds.), Vol. 111, pp. 1865-1890. American Physiological Society, Washington, D. C. Verbeek, N. A. M. 1977. Comparative feeding behavior of immature and adult herring gulls. Wilson Bull. 89, 415-421, Wahlsten, D. 1972. Genetic experiments with animal learning: A critical review. Bchav. Biol. 7 , 143-182. Warren, J . M. 1965. Primate learning in comparative perspective. In “Behavior of Non-Human Primates” (A. M. Schrier, H. F. Harlow, and F. Stollnitz, eds.), Vol. 1 , pp. 249-281. Academic Press. New York. Watson, J. B. 1924. “Behaviorism.” Norton, New York. Watson, J. D. 1970. “The Molecular Biology of the Gene.” 2nd ed. Benjamin, New York. Welker, W. 1. 1976. Brain evolution in mammals: A review of concepts, problems, and methods. In “Evolution of Brain and Behavior in Vertebrates” (R. B. Masterton, M. E. Bitterman, C. B. G. Campbell, and N. Hotton, eds.), pp. 251-344. Erlbaum, Hillsdale, New Jersey. Wheeler, P. H. 1978. Elaborate CNS cooling structures in large dinosaurs. Nature (London) 275, 441-443. Williams, G . C. 1966. “Adaptation and Natural Selection.” Princeton Univ. Press, Princeton, New Jersey. Wooller, R. D., and Coulson, J. C. 1977. Factors affecting the age of first breeding of the kittiwake Rissa tridacryla. Ibis 119, 339-349. Wright, S. 1932. The roles of mutation, inbreeding, crossbreeding, and selection in evolution. Proc. X I In:. Congr. Genet. 1, 356-366. Yarczower, M.,and Hazlett, L. 1977. Evolutionary scales and anagenesis. Psycho/. Bull. 84, 1088- 1097.

ADVANCES M THE STUDY OF BEHAVIOR, VOL. 12

Visceral-Somatic Integration in Behavior, Cognition, and “Psychosomatic” Disease BARRY R. KOMISARUK INSTITUTE OF A N I M A L BEHAVIOR RUTGERS UNIVERSITY NEWARK, N E W JERSEY

I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

..............

11. Visceral-Somatic Relationships

111.

IV.

V. VI.

A. Somatovisceral Reflexes. . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Viscerosomatic Reflexes C. Neuroendocrine Reflexes . . . . . . . . . . . . Higher Order Integration of Visceral and So A. Integration via Coupling B. Neural Mechanisms of Visceral-Somatic Integration Underlying Cardiovascular, Respiratory, and Locomotor Activity . . . . . . . . . . . . . . . . C. Perceptual-Motor Aspects of Visceral-Somatic Integration: Responses to Exteroceptive Stimulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. Subjective Aspects of Visceral-Somatic Integration . . . . . . . . . . . . . . . . . . Visceral Activity and Ideational Imagery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Ideational Imagery Associated with the Cardiac Cycle . . . . . . B. Body Imagery and Metaphoric Language . . Toward a Concept of Psychogenic Organic (‘ Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . ........... References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

I.

I07 109

109 110 111

112 112 113

116 119 I23 123 124 129 134 I35

INTRODUCTION

The behavior of organisms is adapted to their internal as well as external environment. Their bkhavior is integrated with their own basic physiological processes (e.g., respiratory, cardiovascular, and hormonal). A major role in the integration of overt behavior with internal events is played by afferent neural activity of visceral origin, as exemplified in ingestion, excretion, copulation, and parturition. Even the timing of locomotion is coordinated with visceral (e.g., cardiac and respiratory) rhythms. In the present article, we focus on certain aspects of modulation of somatic behavior by visceral events, and the integration of visceral and somatic activity within the organism. Adequate integration of I07

Copyright @ 1982 by Acdcrnic Rcss. lnc. All rights of reproduction in MY form rcuncd. ISBN CL12-004512-5

108

BARRY R . KOMISARUK

these systems is essential to the organism’s adaptation to physical and social factors in its external environment. Traditionally, two bodily action systems have been recognized-the “voluntary” system which includes the striated muscles associated with the skeleton, and the “involuntary” system which includes the smooth muscle associated with the viscera. These two systems have also been referred to as “somatic” and “visceral,” respectively. The latter is under the control of the “autonomic” nervous system, so-called because its activity is relatively autonomous of “voluntary” control. Although these two systems can operate independently of each other, they,can also act in concert. This article analyzes some of the properties that emerge when the visceral and somatic systems operate in concert vs when they operate independently, and speculates on the implications for cognition and pathology. These properties are examined at several levels, starting with segmental spinal reflexes. Local reflexes demonstrate that somatic sensory stimulation can modify local visceral motor activity, and, conversely, that visceral sensory stimulation can modify somatic motor activity. The functional significance of these basic building blocks of somatic-visceral relationships is examined, e.g., in the coordination of internal and external body muscular movements during parturition. At a higher level of integration, such as in behavioral patterns that involve precise timing and coordination of the entire body, multiple visceral and somatic systems act in concert, e.g., when locomotor activity is coupled with cardiovascular and respiratory activity. The reflexive and central coordination of these systems is discussed and related to rhythmical activity of the brain. Following the discussion of visceral-somatic integration in terms of motor function, this integration is then considered in terms of sensory and cognitive function. Changes in sensory threshold and perceptual effects that occur in relation to visceral activity (e.g., phasic changes of ideational imagery in relation to the cardiac cycle) are reviewed. Evidence that visceral sensory activity exerts cognitive effects is presented, based upon dream content in relation to visceral disease, and literal metaphors derived from bodily sensations (e.g., “nausea, ” ‘‘catching one’s breath ”). Finally, coordination of visceral and somatic activity is examined in symbolic expression (e.g., change in blood flow to the limbs in association with inhibition against using the limbs). The speculation is made that certain forms of somatic or visceral psychogenic pathology may be understood in terms of the metaphoric “use” of specific bodily activity. The viscera may react in the absence of concomitant somatic action, and the viscera may, through their activity, provide afferent stimulation under conditions of suboptimal somatic sensory stimulation. According to this view, an understanding of pathological aspects of visceralsomatic relationships can be aided by an understanding of the principles of visceral-somatic coordination at various levels, from reflexive to cognitive.

VISCERAL-SOMATIC INTEGRATION

109

11. VISCERAL-SOMATIC RELATIONSHIPS

In this section we shall be concerned only with reflexes that pass between visceral and somatic systems, not those that occur only within either visceral or somatic systems. Thus, we will not consider visceral-visceral reflexes, such as peristalsis in response to gut distention, or somatic-somatic reflexes, such as the stretch reflex. We will consider somatic-visceral reflexes, i.e., those with somatic afferents and visceral efferents, and visceral-somatic reflexes, i.e., those with visceral afferents and somatic efferents. A.

SOMATOVISCERAL REFLEXES

Somatic stimulation in the form of maternal licking of the pups’ anogenital or rump skin elicits the visceral responses of urinary and defecatory emptying. In rats and guinea pigs, this stimulation is essential for the survival of the young, since their endogenous visceral emptying reflexes are not operational at birth (Beach, 1966; Harper, 1972; Boling et al., 1939). In adult rats, the bladderemptying response to tactile stimulation of the anogenital skin is considered to be a spinal reflex, on the basis that it can be elicited after mid-thoracic spinal transection in adult rats (Sato, 1975). Somatic stimulation may also elicit a vascular response. Thermal (cold) stimulation of the lower thoracic skin elicits ischemia in the duodenum and lower intestine, corresponding to the segmental level of the skin that is stimulated (Kuntz and Haselwood, 1940; Kuntz, 1945; Richins and Brizzee, 1949). A spinally mediated vascular response (dilatation) in response to somatic stimulation also occurs in the case of penile erection, resulting from stimulation of the penile skin in spinal-transected rats and dogs (see review, Hart, 1978). In spinal-transected dogs at least 30 days postoperative, the visceral responses of erection and ejaculation of seminal fluid can be elicited by mechanical stimulation of the penile shaft. Furthermore, detumescence can be elicited i n these dogs by stimulation of the glans penis. These responses are therefore somatovisceral reflexes that are organized at the spinal level. It is particularly interesting that the duration of penile engorgement is reduced by withdrawal of androgen treatment in spinal dogs, suggesting the possibility of spinal androgen sensitivity and/or peripheral androgenic effects on the target motor and sensory systems (Hart, 1978). Somatovisceral reflexes are mediated in part at the spinal level by the sympathetic and parasympathetic divisions of the autonomic nervous system. A role of the sympathetic division was shown in the rat by the observation that noxious electrical stimulation of the thoracic skin increased the heart rate after spinal transection at the cervical level, and noxious electrical stimulation of the abdominal skin decreased gastric motility after spinal cord destruction at levels T5-TI 1

110

BARRY R . KOMISARUK

(Sato, 1975). This noxious cutaneous stimulation increased the firing activity in the splanchnic (sympathetic) but not the vagus (parasympathetic) nerves, and the gastric response was abolished by destruction of the splanchnic nerves (Sato, 1975). The parasympathetic division also mediates somatovisceral reflexes at the spinal level: gentle tactile stimulation of the perineal skin induced urinary bladder contractions, such as those involved in the emptying response, and increased firing activity in the pelvic (parasympathetic) but not hypogastric (sympathetic) nerves which innervate the bladder, after spinal transection at C2 in the rat (Sato, 1975).

B.

VISCEROSOMATIC REFLEXES

Sensory stimulation originating in the viscera can modify somatic motor activity. This activity can be viewed in general form as postural adjustment to visceral distention, such as leg and trunk extension during voiding and parturition. Mechanical stimulation of the vagina, cervix, or rectum elicits leg extension, and blocks the leg-withdrawal response to noxious stimulation of the foot in rats in which the spinal cord has been transected at the mid-thoracic level (Komisaruk and Larsson, 1971). Electrical stimulation of the pelvic nerve, which provides afferent innervation of the urinary bladder, rectum, and genital tract (Komisaruk et al., 1972), induces combined relaxation of the muscles of the perineum and contraction of the striated muscles of the abdomen and diaphragm (Kuru, 1965). This adaptive “pelvico-perineo-abdominal reflex” is utilized in urination and defecation, and is also active during parturition. Additional examples of local viscerosomatic reflexes are represented in the following cases, although their function in natural contexts is not known. Compression of the fallopian tube induces contraction of the paravertebral striated muscles in rabbits (Eble, 1960). Distention of the uterine horn, urinary bladder, or rectum inhibits polysynaptic reflexes of the hind limb in cats (Evans and McPherson, 1958). Polysynaptic reflex responses of the hind leg are suppressed during spontaneous contractions of the urinary bladder (McPherson, 1966). Stimulation of the pelvic or splanchnic nerves, or sympathetic chain inhibits polysynaptic reflexes of the hind leg (Evans and McPherson, 1960). Pathologically, somatic muscular spasm occurs in response to visceral irritation, at the same segmental levels, e.g., in spasm of the lower abdominal muscles occurring in appendicitis, or of the thoracic muscles in tuberculosis (Ruch, 1961). In addition to the local segmental responses described above, multisegmental somatic responses to visceral afferent stimulation also occur. One example is the lordosis mating stance of the female rat, which involves elevation of the head and rump and depression of the lumbar area. In estrous females, this posture is elicited by mechanical stimulation of the skin of the flanks and perigenital re-

VISCERAL-SOMATIC INTEGRATION

111

gion, but it fails to occur to this stimulation in nonestrogenized females (Komisaruk and Diakow, 1973; Kow ef uf., 1979). Vaginal or rectal mechanical stimulation added to the skin stimulation increases the lordosis intensity in estrogenized females. This visceral stimulation can facilitate lordosis even in nonestrogenized females. Vaginal or rectal stimulation without skin stimulation does not elicit lordosis (Komisaruk and Diakow, 1973). It is not known which sensory endings mediate this response. However, the vaginal epithelium is at least in part of entodermal origin (although the anal epithelium is of ectodermal origin) (Arey, 1954), and the stretch receptors of the vaginal and rectal smooth muscle are probably activated by the adequate stimulus; hence, this reflex is provisionally considered to be visceral-somatic rather than somatic-somatic. A long-lasting somatic effect of this visceral stimulus has been described. In ovariectomized rats in which perigenital skin stimulation initially does not elicit lordosis, brief (2-sec) vaginal stimulation enables subsequent perigenital skin stimulation to elicit lordosis, an effect that begins immediately and persists for 2-3 hr after the single application of the brief vaginal stimulation (RodriguezSierra ef a l . , 1975). Other examples of multisegmental somatic effects of visceral stimulation are as follows. A posture in neonatal rats which strongly resembles lordosis is elicited by milk entering the throat as pups suckle (Drewett er al., 1974; Hall and Rosenblatt, 1977; Martin and Alberts, 1979). Characteristic extensor postures occur during defecation and urination in cats (Hess, 1954) and parturition in Anfilocupru (Muller-Schwarze, 1974), probably in response to visceral distention. And in pigeons, an effect has been described that is dependent on afferent activity generated in spinal afferents by movement of the viscera as the axis of the body shifts (Delius and Vollrath, 1973). In response to passive body tilt or rotation, pigeons show postural compensatory reflexes of the wings, tail, and neck even after labyrinthectomy (Biederman-Thorson and Thorson, 1973). The former authors suggest that this visceral afferent postural control mechanism provides an initial rapid, albeit gross, compensation before the slower vestibular reflexes come into play to provide the fine correction for passive body rotations. Diakow (1978) has shown that intraabdominal pressure affects behavior in female frogs. If a nonreceptive female is clasped by a male, she emits a “release call,” but this call is inhibited in receptive females, in which there is a developed egg mass, thereby facilitating oviposition and spawning. Intraabdominal distention produced normally by vasotocin-induced fluid accumulation, or artificially by injection of fluid or inflation of an implanted balloon, inhibits this release call. C.

NEUROENDOCRINE REFLEXES

There is another class of smooth (nonskeletal) muscle response to afferent somatic stimulation-it comprises responses that are mediated by the endocrine

112

BARRY R. KOMISARUK

system. The milk-ejection reflex is a somatovisceral reflex in which the effector is activated indirectly via the systemic circulation, rather than directly via efferent neurons. In this reflex, in response to suckling stimulation, the posterior lobe of the pituitary gland releases oxytocin into the systemic circulation (Cross, 1959). The oxytocin stimulates contractions of the smooth muscle myoepithelium of the mammary gland, thereby ejecting the milk into the mammary ducts from which it is expelled into the nursing young (Zaks, 1958). A related example is stimulation of uterine smooth muscle by oxytocin released reflexively in response to nuzzling of the genital skin in cattle (Vandemark and Hays, 1952). (For further discussion, see Komisaruk et al., 1981.) A similar type of somatovisceral response mediated by the neuroendocrine system is the increase in cardiac output and rate in response to epinephrine and norepinephrine which are released into the systemic circulation from the adrenal medulla following noxious somatic stimulation (see Koizumi and Brooks, 1974, for review).

111.

A.

HIGHERORDERINTEGRATION OF VISCERAL AND SOMATIC ACTIVITY

INTEGRATION

VIA

COUPLING OF ACTIVITY CYCLES

Higher order integration of visceral and somatic activity is exemplified by the coupling among locomotor, respiratory, and cardiac activity cycles. In this section, the occurrence, possible underlying mechanisms, and possible adaptive significance of such integration are presented. Bechbache and Duffin (1977) showed that during exercise in humans, the respiration cycle becomes entrained to the locomotion cycle. Volunteers were instructed to pedal on a bicycle ergometer, timing their leg movements to a metronome at 50 cycles per minute. When their respiration rate was recorded, it was found that respiration rate was entrained to the pedaling rhythm in 53% of 15 individuals. That is, they showed one respiratory cycle per pedal revolution. When the individuals were instructed to run on a motorized treadmill at a comfortable speed, the proportion showing entrainment of respiration to leg movement was even greater (80% of 15 individuals). The oxygen uptake in 6 of 8 subjects who showed respiratory-locomotor entrainment on the bicycle ergometer was lower (i.e., oxygen was used more efficiently) than predicted independently of entrainment. Therefore, a natural tendency exists to couple respiration with the locomotor cycle; this increases the efficiency of oxygen utilization (Bechbache and Duffin, 1977), which in turn increases stamina in locomotion. This appears to be an expression of a biological “Conservation of Energy Law” (Ralston, 1976) which states: “In freely chosen rate of activity, a rate is chosen

VISCERAL-SOMATIC INTEGRATION

113

that represents minimal energy expenditure per unit task.” Ralston’s law is based on his findings that a person in a natural walk tends to adopt a speed close to that of minimal energy expenditure per unit distance walked. These findings taken together suggest that a locomotor speed is most comfortable and energy efficient when the timing of the locomotion cycle coincides with that of the respiratory cycle. It would be of interest to determine the range of locomotor and respiratory frequencies over which this relationship holds. The cardiac cycle can become entrained to the respiration-locomotion cycle. Pessenhofer and Kenner (1975) showed that there is a marked tendency for individual respiratory movements to start at a constant phase of the cardiac cycle in humans. The evolutionary origin of respiratory-cardiac entrainment may be found in a primitive vertebrate-the dogfish shark, in which the timing of the heart beat is synchronized with the gulping contraction of the jaw muscles, which send surges of oxygenated water across the gill surfaces. There is usually one respiratory beat for each heartbeat (Satchell, 1968). The significance of this process is that as the oxygenated water surges across the gill surfaces, the blood surges through the gills, thereby optimizing the exchange of gases. One function of the coupling of cardiac, respiratory, and locomotor cycles may be to enable optimal efficiency of energy exchange among these systems, thereby maximizing stamina and strength. Furthermore, since such coupling exists in sharks and humans, it is likely to be phylogenetically widespread.

B.

N E U R A LMECHANISMS OF VISCERAL-SOMATIC INTEGRATION UNDERLYING CARDIOVASCULAR, RESPIRATORY, AND LOCOMOTOR ACTIVITY

I.

Rejlexive Mechanisms

Let us examine some of the ways in which visceral-somatic integration occurs in the cardiac, respiratory, and locomotor systems. Spinal motor neurons (at the lumbosacral level) fire in a phase-coupled relationship with respiration in decorticate cats (Frankstein et al., 1974), demonstrating a fundamental linkage between locomotion and respiration. With each respiratory cycle, bursts of neuronal discharges are recorded in the sympathetic nerve bundles innervating the leg in humans (Hagbarth and Vallbo, 1968) and also in the splanchnic nerves in cats (Preiss et al., 1975; Gootman et al., 1975). Furthermore, the timing of blood pressure rises and falls coincides with that of the respiratory cycle in lightly anesthetized cats (Borgdorff, 1975). This coupling could provide rhythmical modulation of blood flow to the limbs. Particularly when the locomotion cycle is coupled with the respiratory cycle (Bechbache and Duffin, 1977), it could maximize the efficiency of delivery of blood to the muscles of locomotion.

114

BARRY R. KOMISARUK

Coupling of these visceral and somatic activity cycles depends, at least in part, upon reflexive peripheral feedback mechanisms. Several reflexive mechanisms which utilize musculoskeletal body movements (Shepherd and Vanhoutte, 1979) are involved in cardiac-respiratory-locomotor entrainment. First, in the “respiratory sinus arrhythmia” reflex, deep inspiration stimulates sympathetic fibers and inhibits vagal inhibitory fibers to the heart, thereby quickening the heartbeat and coupling its rhythm to that of the respiratory cycle (Porges el al., 1980). Second, deep inspiration also produces a negative pressure in the thoracic cavity relative to the peripheral circulation. This suction forces the blood back to the heart in a surge, thereby abruptly stimulating the mechanoreceptors at the venal-atrial junction, and quickening the heartbeat via the cardioacceleratory system in the medulla. This respiration-linked cyclical heart stimulation could also help to entrain the cardiac cycle to the respiratory cycle. Third, a “muscle pump” effect exists in which rhythmical muscular contractions occurring in locomotion force blood back into the venous system, thereby rhythmically activating the mechanoreceptors at the venal-atrial junction. These reflexes phasically increase heart stroke volume in response to returning blood flow (Shepherd and Vanhoutte, 1979), which surges in time with rhythmical locomotor and respiratory movements. They provide an entraining rhythm to the heart, which adjusts its rate, timing, and stroke volume to the changing demand of the peripheral musculature. Another component of the neuromuscular system which regulates cyclical blood flow to the skeletal muscles is afferent activity generated by muscular contraction, such as that which could occur during walking. Coote (1975) found that, in cats, muscular contraction (generated by electrical stimulation of the lumbosacral ventral roots) was followed by increases in arterial blood pressure, heart rate, and pulmonary ventilation, but these effects were abolished when the muscle afferent activity was blocked by cutting the dorsal roots or by blocking muscle contraction with gallamine. Rhythmical muscular contraction utilizing this reflex in locomotion would therefore tend to entrain these systems to its rhythm. The reflexive relationships among the systems reviewed above facilitate the phase-coupling of their activity cycles. This may serve to optimize the efficiency of energy exchange among the systems, thereby optimizing the organism’s strength and stamina. This optimal condition may represent the simplest relationship among the cardiovascular, respiratory, and locomotor systems. That is, in the state of synchrony among these activity cycles, the variability in their timing is minimized, i.e., their relationship is least differentiated, and this may be the most “primitive” and efficient way in which they operate. This may be virtually the only way in which the most primitive vertebrates (e.g., shark) function (hence, by necessity, the most efficient means of functioning), and also the most efficient means of functioning in higher vertebrates.

VISCERAL-SOMATIC INTEGRATION

2.

115

Central Mechanisms: The Role of Brain Rhythms in Visceral-Somatic Coordination

Coordination of respiratory, cardiovascular, and motor systems is also based upon the phasic entrainment of their activity cycles to delta-theta, and alpha EEG rhythms of the brain. The following studies describe phasic correlations between brain rhythms and bodily activity rhythms. Langhorst et al. (1975) found that medullary reticular formation neurons that respond to stimulation of the aortic baroceptors, by rapid phasic increases in blood pressure, fire in phasic relation with the theta-delta rhythm in chloralose-urethane anesthetized cats. Similarly, during exploratory sniffing behavior in rats, cardiac and respiratory cycles (including muscular contraction cycles of the face, neck, and thorax), and respiratory cycles in hamsters (Macrides, 1975). are coupled with theta waves recorded as hippocampal andor hypothalamic EEG, at a frequency centered near 7 Hz (Komisaruk, 1970, 1977). In humans, the cardiac cycle can also become phase-coupled with the alpha rhythm (Callaway and Buchsbaum, 1965; Birren et al., 1963; Buchsbaum and Callaway, 1965; Callaway and Layne, 1964; Callaway, 1965). Therefore, the three major rhythmical brain systems (lower brain stem: delta; limbic system: theta; thalamocortical: alpha) can be phase-coupled to the cardiac cycle. We can speculate that different kinds of motor activities are entrained to brain rhythms of different rates. Perhaps rapid-movement rhythmical muscular systems are coupled primarily with the high frequency brain systems (e.g., thalamoneocortical alpha rhythm; 8- 12 Hz, related to physiological finger tremor (Jasper and Andrews, 1938) and eye saccades (Gaardner et a l . , 1966), whereas slower-movement, rhythmical muscular systems (respiratory, cardiovascular, locomotor) are coupled primarily with lower frequency brain rhythms [e.g., limbic system theta rhythm (3-8 Hz) and lower brain stem delta rhythm (less than 1-3 Hz)].One form of integration among the systems that could occur is that in which a rapid movement is timed to occur at an optimal moment during a slower movement (e.g., the respiratory, hand, and arm movements involved in throwing a curve ball or cracking a whip). It is tempting to speculate that when brain rhythms overlap in frequency (e.g., the low end of the alpha rhythm range and the upper end of the theta rhythm range overlap at 8 Hz) they share coupling with the neuromuscular systems that may ordinarily be coupled with only one of the brain rhythms when it is active at a different rate. This may be involved in the establishment of fine control over large-mass systems, e.g., eye-finger precision of movement brought to bear on trunk and leg orientation as in dancing and athletics. Thus far we have focused on the performance aspects of visceral-somatic integration, noting that respiration, cardiovascular, and locomotor activities are interrelated at spinal and supraspinal levels of the central nervous system. This

I I6

BARRY R. KOMISARUK

integration is based in part upon afferent stimuli which provide feedback information that coordinates the activity. The temporal patterning of this activity suggests that somatic movement is adapted to, and integrated with, the cardiovascular and respiratory supply systems and with the visceral components of the movements, as in mating, parturition, defecation, and urination. The rates and rhythms of movements, the segmental level of a movement, the muscles employed in the movement, and the duration of the movement may be understood in terms of the underlying visceral process with which they are neurally integrated, and to which they are functionally related. Evidence in humans will be presented that we sense the various kinds of somatic-visceral integrations, assess their "quality," and utilize this perception in our behavior and language.

c.

PERCEPTUAL-MOTOR ASPECTS OF VISCERAL-SOMATIC INTEGRATION: RESPONSESTO EXTEROCEPTIVE STIMULATION

Visceral-somatic integration operates in two directions: in one direction it incorporates internal sources of stimulation, as discussed earlier. In the other, discussed in the present section, it deals with exteroceptive stimulation and adapts the organism's behavior to its external environment. Over the past decade it has become clear that sensory input is regulated by brain processes that exhibit wave-forms. Thus continuous sensory stimulation at the periphery may be "gated" to affect behavior only at certain times, in relation to the brain wave activity with which the stimulation interacts (Komisaruk, 1977). We have seen that these wave patterns of neural activity are related to visceral-somatic integration. Let us now examine ways in which these may be involved in exteroceptive perception and in the timing of motor responses. In the following section, evidence is reviewed that when visceral and somatic activity cycles become coupled with each other, they may jointly or singly become entrained to brain rhythms; consequently, perceptual sensitivity and motor activity fluctuate in time with these rhythms. When coupling of these systems occurs, it is likely that unique perceptual and motor properties emerge. The existence of entrainment of neurornuscular excitation cycles to brain rhythms implies that the activity (e.g., excitation level, responsivity) of the neuromuscular systems and their associated sensory input does not remain constant, but instead varies over time. There is extensive evidence that brain rhythms do in fact represent fluctuating levels of sensory and neuromuscular excitability. Evidence for this has been presented in an analysis of sniffing in the rat. In the rhythmical oscillatory pattern of the rat's exploratory sniffing behavior, the vibrissae are whisked to and fro in synchrony with discrete rhythmical sniffs, head-neck-thorax movements, and individual theta waves. This sequence can be viewed as a chain of individual cycles during each of which the state of excitation of limbic system neurons and the associated neuromuscular system varies cycli-

VISCERAL-SOMATIC INTEGRATION

I17

cally over time. This suggests that there may exist specifiable phases of the cycle during which perceptual processes associated with the movements can be performed optimally. I n other words, the underlying neural system may perform different operations (i.e., sensory and motor) at specific phases of the cycle (Komisaruk, 1977). It was these considerations that led us to predict that in rats a self-initiated act (e.g., a lever press which delivers a food pellet) would be initiated nonrandomly in time and in phasic relation to an individual theta cycle. In a test of this hypothesis (Semba and Komisaruk, 1978) we found that, indeed, bar presses were most likely to occur at wave peaks during the theta cycle (recorded as hippocampal EEG), whereas bar releases were most likely to occur at theta-wave valleys. The difference in the behavioral movements involved in pressing vs releasing the bar suggests that the flexor and extensor forelimb systems are preferentially activated at different phases of the theta cycle. A “gating” effect related to the theta cycle was also observed by Buzsaki et a / . (1982) who found that the amplitude of evoked potentials recorded in the dentate gyrus of the hippocampus, which were generated by electrical stimulation of the perforant path, was greater when the stimulus was delivered during the negative-going phase of the theta wave cycle than during the positive-going phase. It is not known whether this effect is more closely related to sensory or motor systems. The theta cycle is not unique in its relationship to sensorimotor excitability cycles; the higher frequency alpha cycle and lower frequency cardiac cycle have been shown to be related phasically to cycles of visual and auditory sensitivity and reaction time. For example, the subjective intensity of a constant-intensity light flash varied as a function of the phase of the alpha wave during which the flash was presented (Callaway, 1965). Using a different approach, Harter and White (1967) showed that there was a marked tendency for a light flashing at 33.3 times per second to be perceived as flashing only as many times as the cortical alpha wave (10 Hz)peaked during the flash train. That is, when a train of three flashes occurred during which only one alpha wave peak occurred, the subjects reported that only one flash had occurred. When a train of six flashes was presented, only two flashes were reported (only two alpha peaks had occurred). Thus, the alpha wave apparently modulates or “gates” the visual input cyclically. In related experiments, the galvanic skin response was measured when the word “danger” was presented, and the magnitude of the response varied as a function of the phase of the alpha-rhythm cycle during which the word was presented (Nunn and Osselton, 1974). The authors concluded: “it seems likely that the alpha rhythm is indeed a correlate of the activity of a ‘neuronic shutter’ which periodically prevents the reception or processing of visual information by the cortex.” Similarly, the probability of eye opening in response to an auditory

1 I8

BARRY R. KOMISARUK

stimulus varied in relation to the phase of the alpha rhythm (Boreham et al., 1949). The amplitude of the visual evoked cortical response has been shown to vary as a function of the phase of the alpha rhythm at which the light flash occurs (Callaway and Layne, 1964; Remond and Lesevre, 1967). Muscular movements have also been shown to occur in phasic relation to the alpha wave. Thus, Jasper and Andrews (1938) found that normal physiological tremor of the fingers is at times synchronized with the alpha rhythm recorded from the motor cortex. In rats, tremor of the vibrissae and jaw at a frequency of approximately 10 Hz occurs normally when the rats are crouched and showing no gross motor activity. These movements are precisely coupled with individual waves of a characteristic neocortical alpha rhythm and with bursts of thalamic neuronal activity (Semba et al., 1980). The tremor activity can in most cases be abolished by surgical ablation of the contralateral sensorimotor cortex (Semba and Komisaruk, submitted for publication). In humans, saccadic eye fixations are phase-locked with the alpha wave cycle (Gaardner et a l . , 1966). These authors suggested that “each saccade results in a packet of information being presented for storage. . . a single alpha component cycle may reflect a basic unit of storage.” This is of particular interest when considered in conjunction with the findings of Lansing (1957), who showed that reaction time measured as a finger response to a flash of light presented at various phases of the alpha-wave cycle was shortest at a particular phase of the cycle. That is, the neuromuscular actions involving eye fixation, visual perception, and behavioral response may be performed in basic units, rather than continuously, like individual ‘‘stills’’ of a motion picture. Sensorimotor modulation occurs in phasic relation to cardiac cycles as well as brain rhythms. Reaction times are longer during the heart systole than during the diastole (Birren ef al., 1963; Callaway and Layne, 1964). This was suggested as being due to response inhibition generated by the pulsatile increase in blood pressure during systole, since a similar slowing of reaction time during systole was reported in patients with transistorized pacemakers (Callaway and Layne, 1964). As an example of another visceral-somatic pacing mechanism, Pessenhoffer and Kenner (1975) showed that there is a marked tendency for inspiratory movements to start at a constant phase of the cardiac cycle. An analysis of this phenomenon, described later, indicates that this effect may be related to motor inhibition induced by increased afferent activity from the baroceptors and lung afferents. In dogs, Heymans (cited in Liljestrand, 1965) separated the circulation of the head from that of the trunk, but left the vagus and aortic nerves intact. Administration of adrenaline to the trunk increased the blood pressure, and inhibited facial motor activity that was associated with respiratory movements. In addition, expansion of the lungs inhibited facial respiratory movements. Both inhibitory effects were abolished by cutting the vagus and aortic nerves. These

VISCERAL-SOMATIC INTEGRATION

119

studies demonstrate that afferent impulses from stretch receptors in the lungs and chemo- and/or baroceptors in the cardioaortic vascular area can inhibit motor activity of the face. The studies described above suggest a neurological basis for the coupled activity of cardiovascular, respiratory, autonomic, and musculoskeletal systems. This mode of activity can be viewed as a relatively undifferentiated and, hence, primitive condition. When it occurs, certain properties emerge, which are probably perceived as “peak experiences.” Lowen has expressed some of the properties of such integrated visceral-somatic activity as follows: The body’s involuntary movements are the essence of its life. The beat of the heart, the cycle of respiration, the peristaltic movements of the intestines-all are involuntary actions. But even on the total body level, these involuntary movements are the most meaningful! We convluse with laughter, cry for pain or sorrow, tremble with anger, jump for joy, leap with excitement and smile with pleasure. Because these are spontaneous, unwilled or involuntary actions, they move us in a deep, meaningful way. And most fulfilling, most satisfying and most meaningful of these involuntary responses is the orgasm in which the pelvis moves spontaneously and the whole body convulses with ecstasy of release (Lowen, 1976, p. 244).

The following section examines some other cognitive aspects of integrated visceral-somatic activity.

D. SUBJECTIVE ASPECTSOF VISCERAL-SOMATIC INTEGRATION Not all of the manifestations of visceral-somatic integration have been measured objectively as yet, but humans are apparently finely tuned subjectively to sensations of comfort and discomfort during locomotion, to the sensations which accompany great effort or relative ease in movement, to the pressures arising from contact between internal organs and from blood engorgement of the extremities, as well as to many other internally based sensations. Humans make use of these sensations to adjust their movements, their posture, and their effort to minimize discomfort, and also to maximize the force of their actions when this is required. In this section I shall discuss the visceral-somatic integrations in humans that underlie the subjective sensations of comfort and discomfort, forcefulness or lack of force, ease or difficulty in movement, and the sense of harmony or its absence during activity. During locomotion, the thorax, shoulders, and arm form a subsystem whose components are jointed mechanically such that swinging the arms outward and upward forces the rib cage to increase its internal volume. Conversely, during locomotion, expanding the thorax from within can be felt mechanically to lift the shoulders and arms outward and upward, against gravity. The activity rhythm of this subsystem is synchronized with that of the legs in comfortable, energy-

120

B A R R Y R . KOMISARUK

efficient locomotion (Bechbache and Duffin, 1977). This is a form of structural resonance: the arms help the thorax to move, that inertia helps the legs to swing, and inertia of the legs in turn helps the arms to swing. Thus, when the limbs and thorax are in resonance, whole-body forward momentum is increased, thereby mechanically increasing the speed of locomotion. When running at a comfortable gait, the lung expansion-contraction cycle can be adjusted to the limb cycle and this respiration cycle duration can be adjusted, in turn, to the heartbeat cycle. When such coupling is achieved, a minimum of pressure is felt at the boundary between the expanding-contracting heart and lungs (which are contiguous with each other) within the thoracic cavity. If the coupled activity is suddenly disrupted, physical clashing is felt at the boundaries between lungs and heart, which is reduced again if coupling is reestablished. During locomotion, as the arms move back and forth, pulse pressure, felt at the fingertips, is increased by the mechanical pumping action of the arms casting the blood toward the periphery. If the phases of arm movement and heartbeat are adjusted such that this pulse pressure is minimized in the fingertips, a feeling of ease of arm movement results. In other words, when the forward thrust of the arm is timed to coincide with the time of the surge of blood to the fingertips, there is a feeling of minimal pulse pressure at the fingertips and, consequently, minimal effort of arm motion. Coleman (1921) made the following interesting observation: "One who always became breathless when halfway up a hill felt his pulse and began the climb breathing and stepping in unison with the pulse and climbed the hill without breathlessness, and the rise in blood pressure was only half as great. A one-to-one correspondence between heartbeats and leg kicks has also been described in diving tufted ducks during the underwater descent and ascent phases of the dive (P. Butler and T. Woakes, cited in MacDonald and Amlaner, 198 1). This is likely to be a situation in which energy utilization is of necessity highly efficient, and thereby indicates the adaptive significance of this coordination pattern. Schlant (1978) noted the mechanical effect of musculoskeletal activity on cardiovascular activity: the lungs expand to fill the increased volume of the thorax as it expands while the diaphragm contracts synchronously. Inspiration and expansion of the lungs exerts a compressive force against the heart, which is surrounded by, and in contact with, the lungs. Hence, there is a mechanical advantage of having limb, trunk, respiration, and cardiovascular cycles the same length and phase-coupled. This pumping action varies the filling rate of the heart and the blood pressure, thereby cyclically modulating neural reflex activity via mechano- and baroceptors, thus influencing atrial and ventricular contractility and vasomotor tone via the autonomic nervous system (Schlant, 1978; Shepherd and Vanhoutte, 1979). We may speculate that the sense of ease of movement corresponds, therefore, to a state of synchrony among the activity cycles of the respiratory, cardiovascular, and locomotor systems. "

VISCERAL-SOMATIC INTEGRATION

I.

121

Linguistic, Ititerpersotrul, atid Aesthetic Expressions of Visceral-Sotnatic Rhythmic Integration

Visceral-somatic sensations and their rhythms permeate our linguistic, interpersonal, and aesthetic expressions as the reference for temporally ordering our activities. In music, the sense of timing (i.e.. “tactus,” literally “time beating”) was originally based on the natural rhythm of the pulse, rather than an absolute external standard (Sachs, 1948). The interval between pulse beats during quiet respiration was used to define the duration of the semibreve (i.e., whole note). Natural rhythms of movement occur in a variety of contexts in humans in relation to biological rhythms. Lourie (1949) claims that in a pediatric clinic population, 1~20% of 130 children showed rhythmical body movements. Furthermore, “In the great majority of the children who rock, roll, bang, or sway the pacemaker is the heartbeat .... In a minority of the children in this series the pacemaker is the breathing rate” (p. 657). Byers (1979), on the basis of cinegraphic analysis, has found that people in small groups perform gestural movements toward each other in what appears to be a common rhythm, the intermovement interval of which is approximately 0 . 4 sec. He states, “This duration of 0.4 sec is the same (recognizing that ‘same’ embraces a small range of possible variation) as that of the military march rate, is half that of the accepted heart rate at rest, and is equivalent to four cycles of the familiar “alpha” rhythm of encephalography. ” Byers (1977) further points out “some form of this synchronizing process is used everywhere in the world. It is familiar in church ceremonies, cheerleading at sports events, rock concerts, dances. ” Perhaps the cycles of systole-diastole, adduction-abduction, and inspirationexpiration provide a body-based sensory input which underlies our concepts of the beginning, duration, and end of movements, i.e., phrasing. The modulation of expenditure of effort in relation to time during each such movement cycle or phrase provides “expression” in the musical sense (Byers. 1977, 1979). With regard to expression in the motoric sense, Clynes (1979) points out An expressive movement is an entity in time; it has a beginning, middle, and end. The first step in studying such entities is to realize that it takes a certain amount of time to execute an expression o f joy, of anger, of sadness, o f love and so on, and these times differ for different emotions.. .we have called these elemental chunks of entities o f expression essrnfi~.,~,rrns.. .the expressive quality is recognized from the dynamic contour of the motion. rather than the particular part o f the body used.

Clynes (1979) has shown that when human subjects press a pressure transducer on hearing a click, expressing the quality of anger, hate, love, etc. as precisely as they can with single, expressive pressure actions, the change in pressure over time generates, on a polygraph, “sentic forms” which differ markedly from each other in characteristic ways. For example, the sentic form for “hate” has a more

I22

BARRY R. KOMISARUK

abrupt onset than that for “love.” Biological movement cycles of exertion and relaxation (e.g., cardiac, respiratory, locomotor) may therefore provide the basic building blocks out of which more complex behavior patterns (e.g., vocalization, gesturing) are differentiated, the more primitive biological rhythmicity nevertheless being retained.

2. Physiological Consequences of Rhythmic Visceral-Somatic Activity Rhythmical repetition of somatic sensory stimulation in the genital system leads to an increase in the intensity of visceral and somatic muscular contraction leading to an efferent excitation peak, i.e., orgasm. In mammalian male orgasm, there is a sudden surpassing of the contraction threshold of the visceral muscles of the seminal vesicles and prostate, thereby producing ejaculation (Monnier , 1968). Perhaps it is the intense but nonaversive sensory input generated by synchronous, peak visceral and somatic muscular contraction in genital orgasm that is perceived as intensely pleasurable. The “organ pleasure” that occurs in the performance of sucking and retention of excreta in addition to genital stimulation (Freud, 1924) may be based at least in part upon this property of afferent visceral-somatic excitation. The orgasmic process as described above need not be restricted to the genital system. Indeed, sensations with orgasmic qualities have been described to occur during breathing (Scott, 1948), crying (Hite, 1976), and vomiting (Dodson, 1974). A sneeze may be described as a respiratory orgasm. Masters and Johnson (1979) claim that “the total body is a potentially erotic organ ... There can be back-of-the-neck orgasm, bottom-of-the-foot orgasm and palm-of-the-hand orgasms” (p. 1 10). Genital orgasmic response has been described in response to breast stimulation alone (Masters and Johnson, 1979), and to other forms of nongenital stimulation (Hollander, 1981). A similar statement was made in popular literature: “It is possible to generate an orgasm at any spot on the human body” (Alther, 1974, p. 43). We may speculate that orgasm is generated by a process of increasingly synchronous afferent discharge (at least in part generated by muscular contraction) which generates a peak of sensory excitation. Synchrony of discharge among different systems implies undifferentiated activity, i.e., unity among otherwise differentiated elements. Hypersynchrony is characteristic of epileptiform seizure discharges in the EEG, and reports of pleasurable sensations occurring during epileptic seizures, sometimes bordering on orgasm, have been described (Myslobodsky, 1976). The following analogy may be considered to be a model of orgasm which depicts the relationships among rhythmicity, synchronicity, and intensity of excitation. At times, in a noisy sports stadium, some individuals start clapping in a slow steady rhythm, then others join them (i.e., are “recruited”) in the same

VISCERAL-SOMATIC INTEGRATION

123

rhythm, and there develops a steadily increasing contrast between peaks and valleys of sound intensity. Thus, a relatively undifferentiated (i.e., simple, “primitive”) synchronized pattern of alternating bursts of sounds and silences emerges out of a more differentiated (i.e., complex) continuous din, reaching a climax of intensity. In orgasm, perhaps the activity cycles of somatic and visceral systems are entrained by rhythmical stimulation, peak efferent activity generates undifferentiated peak afferent stimulation, and at a cognitive level this is described as intensely pleasurable.

Iv.

VISCERAL ACTIVITY A N D IDEATIONAL IMAGERY

The findings we have reviewed concerning the integration of cardiovascular activity, respiration, and somatic motor activity have subjective counterparts in feelings of well-being, comfort, ease of movement, or the opposites of these. Highly specific ideational imagery has also been reported to be related to visceral activity. A.

WITH THE CARDIAC CYCLE IDEATIONAL IMAGERY ASSOCIATED

An individual using marijuana has been described (see Komisaruk, 1977) in whom visceral, visual, and somatic imagery fluctuated in synchrony with the heartbeat. The individual described an experience that was as if one were suddenly thrown high into the sky, soaring like a bird, looking down, feeling warm, light, yellow, free, limbs extended, strong, good, and hearing a highpitched ringing hum; then suddenly being plunged deep down into a small, cramped, blue-black dark hole, hearing a low-pitched hum, feeling cold, terrified, and minuscule in size and importance. Then the upsurge occurred again, followed by the plunge; up and down, over and over again in a steady rhythm which was precisely that of the heartbeat. This kind of “synesthesia,” e.g., between a sense of bodily motion and visual image, may be based in part upon specific neuronal circuitry. For example, Daunton and Thomsen (1976) have shown that in cats suspended in a harness, neurons in the vestibular nucleus that respond to linear acceleration of the cat in a given direction also respond to a visual stimulus which simulates actual movement of the cat in the same direction: a cell excited by movement of the cat to the left was also excited by movement of the visual stimulus to the right. Perceptions similar to those described earlier were reported by Castaneda under the influence of peyote. At moments everything was so clear it seemed to be early morning or dusk. Then it would get dark; then it would clear again. Soon I realized that the brightness corresponded to my heart’s diastole, and the darkness to its systole. The world changed from bright to dark to

124

BARRY R. KOMISARUK

bright again with every beat of my heart . . . . I was able to detect a definite melody. It was a composite of high-pitched sounds like human voices, accompanied by a deep bass drum. I focused all my attention on the melody, and again noticed that the systole and diastole o f my heart coincided with the sound o f the bass drum, and with the pattern of the music (Castaneda. 1974, pp. 98-99).

These perceptions are also similar to those described by Custance (1964) in a psychotic patient: “The great male and female organs of love hung there in mid-air . .. pulsing rhythmically in a circular clockwise motion, each revolution taking approximately the time of a human pulse or heartbeat, as though the vision was associated in some way with the circulation of the blood.” Since these perceptions fluctuate along the gradients of warm-cold, risingfalling, light-dark, extension-flexion, yellow-blue-black, expansion-contraction, and free-confined, cyclically in time with the heartbeat, it seems likely that they would be temporarily associated with fluctuations in autonomic tone. Hagbarth and Vallbo (1968) have shown autonomic nerve activity to be phase-coupled with the heartbeat in humans. Sympathetic tone increases suddenly in response to changing from a reclining to an upright posture, providing increased blood pressure necessary to maintain the blood supply to the head (McLaughlin ef a l . , 1978). Conversely, increased parasympathetic tone has been termed “gravity-submissive’’ (Kempf, 1953). Therefore, perhaps the feelings which accompany the sympathetic-dominant phase of the cardiac cycle are those related to accelerating upward against gravity, feeling strong, expansive, and good, whereas those associated with the parasympathetic-dominant phase are perhaps those related to falling with gravity, feeling weak, contractive, and bad. At first, this formulation may seem to be at odds with that of Reich (1942) who proposed that the parasympathetic system is “operative wherever there is expansion.. .out of the self-toward the world ... and pleasure. Conversely, the sympathetic is found functioning wherever the organism contracts.. .away from the world-back into the self. . . [in] anxiety. . . sorrow, and pain” (p. 257). But in this formulation, Reich does not take into account the sympathetic activation that occurs at orgasm (Wenger et al., 1968; Zurkerman, 1971) and during exercise (e.g., Shepherd and Vanhoutte, 1979). Resolution of the issue would require assessment of the covariation between spontaneous mood and autonomic tone.

B.

BODYIMAGERY A N D METAPHORIC LANGUAGE

The previous section provided evidence suggesting that visceral activity can influence ideational imagery of the position and appearance of the body. The present section provides evidence suggesting that body imagery is used extensively in the formation of words and idioms. Since many words and idioms are metaphors of bodily sensations, an understanding of the metaphors that refer to bodily sensations may provide insight into our perception of such sensations.

VISCERAL-SOMATIC INTEGRATION

125

The physical body is used as a point of reference in forming word meanings, by which an identity with an external object is established. For example, as compiled by Swadesh (1971), “number” = “digit” = “finger or toe” (English; “five” = “hand” (Sumerian); “ten” = “belonging to the hands” (Chukchee); “twenty” = “man” (Mayan). Thass-Thienemann (1968) points out that expressions such as “I have a sinking feeling in the pit of my stomach” or “my heart goes out” are literally absurd, yet such words describe true subjective sensations. Similarly, abstract feelings are communicated by expressions which describe the bodily sensations by which they are accompanied; for example: hair-raising, skin- (or flesh-) crawling, spine-chilling (or -tingling), cold feet, no sweat (Sperling, 1981). Such viscerally based metaphors are common in language and provide meaning to words. For example, “nausea” has the same origin as “nautical”; the bodily sensation of nausea feels similar to the sensation that is produced by being on a ship in an undulating sea. According to Freud (1924), “The name Angst (anxiety)angustioe, Enge, a narrow place, a strait-accentuates the characteristic tightening in the breathing which was then the consequence of a real situation and is subsequently repeated almost invariably with an affect” (p. 404). Similarly, in onomatopoeia, the meaning of a particular word is congruent with the sound produced when pronouncing the word. These are forms of “iconic signals,” i.e., “literal images,” as are hand signals in American Sign Language (see Green and Marler, 1979). An example of such iconicity is seen in the following quote from Thass-Thienemann (1968). Considering such words as psycht, pneuma, and spiritus phonemically, it is difficult to deny ...that some sound symbolism is also operative in their meanings. The combined explosive and sibilant consonant cluster in the initial psy-, pneu-, spi-, suggests the explosive exit of air (pneumatic). The implication o f the odor perception o f the outflowing air may explain how this sound cluster became expressive of anal fantasies connected with disgust, loathing and contempt. Such are some interjections in English as pooh, pshaw, pish, fie. The same sound complex is present in the German pfui or the French f i . ., , The phonemic equivalent of the Greek pneuma, “spirit,” is the Old English fneo-san, Dutch fniezen, and Old Nordic fnysa, all meaning “sneeze.”

Hinde (personal communication) has pointed out that many “sn-” words are also related to the nose (e.g., snort, snore, snuffle, snivel, sniff, snout, snorkel) (Onions, 1966). Similarly, Swadesh (1971) suggests that many of our words originate from sound utterances and facial-tongue movements which imitate or metaphorically represent certain objects. For example, the dental “t” sound is produced by, and thus provides the effect (meaning) of the contact of a point, whereas the “ k ” sound provides that of a blunt object. Sounded together, they iconically represent “pointed to blunt” as in “tack.” Although the many exceptions to this line of reasoning restrict its generality, they do not necessarily deny its heuristic value.

126

BARRY R . KOMISARUK

There may exist a “bodily geometry” to feeling good or bad; falling and shrinking feel bad, whereas soaring up and expanding feel good. G. Lakoff and M. Johnson in an unpublished manuscript (‘‘Toward an experimentalist philosophy: The case from literal metaphor”) have catalogued the English phrases which describe feelings of good and bad in the form of physical metaphors based on body geometry. For example: “Happy is up; sad is down. I’m feeling up. That boosted my spirits. My spirits rose. You’re in high spirits. Thinking about her always gives me a lift. I’m feeling down. I’m depressed. He’s really low these days. I fell into a depression. My spirits sank.” They relate these metaphors to body imagery by pointing out that droopy posture typically goes along with sadness and depression, and erect posture with a positive emotional state. As noted in the previous section, these postures are associated with parasympathetic and sympathetic dominance, respectively. The words and idiomatic expressions may be based on specific feelings which are related to specific moods, postures, and autonomic tone. Other examples of body-based metaphor are as follows. The head in its role of standing for the entire body also has a tendency to exhibit its own dichotomy of front and back. Unwelcome thoughts are pushed 10 the back of the mind, whereas other ideas are accepted into the forefront of consciousness. What is in front is known, what is in back is unknown. Also, what is in back is held back, whereas what is in front is faced. We look to the future and push the past behind us. These metaphorical expressions may actually refer to physical sensations, and perhaps to physiological processes. (Reprinted by permission of the publisher from Kepecs, Psychosomatic Medicine IS, p. 427. Copyright 1953 by The American Gastroenterological Assofiation.)

Kepecs ( 1 953, p. 426) also points out that the expression “keep a stiff upper lip” refers “to the employment of muscle tension as a defense against vegetative discharge of feeling. This is similar to Reich’s (1942) identification of “character armor” with “muscular armor” and Dunbar’s (1943, p. 86) notion of the “relationship that exists between muscle tension and the keeping of important emotional material in repression, ” indicating that “muscle tension is a general defense against expression of vegetative energies. Similarly, Schwartz et al. (1976), on the basis of their findings that specific facial muscles are differentially active during mental imagery associated with happiness, sadness, and anger, and that the profiles of muscular activity differ between normal and depressed persons, postulate that “peripheral feedback from discrete, innate patterns of facial muscle activity provides an important component underlying the subjective experience of emotion. This conclusion is similar to that of Gellhorn (19641, who suggested that “mood depends to an important degree on posture ... the inner attitude may be induced through the external posture, and vice versa” (p. 413). These statements are consistent with the notion of William James (1884) that “we feel sorry because we cry, angry because we strike, afraid because we tremble, and not that we cry, strike, or ”

”

”

VISCERAL-SOMATIC INTEGRATION

127

tremble, because we are sorry, angry, or fearful.. .” Lowen provides support for this point of view in the following quote from one of his patients during a “falling exercise” in which the patient is encouraged to stand on one leg with bent knee until she falls to a cushion on the floor: “ ‘I am not going to fall. I am not going to fail. I’ve always failed.’ And with that remark she fell and began to cry deeply” (Lowen, 1976, p. 205). A link between the form of body expression and specific types of psychosomatic disease was demonstrated experimentally by Williams and Krasnoff (1964). They found that peptic ulcer patients gave significantly more Rohrschach responses emphasizing “penetration of (bodily) boundary” (e.g., mashed bug, person bleeding, soft mud) than did rheumatoid arthritis (RA) patients. The RA patients tended to give more “barrier” responses (e.g., “bottle,” “knight in armor,” “turtle with a hard shell”) than the peptic ulcer patients. Patients in both groups who had “high bamer” scores showed significantly higher muscle tension during an emotionally stressful test than patients with “low bamer” scores, indicating a relationship between body image (i.e., projecting the body as having a barrier or being penetrated) and bodily muscular expression. The generation of dream imagery by bodily sensations was suggested by Freud (1900) in the sixth chapter of “The Interpretation of Dreams.” He pointed out that Hippocrates had noted that disorders of the bladder were associated with dreams of fountains and springs, and disorders of the intestine, Freud believed, are associated with dream symbols of buried treasure, gold, and feces. Similarly, in a hypnagogic state between waking and sleep, Silberer (1951) observed visual imagery of body pantomimes which were metaphors of his thoughts. For example, a thought of his was “I am to improve a halting passage in an essay,” and his imagery was “I see myself planing a piece of wood.. . the position of the piece of wood I am planing is that of my lower arm; I really feel that my lower arm represents this piece of wood” (p. 206). Since dream or dreamlike imagery can apparently be generated by visceral afferent activity, it is likely that visceral motor activity would generate reafferent activity and thereby induce such imagery. A vivid example in support of this notion occurred in an instance when my auto skidded on an icy road and I could feel my stomach contract suddenly as I turned the steering wheel. I had the distinct imagery of my stomach feeling as if it was an auxiliary hand that reached out and gripped the wheel, helping me steer. The motor and sensory components of this process may well represent what Freud characterized as “primary process, ’’ which creates unconsciously generated, wish-fulfilling dream imagery (Mack and Semrad, 1967). In the present context, the visceral motor activity “acts out” an ineffectual but comprehensible response to a perceived situation (i.e., the stomach “helps” as an auxiliary hand) and that motor activity generates afferent activity which is perceived as an auxiliary hand in a dreamlike manner.

128

BARRY R. KOMISARUK

Another kind of transformation of body imagery, that occurs in organic disease was described by Schilder. Every protrusion can take the place of another. We have possibilities of transformation between phallus, nose, ear, hands, feet, fingers, toes, nipples and breasts; every round part can represent another-head, breasts, buttocks; every hole can be interchanged with another-mouth, ears (in some respects, eyes and pupils), openings of the nose and anus.. . . Actions may create artificial caves in the body; the inside of the hand and the inside of the mouth and the inside of the genital region may be substituted for each other (Schilder. 1950, pp. 182-183).

This transformation is also evident in more prosaic examples, such as the facial and tongue movements of a child trying to copy a figure by hand, or of a mother trying to coax a spoonful of food into her baby’s mouth. If some food drops off the spoon as the baby opens its mouth, the mother might suddenly drop the edge of her own mouth as if to catch it. She might use the edge of her mouth as she would her hand or her shoulder and arm in that situation. The expression “down at the mouth” refers to an attitude in which everything is drooping, as if the shoulders are bearing heavy weights. A related example is the facial grimace accompanying lifting a heavy weight. Perhaps the mouth corners are used as metaphors of the wrists. A contrast, with respect to gravity, is arms extended up in victory and a victory smile with the edges of the mouth raised high. The face apparently can mimic what the hand does. Significantly, the facial and scalp muscles are derived embryonically from the visceral branchial arches, and are innervated by the fifth and seventh cranial nerves, which are “visceral” nerves (Romer, 1962; see Komisaruk, 1977). Since the face performs actions directed at the environment which are metaphors of what the hands do, but the facial muscles are of “visceral” origin, face-hand coordination may represent a form of visceral-somatic integration. Perhaps the face-scalp can act as a “hollow” viscus representing both the stomach and the hand. For example, when one is nauseated and feels a vomiting-like movement in the stomach, a characteristic movement of the face appears, in which the lips are pursed and the cheeks are billowed out. Thus, the face may adaptively “act out” a pantomime, i.e., create a physical metaphor, of a specific movement of the stomach or of the hand. In this section I have speculated on some ways in which somatic and visceral integrated activity gives rise to expressive imagery, metaphor, language, and symbolic behavior. It should be emphasized that the gradients of sadnesshappiness, and droopy-erect posture appear to parallel the parasympatheticsympathetic and gravity submissive-gravity opposing gradients described in the previous section. These also parallel the trophotropic-ergotropic gradients described by Hess (1954). It is thus tempting to speculate that skeletal postures, visceral tone, and mood, feelings, and body-metaphor-based language are all integrated along congruent physiological and cognitive gradients.

VISCERAL-SOMATIC INTEGRATION

V.

129

TOWARD A CONCEPT OF PSYCHOGENIC ORGANIC (“PSYCHOSOMATIC”) DISEASE

The visceral system can act out the primary process without the participation of the usually associated somatic system, as in the report of increased blood flow to an arm when one thinks of moving the arm (Dunbar, 1946). This is a form of “motoric thinking. ” In the somatic system, L. W. Max (see Crafts et a l ., 1938) found that when subjects were told to imagine acts such as holding a wriggling snake behind the head, holding a squirming fish in the hands, telegraphing an SOS signal, and typewriting the subjective experience of kinesthetic image was usually accompanied by contractions of the muscles which would be active in the performance of the acts imagined, although the contractions themselves were not perceived by the subjects. In the realm of the visceral system, motoric thinking can be viewed as, e.g., the activity of the cardiovascular system being adjusted appropriately to the holistic attitude of the body. Kempf (1953) vividly contrasts its activity in panic and rage. In panic, when unable to escape from deadly force, the pulse grows extremely fast and small, accompanied by cutaneous pallor and visceral vasodilation, with fall in blood pressure and great general weakness, trembling and fumbling, and even fainting of submission. In rage or hate, when the offended person feels stronger than the offensive force, the heart develops slower, stronger and larger contractions than in fear. This sustains a greater volume of blood at high pressure, and supports violent somatic compulsions to destroy the cause (Kempf. 1953, p. 316).

If we propose that the motor activities of the viscera enact our fantasies, then we can assume that the visceral sensations that result from the enactment bear some relationship to those that would occur if we actually enacted the specific fantasy with our entire body. Perhaps individuals who eventually develop gastric ulcers use their stomach as they would their hand, the stomach contracting as a sphere when the hand, face, and thorax-abdomen do so. Perhaps the pain that these actions of the stomach create gives them the feeling that they would get from a powerful crush, pinch, or twist. This implies that the symptoms in psychogenic organic disease may be unconsciously created bodily representations of the way in which the individual deals with a given situation at a primary process level. Wright (1976) in referring to phobic symptoms proposes that “the symptom is a wordless presentation of an unnameable dilemma-an abortive metaphor that stops below the level of speech ... the undoing of a symptom is in part the creation of (conscious) metaphor from symptom. ” By performing viscerally a metaphor of what we feel like doing but refrain from doing somatically, we may generate a cognitive representation of our fan-

130

BARRY R . KOMISARUK

tasied response. That is, we may provide ourselves with the visceral equivalent feelings that fulfill our wish, e.g., by acting out an angry punch, using the stomach but not the arms. If the blood vessels to the arm perseverate in acting out the punch, but we prevent the arm from actually doing so, this could, perhaps, eventually lead to pathological changes in the muscles of the arm. Indeed, Travel1 (1960) has described a process of muscle contracture and eventual locking of skeletal joints in cases of prolonged conversion reactions, the musculoskeletal system eventually becoming physically incapable of performing the actions which were previously inhibited hysterically. According to Weiner (1977, pp. 457-458) “physical diseases such as rheumatoid arthritis are often accompanied by conversion symptoms.. . . The hypothesis that psychological tension is translated into tension in muscles and tendons around joints is central to many past formulations about the initiation of rheumatoid arthritis. For example, Reich (1942, p. 267) stated, “Every muscular rigidity contains the history and the meaning of its origin.” Weiner (1977) has raised the question of whether psychological factors could be mediated through autonomic outflow channels to produce vasoconstriction in the arterioles that supply the joints, cartilage, or synovial cells. Thus, the overriding wish not to perform some act would eventually be fulfilled by the body’s becoming incapable of performing it. According to Dunbar (19461, “somatic symptoms appear as safety valves against the appearance of repressed material.” Following are some examples, in the case of rheumatoid arthritis, of what is interpreted to be metaphoric use of body parts in expressing conflicts arising from repressed desires. ”

The close relationship between the psychological life history of the patient and arthritic process receives additional emphasis from the fact that the localization of the disease seems to be in the joint or joints in which the conflict is focussed.. . . The joints play an important part in an activity that is disliked, and which, as the ultimate result of the disease, cannot be continued. Thus, in 12 cases, the joint was strained in an occupation which the patient hated, and in nine cases the sacro-iliac joints were particularly involved, where repression of the sexual act constituted an integral part of the patient’s dilemma. The joint is essential to the assumption of an attitude which is expressive, in a symbolical way, of the patient’s general demeanor or conduct. Thus, in seven cases of primary stiffening of the spine, conceited haughtiness was very evident (Booth, 1937, p. 144).

Booth ( I 937, p. 645) concluded that, “The arthritic process tends to localize in joints which the patient, for conscious and/or unconscious reasons, desires immobilized rather than active.” Similarly, Lowen (1976, p. 86) stated, “Tensions in the small muscles of the hand are the result of repressed impulses to grasp or seize, to claw or to strangle. I believe such tensions are responsible for rheumatoid arthritis in the hands.” The use of specific bodily parts to express specific conflicts led Engel ( 1 968) to consider rheumatoid arthritis as a form of conversion reaction. He stated that a complication of the conversion ultimately results in local tissue damage.. .involving. ..joints and even parts of the vascular system. For example, we have seen Reynaud’s

VISCERAL-SOMATIC INTEGRATION

131

phenomenon appear first in the index finger of a woman about to dial the phone and “tell off” her mother, and rheumatoid arthritis first in the ankle of a man upon impulse to kick down the door of a rejecting girlfriend (Engel, 1968, p. 321).

Deutsch (1959) has developed a concept of somatic disorders based upon the conversion process. External objects are perceived as if severed from the body and lost. This separation leads to the continual wish to restore the loss of the bodily wholeness.. .. The child reacts to this loss of an object with the attempt to regain it, to retrieve this part of himself, by imagining it. Attempts of this nature continue throughout life and can be considered as the origin of the conversion process.. .. The objects outside become reunited with the body by way of symbolization ... the source of the symptom is the wish for, and the flight from, a symbolized object which stirs up an emotional process aimed at undoing the loss. ...However, if no surrogate object is available, conversion symptoms are formed. This can lead to partial or complete inhibition or to hyperactivity of a bodily function. The organic symptom is the protective device against an impending loss of the object (Deutsch, 1959, pp. 76-77).

We may speculate that visceral disease may develop when we do not perceive, understand, or act out responses to, our visceral afferent messages. This notion is supported by the following quote from Dunbar (1946). Important in the psychic situation of organic heart patients is the absence of a definite correlation between the seriousness of the illness and the subjective experience of it. In striking contrast to the objective findings, these patients show little or no consciousness of disease, and consequently, no insight; complaints (shortness of breath on exertion, etc.) are minimized or dissimulated. This situation is peculiarly characteristic of a majority of serious organic heart patients (p. 208).

Perhaps the intensity of the sensory activity generated by visceral efferent activity continues to increase until one responds to it consciously. If one does not respond to it, it persists and/or intensifies, leading eventually to organic disorder. In other words, the viscera may act out our fantasies and in so doing, provide us with some, but not all, of the components of the stimulation which we would receive from the body if we acted out our fantasies. What is missing is the stimulation provided by reality. Thus, in the visceral, but not the somatic, realm we act out our primary process. Perhaps stomach spasm is a visceral pantomime which resembles the feeling of stomach fullness, and which the primary process might initiate if one felt abandoned, as a means of recreating the feeling of being fed, thereby mitigating the feeling of abandonment. The stomach spasm pantomime could also provide the feeling of a hand grasping a desired object. In both cases, the stomach can be viewed as performing a wish-fulfilling pantomime, in which one “gets what one wants” in a dreamlike, but not a real, sense. If one feels lonely, the primary process might create the feeling of pressure in the chest which one would obtain from a hug. Lung congestion against which we

132

B A R R Y R. KOMISARUK

strain to breathe creates a feeling of pressure. Perhaps asthmatic attacks which are precipitated by anxiety could be understood as an attempt by the body to generate this specific sensory stimulation. This “visceral strategy” would operate when one feels unable to obtain appropriate desired sensory stimulation from the external world. If specific somatic stimulation (e.g., a hug from a loved one) is impossible or thwarted, the closest substitute may be viscrrul stimulation originating in a body region whose sensory input is convergent with the desired somatic stimulation. Examples of such convergence are seen in referred sensation, e.g., forearm cutaneous input converging with cardiac input to the same spinal neurons (Foreman, 1977) or urinary bladder and cutaneous thigh afferents converging on the same spinal neurons (Fields et ul., 1970). Thus, cigarette smoke may stimulate visceral spinal afferents whose spinal projection fields overlap with those of somatic afferents that are stimulated by a hug. If the smoking strategy is not used, perhaps production of lung congestion would be the body’s next resort. In other words, a “strategy” of generating lung congestion as a substitute for thoracic contact stimulation (hugging) might characterize certain forms of asthma. This notion is supported by Weiner’s (1977, p. 297) characterization of some asthmatic patients. In about one-half of asthmatic patients the dependency conflicts take the form of the unconscious wish to be engulfed and protected. The threat or actuality of separation may then mobilize the cry for the mother which may not be fully expressed. and which is associated with the asthmatic attack.

Perhaps the cry generates sensory stimulation via increased muscular tension in limbs, trunk, and throat as a compensatory substitute for a perceived inadequate level of sensory stimulation. When the stimulation is provided by an external source (e.g., holding, hugging, rocking) or the symbolic equivalent, it is no longer necessary to generate the stimulation by crying. A similar process of trying to provide a form of sensory stimulation which compensates or substitutes for that which is lost, could account for the series of events leading to the development of a peptic ulcer in a patient described by Savitt (1977, p. 609). In the course of the analysis, the patient revealed a striking example of an earlier somatization of the oral zone. At sixteen, about a year before he developed his stomach ulcer, he had experienced an overpowering need to suck. It was as persistent a desire as his earlier wish to suck his thumb, which had lasted until he was twelve. This need to suck seemed beyond his control; while sucking, he pulled at the buccal mucosa of his right cheek until a buccal papilloma developed. He had “created” in this papilloma a breastnipple equivalent which he could mouth and suck at will. This constituted an internal maternal breast that could not be taken away. But it was a “dry breast” which gave no nourishment; it was as nongiving as his mother had been. Consistent biting, chewing, and

VISCERAL-SOMATIC INTEGRATION

133

sucking enlarged and macerated the papilloma; ultimately it had to be removed surgically. With the loss of this part-object breast substitute, he sustained another oral deprivation, which may have been a factor in the original somatization of his oral conflicts that led to the peptic ulcer at age seventeen.

Similarly, Weiner (1977, p. 83) concluded that “many adult male patients are indeed predisposed to peptic duodenal ulcer by long-standing and unconscious wishes to be loved, cared for, and fed, and given that they also have the anatomical and physiological predispositions to the illness. ” It is also relevant to note that Fenichel (1945) called attention to the permanent hunger for love and its frustration as a factor in the etiology of stomach ulcer. Anxiety has been considered to provide a common basis for psychosomatic disease. Fromm (1 956) considered anxiety to be the expectation of isolation in some form, such as separation, abandonment, or ostracism. This notion is supported in instances of essential hypertension and rheumatoid arthritis, as well as asthma and ulcer which have been discussed. Thus, in the case of rheumatoid arthritis, Weiner ( I 977, p. 477) stated, “separation from another person, especially from someone who had previously been submissive to the patient, and felt trapped in the relationship, are common onset conditions.” And in the case of essential hypertension, “separation may antecede the malignant phase. ” The implication of these conclusions regarding a common basis for psychogenic organic disorders is that visceral activation occurs when the individual perceives that he is receiving inadequate sensory stimulation. As Kaufman (1960, pp. 32 1-325) pointed out, “Most gratifications are in fact derived from stimulation, not the lack of it; people deprived of sensory experience hallucinate it; . . . Freud . . . said that the child sought this experience (nursing) again for the pleasurable state it produced, which it should be noted is a state of stimulation.. . ” McCray (1978) reports that compulsive excessive masturbation in children (10-15 times per day) has a common theme of withdrawal of affectionate parental tactile stimulation, and is reversed by reinstatement of affectionate nonsexual tactile contact by the parents. The pleasurable aspect of the various types of orgasm may be due to the nonaversively intense sensory stimulation that is generated by movement of visceral and somatic structures. This implies that the lack of adequate sensory stimulation may induce the organism to generate sensory stimulation “homeostatically. ” Perhaps psychogenic organic disorders arise if the self- or other-generated somatic stimulation, or its symbolic equivalent, is inadequate or thwarted. The individual might then seek compensatory physical stimulation by organizing his or her behavior to generate stimulation of the viscera at the equivalent segmental level, such as by eating, drinking, or smoking. If this visceral self-stirnulation is perceived as being an inadequate substitute for somatic stimulation, the visceral efferent system (e.g., parasympathetic, controlling gastric secretion or lung congestion) may compensate by becoming hyperactive, thereby generating visceral

134

BARRY R. KOMISARUK

sensory activity. If this inadequate compensatory input persists for an extended period, the results could be organic disorder represented by intestinal ulceration, asthma, etc. Although the autonomic system is characterized by negative feedback leading to homeostasis (Gellhorn, 1957), it is likely that cognitive factors (not necessarily “conscious” or “voluntary”) affect the “set-point’’ of the system, so that certain levels of sensory input are perceived as being inadequate, thus allowing the activation of compensatory processes. When visceral afferent activity is generated by compensatory efferent visceral hyperactivity, it may tend toward providing the calming, reassuring effect that the thwarted somatic input that was previously experienced had provided. Even visceral discomfort or pain might actually be reassuring: it may be perceived as preferable to the absence of stimulation, which is equivalent to being isolated. Therapeutically, this formulation implies that fulfilling the wish somatically or viscerally might reverse the necessity to perpetuate the compensatory visceral efferent activity, thereby perhaps reversing the organic pathology.

VI. CONCLUSION A major function of the visceral system is to provide metabolic support for the somatic (musculoskeletal) system. It prepares the organism for, and enacts, basic energetic adaptations to the environment. Therefore, the “language.’’ of covert visceral activity can be understood in terms of the overt somatic activity which the visceral activity normally supports. The visceral system can be viewed as enacting the “primary process” level of thinking (i.e., wish fulfillment) with, and even without, the participation of the somatic system. Pathological states may develop under chronic conditions of dissonance between the activity of the two systems. Specifically, during conflict situations the activity of the cardiovascular system may not match that of the musculoskeletal system. As an example, Reynaud’s disease may develop out of chronic unresolved conflict represented by the body metaphor expressed in the extremities and referred to idiomatically as “cold feet or hands.” As the primary process is enacted by the visceral effector organs, afferent activity emanates from them. A state of dissonance exists when the visceral system provides such afferent activity in the absence of either matched somatic motor activity or appropriate sensory stimulation, a consequence of which may be organic pathology. In contrast to the condition of dissonance, consonance between the activity patterns of the visceral and somatic systems probably provides optimal strength and stamina, energy efficiency, and ease of movement. The temporal coordination between these systems is probably facilitated by the existence of rhythmical activity cycles of the cardiovascular, respiratory, and locomotor system, which under certain conditions are coupled to each other and to several rhythmical brain

VISCERAL-SOMATIC INTEGRATION

135

activity systems. Maximum consonance may be represented by orgasm and other “peak experiences. In this contribution, emphasis has been placed on the integration between visceral and somatic activity at various levels of the neuraxis from high-order symbolic activity to simpler segmental reflexes. This approach views the visceral system as both a psychological and a physiological entity, to which common principles of operation may apply. ”

Acknowledgments Contribution No. 291 from the Institute of Animal Behavior. I wish to thank Drs. M.-C. Busnel. R. Hinde, S. W. Porges, J. S. Rosenblatt, J. Rush, and K. Semba for their critical reading of earlier versions of the manuscript. I also wish to thank Drs. A. Naggar, M. Numan, H. Szechtman, and J . Wallman, and Ms. Woodall for their helpful comments in the course of the writing, and Ms. W. Cunningham and Ms. C. Banas for technical assistance. This work was supported in part by NSF grant #BNS 78-24505, and U.S. Public Health Service Grants MH-13279 and MH-14711.

References Alther, L. 1975. “Kinflicks.” New American Library, New York. Arey, L. B. 1954. “Developmental Anatomy, 6th Edition. ” Saunders, Philadelphia, Pennsylvania. Beach, F. A. 1966. Ontogeny of “coitus-related” reflexes in the female guinea pig. Proc. Narl. Acad. Sci. U.S.A. 56, 526-533. Bechbache, R. R., and Duffin, J. 1977. The entrainment of breathing frequency by exercise rhythm. 1.Physiol. (London) 272, 553-561. Biederman-Thorson, M., and Thorson, J. 1973. Rotation-compensating reflexes independent of the labyrinth: Neuromuscular correlates in the pigeon. J. Comp. Physiol. 83, 103-122. Birren, J. E.,Cardon, P. V., and Phillips, S . 1963. Reaction time as a function of the cardiac cycle in young adults. Science 14, 195-196. Boling, J. L., Blandau, R. J . , Wilson, J. G., and Young, W. C. 1939. Postpaxturition heat responses of newborn and adult guinea pigs, dataon parturition. Proc. SOC.Exp. Biol. Med. 42, 128-132. Booth, G. C. 1937. Personality and chronic arthritis. J . New. Menr. Dis. 85, 637-662. Boreham, J. L., Kibbler, G. O., and Richter, D. 1949. A relation between a psychomotor response and the phase of the alpha rhythm. J . Physiol. (London)109, 17 P. Borgdorff, P. 1975. Respiratory fluctuations in pupil size. Am. J. Physiol. 228, 1094-llM. Buchsbaum, M., and Callaway, E. 1965. Influences of respiratory cycle on simple reaction time. Percept. Moror Skills 20, 961-966. Buzsaki, G., Grastyan, E., Czopf, J., Kellenyi, J., and Prohaska, 0. 1982. Changes of neuronal transmission in the rat hippocampus during behavior. Bruin Res. (in press). Byers, P. 1977. A personal view of nonverbal communication. Theory info Pracrice 16, 134-140. Byers, P. 1979. Biological rhythms as information channels in interpersonal communication behavior. In “Nonverbal Communication” ( S . Weitz, ed.), pp. 398-418. Oxford Univ. Press, London and New York. Callaway, E. 1965. Response speed, the EEG alpha cycle, and the autonomic cardiovascularcycle. In “Behavior, Aging and the Nervous System” (A. T. Welfor and J. Birren, eds.). Thomas, Springfield, Illinois.

136

BARRY R. KOMISARUK

Callaway, E., and Buchsbaum, M. 1965. Effects of cardiac and respiratory cycles on averaged visual Clirr. Neuroph\sio/. 19, 476-480. evoked responses. E/Pctri~eric.ephab~r. Callaway. E.. and Layne, R . 1964. Interaction between the visual evoked response and two spontaneous biological rhythms: The EEG alpha cycle and the cardiac arousal cycle. Ariri, N.Y. A(u/. Sci. 112, 421-431. Castaneda, C. 1974. "The Teachings of Don Juan: A Yaqui Way of Knowledge." Simon and Schuster. New York. Clynes. M. 1979. Sentics: Communication and generation of emotion through dynamic expression. IJI "Nonverbal Communication" (S. Weitz. ed.), pp. 386-397. Oxford Univ. Press, Lundon and New York. Coleman, W. M . 1921. The psychological significance of bodily rhythms. J . Corrip. P s j h d . 1, 213-220. Coote. J . H. 1975. Physiological significance of somatic afferent pathways from skeletal muscle and joints with retlex effects on the heart and circulation. Brairi Res. 87, 139-144. Crafts, L. W.. Schneirla, T. C.. Robinson, E. E., and Gilbert. R. W. 1938. The iiiotor theory of thinking. Irr "Rccent Experiments in Psychology." pp. 37 1-384. McGraw-Hill. New York. Cross. B. A . 1959. Neurohypophyseal control of parturition. Irr "Recent Progress in the Endocrinology of Reproduction" (C. W. Lloyd. ed.), pp. 44-445. Academic Press. New York. Custance. J . 1964. The universe of bliss and the universe of horror: A description of a manicdepressive psychosis. 1ri "The Inner World of Mental Illncss" (B. Kaplan. ed.). p. 50. Harper, New York. Daunton. N . G..and Thonisen. D. D. 1976. Otolith-visual interactions in single units of cat vestibular nuclei. Neurosci. Ahsrr. I / 2 , 1057. Delius. J . D.. and Vollrath. F. W . 1973. Rotation compensating reflexes independent of the labyrinth. J . CrJWtp.Physiol. 83, 123-134. Deutsch, F. 1959. "On the Mysterious Leap froni the Mind to the Body. A Study on the Theory of Conversion. International Universities Press, New York. Diakow, C. 1978. Hormonal basis for breeding behavior in female frogs: Vasotocin inhibits the release call of Rana pipiens. Scierice 199, 1456-1457. Dodson. B. 1974. "Liberating Masturbation." Bodysex Designs. New York. Drewett, R. F.. Statham, C., and Wakerley. 1. 8. 1974. A quantitative analysis of the feeding behaviour of suckling rats. Ariini. Bekav. 22, 907-913. Dunbar. F. 1943. "Psychosomatic Diagnosis." Paul B. Hoeber, New York. Dunbar. F. 1946. "Emotions and Bodily Changes.'' 3rd ed. Columbia Univ. Press. New York. Eble. J . N. 1960. Patterns of response of the paravertebral musculature to visceral stimuli. Arn. J . Phy.yid. 198, 429-433. Engel, G. L. 1968. A reconsideration of the role of conversion in somatic disease. Cornpr. Pswhiatr\ 9, 3 16-326. Evans, M. H.. and McPherson. A . 1958. The effects of stimulation of visceral afferent nerve fibers on somatic reflexes. J . Physiol. (Lorirlon) 140, 201-212. Evans, M. H., and McPherson, A. 1960. The effects of electrical stimulation of visceral afferent nerve fibers on monosynaptic and polysynaptic reflex responses. J . Phwiol. (Loridmi) 150, 105-1 13. Fenichel, 0. 1945. "The Psychoanalytic Theory of Neuroses." Norton, New York. Fields, H. L.. Partridge. L. D., Jr., and Winter, D. L. 1970. Somatic and visceral receptive field properties of fibers in ventral quadrant white matter of the cat spinal cord. J . NtwropliJsiol. 33, 827-837. Foreman, R . D. 1977. Viscerosomatic convergence onto spinal neurons responding to afferent fibers located in the inferior cardiac nerve. Brairi Res. 137, 164-168. Frankstein, S. I., Lisin, V . V . . and Sergeeva. L. N. 1974. Is there a special pacemaker for stepping'? E.rpcrierrriu 30, 1414-1415. "

VISCERAL -SOMA TIC INTEGRATION

137

Freud. S . IYOO. “The Interpretation of Dreams” (J. Strachey. ed. and transl.). Avon, New York. 1965).

Freud. S. 1924. ”A General Introduction to Psychoanalysis” (J. Riviere, transl.). Washington Square Press, New York. 1960 Edition. pp. 333-334. Fromm, E. 19.56. “The Art of Loving.” Harper. New York. Gaardner. K.. Koresko. R.. and Kropfl. W . 1966. The phasic relation of a component of alpha rhythm to fixation saccadic eye movements. E / e c . f r [ ~ e ~ i c , c , ~ h uClin. l [ i ~ r .Neurophysiol. 21, 544-55 I. Gellhorn. E. 19.57. “Autonomic Imbalance and the Hypothalamus.” Univ. of Minnesota Press, Minneapolis. Gcllhorn. E. 1964. Motion and emotion: The role of proprioception in the physiology and pathology of the emotions. Psycho/. Rev. 71, 457-472. Gootnian, P. M.. Cohen. M. I . , Piercey, M. P., and Walotsky. P. 1975. A search for medullary neurons with activity patterns similar to those in sympathetic nerves. Brain Res. 87, 395-406. Green, S.. and Marler. P. 1979. Analysis of animal communication. In “Handbook of Behavioral Neurobiology.” Vol. 3. “Social Behavior and Communication” (P. Marler and J . G. Vandenbergh. eds.). pp. 73-158. Plenum, New York. Hagbarth. K.-E.. and Vallbo. A. B. 1968. Pulse and respiratory grouping of sympathetic impulses in human muscle nerves. Acru Physid. S c u d . 74, 96- 108. Hall. W. G.. and Rosenblatt, J . S. 1977. Suckling behavior and intake control in the developing rat PUP. J . COtnp. Physiol. Ps!(’hd. 91, 1232-1247. Harper. L. V . 1972. The transition from filial to reproductive function of “coitus-related” responses in young guinea pigs. D w . Psychobiol. 5 , 2 1-34. Hart. B. L. 1978. Hormones, spinal reflexes, and sexual behaviour. It! “Biological Determinants of Sexual Behaviour” (J. B. Hutchison, ed.). pp. 319-347. Wiley. New York. Harter. M.. and White, C . 1967. Perceived number and evoked cortical potentials. Science 156, 406-408.

Hess. W. R. 1954. ”The Diencephalon.” Grune & Stratton. New York. Hite. S. 1976. ”The Hite Report: A Nationwide Study of Female Sexuality.” Dell, New York. Hollander. X . 1981. Out of touch with sex. Abstracts of the Fifth World Congress of Sexology. Jerusalem. p. 196. James, W. 1884. What is an emotion’! Mind 9, 188-205. Jasper. H.. and Andrews. H. 1938. Brain potentials and voluntary muscle activity in man. J. Nmrophysiol. 1, 87-100. Kaufnian, I . C. 1960. Some theoretical implications from animal behaviour studies for the psychoanalytic concepts of instinct, energy, and drive. Inf. J. Psycho-anal. 41, 318-326. Kempf, E. J. 1953. Neuroses as conditioned, conflicting, holistic, attitudinal, acquisitive-avoidant reactions. A n n . N . Y . A m / . Sci. 56, 307-329. Kepecs. J . G. 1953. Some patterns of somatic displacement. Psychosom. Med. 15, 425-432. Koizumi, K.. and Brooks. C. McC. 1974. The autonomic nervous system and its role in controlling visceral activities. In “Medical Physiology.” Vol. I , 13th ed. ( V . B. Mountcastle, ed.). pp. 783-812. Mosby. St. Louis. Komisaruk. B. R. 1970. Synchrony between limbic system theta activity and rhythmical behavior in rats. J . Coinp. Physiol. Psycho/. 70, 482-492. Komisaruk. B. R. 1977. The role of rhythmical brain activity in sensorimotor integration. Prog. Psychobiol. Physiol. Psycho/. 7, 55-90. Komisaruk. B. R.. and Diakow. C . 1973. Lordosis reflex intensity in rats in relation to the estrous cycle. ovariectoniy. estrogen administration, and mating behavior. Endocrinology 93, 548-557. Komisaruk. B. R.. and Larsson. K. 1971. Suppression of a spinal and a cranial nerve reflex by vaginal or rectal probing in rats. Bruirr Rrs. 35, 231-235. Komisaruk. B. R.. Adler, N. T., and Hutchison. J. 1972. Genital sensory field: Enlargement by estrogen treatment in female rats. Science 178, 1295-1298.

138

BARRY R . KOMISARUK

Komisaruk, B. R., Terasawa. E., and Rodriguez-Sierra, J. F. 1981. How the brain mediates ovarian responses to environmental stimuli: Neuroanatomy and neurophysiology . I n “Neuroendocrinology of Reproduction” (N. T . Adler. ed.), pp. 349-376. Plenum, New York. Kow, L.-M., Montgomery, M. 0..and Pfaff, D. W. 1979. Triggering of lordosis reflex in female rats with somatosensory stimulation: Quantitative determination of stimulus parameters. J . Neurophysiol. 42, 195-202. Kuntz, A. 1945. Anatomic and physiologic properties of cutaneo-visceral vasomotor reflex arcs. J . Neurophysiol. 8, 42 1-429. Kuntz, A,, and Haselwood, L. A. 1940. Circulatory reactions in the gastrointestinal tract elicited by localized cutaneous stimulation. Am. Heart J . 20, 743-749. KUN, M. 1965. Nervous control of micturition. Physiol. Rev. 45, 425-494. Langhorst, P., Stroh-Werz, M., Dittmar, K., and Camerer, H. 1975. Facultative coupling of reticular neuronal activity with peripheral cardiovascular and central cortical rhythms. Brain Res. 87, 407-4 18. Lansing, R. 1957. Relation of brain and tremor rhythms to visual reaction time. EEG Clin. Neurophysiol. 9, 497-504. Liljestrand, G . 1965. Introduction to C.J.F. Heyman’s “the part played by vascular presso- and chemo-receptors in respiratory control. In “Nobel Lectures: Physiology of Medicine 19221941 ,” pp. 455-459. Elsevier, Amsterdam. Lourie, R. S. 1949. The role of rhythmic patterns in childhood. Am. J . P v c h i a t . 105, 653-660. Lowen, A. 1976. “Bioenergetics.” Penguin, New York. Macdonald, D., and Amlaner, C. 1981. Listening in to wildlife. New Sci. 89, 466-469. McCray. G. M. 1978. Excessive masturbation of childhood: A symptom of tactile deprivation? Pediatrics 62, 277-279. Mack, J . E . , and Semrad, E. V. 1967. Classical psychoanalysis. I n “Comprehensive Textbook of Psychiatry” ( A . M. Freedman and H. I. Kaplan. eds.), pp. 269-319. Williams & Wilkins, Baltimore, Maryland. McLaughlin, L. J., Goldman, H., Kleinman, K . M., and Korol, B. 1978. The influence of body position on autonomic nervous system function. Pavlovian J . Biol. Sci. 13, 29-41. McPherson, A. 1966. Vesico-somatic reflexes in the chronic spinal rat. J . Physiol. (London) 185, 197-204. Macrides, F. 1975. Temporal relations between hippocampal slow waves and exploratory sniffing in hamsters. Behav. B i d . 14, 295-308. Martin, L. T . , and Alberts, J . R. 1979. Taste aversions to mother’s milk: The age-related role of nursing in acquisition and expression of a learned association. J . Comp. Physiol. Psycho/. 93, 430-445. Masters, W. H., and Johnson, V. E. 1979. Interview: Masters and Johnson. Playboy, Nov. 1979, p. ”

110.

Monnier. M. 1968. Regulation of reproductive functions. In “Functions of the Nervous System: Vol. I : General Physiology and Autonomic Functions,” pp. 477-478. Elsevier, Amsterdam. Muller-Schwarze. D. 1974. Antilocupra americana (Antilocapridae) birth and behavior of the newborn. I n “Encyclopedia Cinematographica” ( G . Wolf, ed.), Film E 2057. Institut fur den Wissenschaftlichen Film. Myslobodsky, M. 1976. Pleasure rhythm as a precursor of wave-spike activity. In “Petit Ma1 epilepsy. A Search for the precursors of Wave-Spike Activity,” pp. 167-179. Academic Press, New York. Nunn, C. M. H.. and Osselton, J . W. 1974. The influence of the EEG alpha rhythm on the perception of visual stimuli. Psychophysiology 11, 294-303. Onions, C. T. (ed.). 1966. ‘The Oxford Dictionary of English Etymology.” Oxford Univ. Press, London and New York.

VISCERAL-SOMATIC INTEGRATION

139

Pessenhofer, H., and Kenner, T. 1975. Zur Methodik der kontinuierlichen Bestimmung der phasenbeziehung zwischen herzschlag und atmung. Pfluegers Arch. 355, 77-83. Porges, S. W., Bohrer, R. E., Cheung, M. N., Drasgow, F., McCabe, P. M., and Keren, G . 1980. New time-series statistic for detecting rhythmic co-occurrence in the frequency domain: the weighted coherence and its application to psychophysiological research. Psychol. Bull. 88, 580-587.

Preiss, G . , Kirchner, F., and Polosa, C. 1975. Patterning of sympathetic preganglionic neuron firing by the central respiratory drive. Brain Res. 87, 363-374. Ralston, H. J. 1976. Energetics of human walking in neural control of locomotion. In “Advances in Behavioral Biology, Vol. 18” (R.M. Herman, S. Grillner. P. S. G. Stem, and D. G . Stuart, eds.), pp. 77-98. Plenum, New York. Reich, W. 1942. “The Discovery of the Orgone - the Function of the Orgasm.” Noonday, New York. Remond, A., and Lesevre. N. 1967. Variations in average visual evoked potential as a function of the alpha rhythm phase (“autostimulation”). Electroencephalogr. Clin.Neurophysiol. 26 Suppl., 42-52.

Richins, C. A,, and Brizzee, K. 1949. Effect of localized cutaneous stimulation on circulation in duodenal arterioles and capillary beds. J. Neurophysiol. 12, 131-136. Rodriguez-Sierra, J. F., Crowley, W. G., and Komisaruk, B. R. 1975. Vaginal stimulation in rats induced prolonged lordosis responsiveness and sexual receptivity. J . Comp. Physiol. Psychol. 89, 79-85. Romer, A. S. 1962. “The Vertebrate Body,” 3rd ed. Saunders, Philadelphia, Pennsylvania. Ruch, T. C. 1961. Pathophysiology of pain. In “Neurophysiology” (T. C. Ruch, H.D. Patton, J . W. Woodbury, and A. L. Towe, eds.), pp. 350-368. Saunders, Philadelphia, Pennsylvania. Sachs, C. 1948. “Our Musical Heritage.” Prentice-Hall, New York. Satchell, G . H. 1968. A neurological basis for the co-ordination of swimming with respiration in fish. Comp. Biochem. Physiol. 27, 835-841. Sato, A. 1975. The somatosympathetic reflexes: Their physiological and clinical significance. In “The Research Status of Spinal Manipulative Therapy.” NINCDS Monogr. 15. Savitt, R. A. 1977. Conflict and somatization: Psychoanalytic treatment of the psychophysiologic response in the digestive tract. Psychoanalyric Q. 46, 605-622. Schilder, P. 1950. “The Image and Appearance of the Human Body.” International Universities Press, New York. Schlant, R. C. 1978. Normal physiology of the cardiovascular system. In “Heart, Arteries. and Veins,” 4th ed. (J. Willis Hurst, ed.), pp. 71-100. McGraw-Hill. New York. Schwartz, G. E., Fair, P. L., Salt, P., Mandel, M. R., and Klerman, G.L. 1976. Facial muscle patterning to affective imagery in depressed and nondepressed subjects. Science 192, 489491.

Scott, W. C. M. 1948. Some embryological neurological, psychiatric and psychoanalytic implications of the body scheme. Inr. J. Psycho-anal. 29, 141-155. Semba, K., and Komisaruk, B. R. 1978. Phase of the theta wave in relation to different limb movements in awake rats. Electroencephalogr. Clin. Neurophysiol. 44, 61 -71. Semba, K., and Komisaruk. B. R. Evidence of a dual neural mechanism related to rhythmical vibrissal movement in the rat. (Submitted for publication.) Semba, K., Szechtman, H., and Komisaruk, B. R. 1980. Synchrony among rhythmical facial tremor, neocortical “alpha” waves, and thalamic non-sensory neuronal bursts in intact awake rats. Bruin Res. 195, 281-298. Shepherd, 1. T., and Vanhoutte, P. M. 1979. “The Human Cardiovascular System.” Raven, New York. Silberer, H. 195 I . Report on a method of eliciting and observing certain symbolic hallucination-

I40

BARRY R. KOMISARUK

phenomena. Iri “Organization and Pathology of Thought: Selected Sources” (D.Rapaport, ed.), pp. 195-207. Columbia Univ. Press, New York. Sperling, S. K . 1981. “Tenderfeet and Ladyfingers: A Visceral Approach to Words and their Origins.” Viking, New York. Swadesh. M. 1971. “The Origin and Diversification of Language” ( J . Sherzer. ed.). p. 184. Aldine/Atherton, New York. Thass-Thienemann. T. 1968. “Symbolic Behavior. Washington Square Press, New York. Travell. J . 1960. Temporomandibular joint pain referred from muscles of the head and neck. J . Prosthe/. D e n / . 10, 745-763. Vandemark. N . L.. and Hays, R . L. 1952. Uterine motility responses to mating. Am. J . Physiol. 170, 518-521. Weiner, W. 1977. ”Psychobiology and Human Disease. Elsevier. Amsterdam. Wenger, M. A , , Averill. J . R., and Smith, D. D. B. 1968. Autonomic activity during sexual arousal. Psychophysiology 4, 468-478. Williams, R. L., and Krasnoff, A. G. 1964. Body image and physiological patterns in patients with peptic ulcer and rheumatoid arthritis. Psychosom. Med. 26, 701 -709. Wright, K . J . T. 1976. Metaphor and symptom: A study of integration and its failure. 1m. Rev. Pqcho-Anal. 3, 97-109. Zaks, M . G. 1958. ”The Motor Apparatus of the Mammary Gland.” Oliver and Boyd, London. Zuckernian, M. 1971. Physiological measures of sexual arousal in the human. P s y h o l . Bid/. 75, 297-329. ”

“

.

ADVANCES IN THE STUDY OF BEHAVIOR VOL. 12

Language in the Great Apes: A Critical Review CAROLYN A . RISTAU THE ROCKEFELLER UNIVERSITY N E W YORK. NEW YORK

DONALDROBBINS DIVISION OF SOCIAL SClENCES FORDHAM UNIVERSITY AT LINCOLN CENTER NEW YORK. NEW YORK

I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I1 . Brief History of the Ape Language Projects . . . . . . . . . . . . . . . . . . . . . . . . . . . 111. Theoretical Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A . Definitions of Language . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B . Can Apes Be Linguistic'? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. C . If Not Linguistic. What Then.?. . . . . . . . . . . . . . . . . . . . . . . . ............... D . Pragmatic Applications . . . . . . . . . . . . E . Summary . . . . . . . . . . . . ............................ IV . The Signing Apes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .......... A . Comparison of the Signing Projects . . . . . . . B . The Gardners' Projects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C . The Nim Project of Terrace and Associates . . . . . . . . . . . . . . . . . . . . . . . . D . The Gorilla Koko. Studied by Patterson . . . . . . . . . . . . . . . . . . . . . . . . . . . E . Other Ape Projects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . F . Concluding Comments about the Signing Projects . . . . . . . . . V . Artificial Lexicons . . . . . . . . . . . . . . . . . . . A . The Lana Project of Rumbaugh and Associates . . . . . . . . . . . . . . . . . . . . . B. The Premack Project . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI . Investigations into Meaning . . . . . . ............................ .......... A . Labeling . . . . . . . . . . . . . . . . . . . B. Novel Uses of a Word . . . . . . . C . Errors Made in Generalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D . Novel Word Combinations . . . . . . . . . . E . Feature Analysis of an Object and Its Label . . . . . . . . . . . . . . . . . . . . . . . . F . Functional Definitions of Words: A Look at Apes' Internal Images G . Categorical Sorting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII . Investigations into Mental States . . . . . . . . ............. VIII . Relation to Animal Cognition and Natural Animal Communication . . . . . . . . A . Relation to Animal Cognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. What Do Animals Naturally Communicate? . . . . . . . . . . . . . . . . . . . . . . . . C . What Aspects of the Apes' "Linguistic" Abilities Are Used in the Natural Environment'? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

142 143 145 145 151

153 154 154 155 155

162 167 171 172 173 178 178 186 188 189 192 192 193 196 I98 205 207 219 219 221 223

.

Copyright 6 1982 by Academic Ress Inc . All rights of reproduction in any fonn reserved. ISBN 0-12-004512-5

142

CAROLYN A. RISTAU A N D DONALD ROBBINS

D. Artificial “Language” Projects in Other Species . . . . . . . . . . . . . . . . . . . . E. Evolutionary Forces That May Have Apes or Early Man . . . . . . . F. Implications of Ape Natural Investigation into Artificial IX. Implications for Human Langu A. Have “Linguistic” Apes Met the Criteria for Human Language‘! . . . . . . B. Are the Apes’ “Linguistic-like’’ Behaviors Analogous or Homologous to Man’s? ............................... C. Implications for Human Children’s Communication . . . . . . . . . . . . . . . . . X. Problems Raised by the Language Research and Suggestions for Future Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. The Clever Hans Phenomenon B. Overly Critical Analysis of the Ape Language Research? . . . . . . . . . . . . . C. Suggestions for Future Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XI. Concluding Statements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

I.

221 228 228 230 23 1 234 236 237 237 239 240 245 247

INTRODUCTION

The ape language studies have captivated public and scientific interest. Reasons for the great curiosity about the subject seem to center on the issue of mental continuity. Do man’s cognitive and linguistic abilities, as well as his morphological structures, exhibit continuity with the nonhuman animals? The issue of Cartesian dualism is raised: Is man unique because he has a soul or mind, while the beasts do not; in what does man’s uniqueness lie? To many, to have a mind, or to be capable of thought, is dependent upon language and, indeed, to some philosophers, is identical to the ability to use language. If some nonhuman can be shown to have language, is man still unique? Man’s unique status did remain intact when other species were shown to use and make tools and to engage in cooperative hunting (Beck, 1980). Now even the issue of morality has been raised; for if an ape can converse with us and reveals understandings and emotions thought only to be human, should we not accord him the moral, perhaps legal, rights that we do humans? Is ape experimentation and confinement justifiable? Ideas such as these reveal many of the reasons for the great public interest in the ape language research. Although not all of these notions strike the scientific fancy, some do. The issue of mental continuity intrigues scientists and public alike, though a scientist may be more likely than a layman to realize that man can exhibit both continuity and uniqueness, for he as well as any other species represents a unique adaptation. The issue of the origins of human language has also stimulated attempts to teach apes linguistic skills in order to reveal the developmental process of language acquisition in humans under conditions more controlled than those feasible with

LANGUAGE IN THE GREAT APES

143

humans. Some hope the research will isolate the basic cognitive capacities or atomic skills that might be the evolutionary precursors to human linguistic skills (McNeill, 1974) or that may constitute language behavior and intelligence (Premack, 1976). Others (Rumbaugh, 1981) cite the pragmatic application of techniques from their own ape language projects to teaching communicative skills to severely mentally deficient children. It may be that yet other consequences are, in the long term, the most important contributions from this research. In this review we will attempt a history of the ape language research, indicating past and current work. We will present a critique of the research methodologies and of the interpretations offered. We will attempt to raise some relevant theoretical issues and begin to relate the research to investigations or issues in the fields of experimental and developmental psychology, linguistic analysis, anthropology, and ethology. Finally, we will suggest some problems for future research.

11.

BRIEFHISTORY OF

THE

APE LANGUAGE PROJECTS

Early attempts were made to teach apes to produce words vocally. Near the beginning of the century, Furness (1916) taught a female orangutan to produce vocally, pupa, and cup, and th over 11 months of instruction. The Hayes (C. Hayes, 195I ; K . Hayes and C. Hayes, 1951), after 6 years of prodigious training of their home-raised chimpanzee, Viki, managed to teach her to produce four raspy single utterances, mama, pupa, up, and cup. The Kelloggs’ (Kellogg and Kellogg, 1933) home-raised chimpanzee, Gua, could not, despite their intensive efforts, produce a single spoken word. Laidler (1978) trained an infant male orangutan, using operant conditioning techniques modeled after those used successfully with autistic children (Hebett, 1965). In the course of 9 months of training, the orangutan was able to produce four sounds, kuh, puh,fuh,and h u h , and to use them appropriately with presumed meanings of milk and other beverages in a mug (kuh), contact-comfort (puh), pan and solid food (fuh), and brushing continuation (rhuh).The experiment then had to be terminated, although the orangutan seemed at this point to be able to learn new “words” very rapidly, i.e., in 1 week or less. His competency, in a variety of contexts, was better than 70% appropriate responding. Subsequent to the initial interest in vocal training, researchers investigated whether the difficulty lay within the structure, mobility, and peripheral or central nervous control of the vocal apparatus. Chimpanzees, in contrast to humans, appear to lack precise muscular control over the vocal apparatus and in particular the larynx, possibly accounting for the apparent inability of the apes to speak language; alternatively apes may lack necessary cortical association areas (see Section 111). Yet, as many have noted (Savage-Rumbaugh et al., 1977; van

144

CAROLYN A . RISTAU A N D DONALD ROBBINS

Lawick-Goodall, 1968, 1973), chimpanzees use manual gestures in their natural communication and have great manual dexterity. I n fact, over 50 years ago, Yerkes and Yerkes (1929) had guessed that the great apes could learn a nonvocal language. The Gardners (R. A. Gardner and B. T. Gardner, 1969) made the pioneering effort to teach hand signs to their home-reared chimpanzee, Washoe; and she did, indeed, apparently master production of 132 signs after about 4 years of training. [The Gardners, and subsequent researchers who teach handsigning, use American Sign Language (Ameslan or ASL), or more precisely, a simpler or pidgin version of ASL in which many spatial nuances are lost.] Fouts, a student of the Gardners, has been attempting to encourage chimpanzees to communicate with each other by hand signs and to teach Ameslan to each other (Fouts, 1974). Patterson, working with young gorillas, specifically uses both spoken English and hand signs (Patterson, 1978a,b). The chimpanzee Nim has been taught hand signing by Terrace and his trainers (Terrace, 1979b; Terrace et a l . , 1980). They have employed videotape extensively in the data collection so as to gather a large corpus of the chimpanzee’s signings and to compare his signs with those of the human trainers. An orangutan has been the subject for Lyn Miles in her project with hand signing. An important similarity between the hand signing projects is that the ape was generally raised in a home environment for a considerable period of his training. Except for Patterson, and Fouts in one instance, the investigators have typically attempted to avoid the use of spoken English in their training procedures. The techniques used for teaching Ameslan include molding and prompting. Other approaches have also been used. Artificial visual designs were employed instead of signing in order to permit more complete and systematic data collection than the sign language projects offer. Rumbaugh (1977), working with the chimpanzee Lana, used a computer-based lexigram system. Lexigrams are visual designs displayed on individual keys on a keyboard that the chimpanzee could depress; a screen above the keyboard displayed the productions of the chimpanzee or a human trainer. Premack (1976) taught several chimpanzees, in particular a very able pupil, Sarah, a plastic chip system in which variously shaped and colored pieces of plastic were placed on a board; these chips could then be rearranged by Sarah or the trainer to serve as communications between them. At an early stage of his research, Premack indicated that a major goal of his work was to investigate basic components of intelligence in the chimpanzee as opposed to a major emphasis on the ability of a chimpanzee to learn humanlike language. More recent developments from these initial efforts are simultaneously more basic and more speculative, including investigations by Premack and Woodruff (1978a,b; Woodruff and Premack, 1979) into the nature of mental states of their primate subjects, such as beliefs, beliefs about beliefs, and lying. SavageRumbaugh et al. (1978a,b) have investigated the apes’ understanding of a

LANGUAGE IN THE GREAT APES

145

“word” through experiments with naming and using food and tools to obtain food; they have also attempted to study two-way communication between chimpanzees. Except for Fouts’ attempts, all the other ape language work involved communication between an ape and a human or a human’s creation, namely a computer program. Criticisms of the appropriate interpretations of apes’ linguistic achievements are becoming widespread. Philosophers, experimental and developmental psychologists, biologists, and on occasion a linguist have offered their own analyses, e.g., in 1978 in The Behavioral and Brain Sciences with open peer commentaries.

111.

A.

THEORETICAL ISSUES

DEFINITIONS OF LANGUAGE

An initial issue raised by the research is the definition of language and the kinds of evidence we need to demonstrate that a human or other animal exhibits language. Views on the definition of language are varied. Let us begin by realizing that no one yet has an adequate definition. On the one extreme stand the views of the linguist Noam Chomsky: To determine whether music, or mathematics, or the communication system of bees, or the system of ape calls, is a language. we must first be told what it is to count as a language. If by larigitage is meant human Imguage the answer will be trivially negative in all thew cases. If by lotrguage we mean symbolic sysrem, or sysrem of rommunicarion, then all these examples will be languages, as will numerous other s y s t e m s 4 . g . . style of walking which is in some respects a conventional culturally determined system used to communicate attitude, etc. If something else is intended, i t must be clarified before inquiry can begin (Chomsky, 1980, p . 430).

In Chomsky’s view, a necessary characteristic of language is the existence of grmnmar. He argues that all human languages have a system of rules, the

grammar, that specify the properties of expressions of that language. This finite set of rules can generate an infinity of expressions. He further argues that a universal gramrnur exists-this refers to a set of principles that all grammars must conform to as a matter of biological necessity. A crucial characteristic of the grammars is principles for constructing a hierarchy of phrases. Recursive embedding is one basic device for creating new phrases. For example, he ate the food can be embedded within a verb phrase such as to get the food that he ate, and so forth. In general, except for a small segment of Premack’s work with Sarah, the ape language projects have not dealt with phrase hierarchies, for most of the ape’s utterances are simply too short.

146

CAROLYN A. RISTAU AND DONALD ROBBINS

Chomsky notes that further rules specify the phonetic and semantic representations of these structures. “These devices provide for the range of expression characteristic of all human languages, allowing us to denote previously unexamined or newly imagined objects, actions, properties, events, etc., and to form propositions of various sorts. These are the most basic and elementary properties of human language” (Chomsky, 1980, p. 431). Chomsky addresses the issue of the function of human language by pointing out that Human language is characteristically used for free expression of thought, for establishing social relations, for communication of information, for clarifying one’s ideas, and in numerous other ways. While some describe its essential purpose as “communication,” there is, to my knowledge, no substantive formulation of this proposal with empirical content; it can be sustained only if the term “communication” is used in so loose a sense as to deprive the proposal of any interest. Cmcially, there is no basis for the belief that human language is used “essentially” for “instrumental ends”-to obtain some benefit (Chomsky, 1980, p. 433).

Other linguists have emphasized that language is characterized as well by creativity, and by the use of metaphor including insults and curses. It is for such reasons in particular that so much attention has been paid to the anecdotal descriptions of chimpanzees’ apparent creative expressions, e.g., calling a swan a water bird and apparently denouncing their trainers at times with statements as Dirty Roger (Fouts, 1974). Well before the ape language projects and controversies, the possible differences between human and animal communication had been studied. Hockett enumerated a set of 16 criteria that appeared to be characteristic only of human language; since that time several of these characteristics have been demonstrated to be attributes of some natural animal communication or of specifically trained artificial ape communication (Hockett, 1958; Hockett and Altmann, 1968; Thorpe, 1974, p. 13). Following are the design features proposed by Hockett. 1. Vocal-auditory channel, The restriction of linguistic communication to a single channel allows the communicator to engage in other activities, and, in particular, in nonverbal communication. Furthermore, since not all vocal characteristics are linguistic, prosody and other characteristics of the acoustic productions can be used to communicate paralinguistically. 2. Broadcast transmission and directional reception. 3. Rapid fading. A spoken message quickly fades, unlike other communications such as those produced by pheromones and the information purposely or inadvertently left by animal tracks. 4. Interchangeability. Interchangeability refers to role reversal of communicators, namely that adult language users are both transmitters and receivers of language.

LANGUAGE IN THE GREAT APES

I47

5 . Complete feedback. Such feedback implies the speaker hears everything relevant in what he says. As Thorpe (1974) suggests, the significance of features (4) and (5) become apparent through comparisons with animal communication systems, e.g., bird song is in nature most typically a male song, thereby not interchangeable. 6. Specialization. The impact of linguistic signals rather than their energy expenditure is significant. 7. Semanticity . This refers to the associative ties between signal elements and features in the external world, i.e., some linguistic elements have denotations. Thorpe considers that many species communication systems have semanticity. Most researchers of animal communication would probably not agree that semanticity has as yet been demonstrated to be so widespread, although there is great interest in searching for this. Thorpe seems to be using the term semanticity in a much more general sense. For example, he considers many alarm calls to be semantic even though a specific call may be made in the presence of any sort of predator, loud noise, or other apparently threatening object or event. 8. Arbitrariness. A word is not an icon; it does not bear any physical resemblance to its denotation. There are some few exceptions such as onomatopoetic words, e.g., buzz, for which the acoustic structures of the word and its denotation may be similar. 9. Discreteness. Unlike most other primate communication, the lexicon of human language is discrete or digital rather than continuous or analog. 10. Displacement. Signals can refer to events remote in time and space. Again, as Thorpe suggests, the difference between human language and animal communication may reflect a quantitative difference in the remoteness of time and space. 1 1 . Openness. This criterion of openness or productivity has become an important matter in the ape language projects. Openness implies that new messages may be created frequently and communicators can talk about things never talked about before. Yet a bee dance can describe locations never before visited and can refer to needs of the bees such as a proposed location for a new hive, never before encountered in the lifetimes of the communicators. Such needs have been “discussed” over evolutionary times, however, and this may be a constraint on the notion of openness. Truly novel challenges, if such could be devised, could provide a better test for some degree of openness. 12. Tradition. Through teaching and learning, specific items of a lexicon and conditions for their use may be passed from one group to another or from one generation to another. 13. Duality of patterning. Although elements of the signaling system, e.g., the phonemes in human language, may be meaningless, patterned combinations of them are meaningful, such as words and patterns of words in sentences.

The following design features were added (Hockett and Altmann, 1968) to the original 13.

148

CAROLYN A . RISTAU AND DONALD ROBBINS

14. Prevarication. This refers to the ability to lie intentionally. Humans clearly have this ability and some animals exhibit deceptive behaviors in the field, such as injury feigning by some birds. Experimental investigation of this issue has been part of the ape language projects (Premack and Woodruff, 1978b; Woodruff and Premack, 1979). This topic is discussed in Sections VIII and IX. 15. Reflectiveness. The ability to communicate about the communication system itself is considered by Thorpe to be peculiar to human speech, yet recent experimentation in the ape language projects, e.g., the chimpanzees’ use of metalinguistic forms such as name-of (Premack, 1976; Rumbaugh and Gill, 1977) and the chimpanzee Sarah’s apparent descriptions of the meaning of a plastic chip (Premack, 1976) may provide the beginnings of intriguing counterexamples. 16. Learnability. The speaker of one language can learn another; by implication, the speaker of a language was able to learn that language as opposed to being innately endowed with language. Learnability does not necessarily preclude the existence of any innate aspects of language. At this point almost every design feature, with the exception of reflectiveness, is considered by at least some researchers of animal communication to be a characteristic of at least some system of animal communication. Such systems include those of crickets and grasshoppers, honeybee dances, various species of birds including the mynahs, colony-nesting seabirds, primate vocalizing, and the nonvocal communication of Canidae (Thorpe, 1974). Thorpe seems particularly generous in applying these design features in a very general way to animal communication and we would not agree with all of his attributions. However, a thorough discussion of the relation between the design features and animal communication is beyond the scope of this article. In general, even as Thorpe himself suggests, it seems most reasonable not to apply these features in an all-or-none fashion to animal communication or, in fact, to the ape artificial language abilities, but rather to search for precursors and “levels” of the design features as they apply to animal communication. For example, there are numerous forms of prevarication, some more simple than others. One should try to define the attributes of the various sorts of lying, possibly by gathering insights from the development of lying in children, and then seek to discover which animals, under what conditions, exhibit the various abilities. (This issue is discussed further in Sections VIII and IX.) Despite the relevance of linguistic issues to the ape language projects, to our knowledge, only two of the projects have had linguists associated with them. Ernst von Glasersfeld collaborated with the Lana project, and Thomas Bever with the Nim project and then predominantly for the beginning stages of the endeavors. It is unfortunate that these associations were apparently rather brief; if differences arose of theoretical interest, it could be useful to express these to

LANGUAGE IN T H E GREAT APES

149

psychologists and ethologists, who are the main scientific audience for the ape language results. Von Glasersfeld did enumerate three critical characteristics of linguistic communication: ( 1 ) A set or lexicon of arbitrary signs; (2) the symbolic use of those signs; and (3) combinatorial patterning of the signs (von Glasersfeld, 1977). These are, you may note, a subset of Hockett’s criteria and include some of characteristics noted by Chomsky. The most extreme view on the other end of this spectrum is probably represented by B . F. Skinner’s (1957) analysis of human language as verbal behavior that is reinforced. Just as predictions can be made for nonverbal responses, so they could for verbal responses or operants. There were several kinds of verbal operants, the tact, the mand, and echoic response being essential to verbal development. An example of a tact is a child naming an object; in Skinner’s terms the discriminative stimulus for that tact is the object. A mand is a demand; the discriminative stimulus is a deprivation condition, such as hunger or thirst, to which a child might say cookie or juice. Echoic responses mimic sounds that are heard. The community, parents in particular, reinforces the verbal operants until the child has a large repertoire. Eventually, the child begins to build up sentences which are composed of intraverbal associations, i.e., combinations predictable by linear associations as opposed to hierarchical relations. According to Skinner’s analysis, the sentence is not conceived of as a thought. A recent experiment (Epstein et al., 1980) applies this analysis to a demonstration of sequential key-pecking by two pigeons. In a witty and provocative article, the authors attempt to dismiss much of the ape language research results as illustrations of nothing more than operant behavior achievable in the Skinner box, which, by implication, has probably already been achieved in more species than apes and the pigeons in their experiment. (Alternatively, the “mere” pigeons’ behavior may reflect more cognition than the operant conditioner is granting them.) For all their satire, the authors also seem to believe that their analyses are relevant to human language learning as well. Skinnerian concepts may apply to some of the ape research work and to achievements of some severely aphasic humans, even perhaps to some initial stages of second-language learning, as by tourists learning a few idioms for the purposes of more convenient travel. However, Skinnerian analysis simply does not appear to be fruitful for the major part of human language; Chomsky discussed such concerns in 1959. Considering these views and those of others who have written on the subject, the issues that seem to be most central to a discussion of the definition of language appear to be the capacity for a generative grammar and the meaning of an utterance. A capacity for generative grammar is demonstrated by the use of rules in constructing an utterance. Lenneberg (1969, p. 83) notes that a capacity for generative grammar is not demonstrated merely by a subject’s combining words. The word generative in grammar is an abstract metaphor and denotes “principles

150

CAROLYN A. RISTAU AND DONALD ROBBINS

that account for something. As an example, the principle of random selection could generate tables of random numbers. To evaluate the word output of animate or inanimate objects, we would want to know the principles that underlie the productions. Applied to the ape language projects, Lenneberg’s argument would require that researchers do more than enumerate the word combinations made by the subjects, if, indeed, such enumeration can be done. The meaning of an utterance refers both to the conceptual complexity of the individual items in the lexicon, e.g., the meaning of a word, the ability to use a word as a metaphor, and to the meaning of the phrase or proposition. Ryle (1949) has distinguished between “knowing how” and “knowing that.” As used by the philosophers, this distinction seems to be the central issue in various psychologists’ criticisms that ape language usage is instrumental and related merely to the gaining of reward, i.e., the ape is demonstrating “knowing how” to get a reward but does not understand what he is “saying.” The distinction between “knowing how” and “knowing that” seems intuitively accurate for certain kinds of descriptions. For example, the accomplished pianist “knows how” to play a familiar piece of music, yet, typically, he is not, at the time, aware that his fingers are moving in a specific sequence, pausing prescribed units of time. Later, he can “know that” his fingers were moving in particular ways by mentally “replaying” the music, possibly even making abbreviated finger motions as he does so. We recognize that the distinction between “knowing how” and “knowing that” is not absolute; there are, indeed, prenumbral cases as philosophers have noted [e.g., Dennett (1978a, p. 184)]. There are also different kinds of “knowing how, ranging from automatous walking upstairs to skilled mountain climbing. Arguments, beyond the scope of this article, have been offered in which “knowing how” can be reduced to “knowing that.” Yet the distinction may nevertheless be useful in discussing the ape language research, insofar as it illuminates the difference between rote performance to obtain reward vs knowledge that entails awareness both of the behavior as communicative and of the possible outcomes, and ascribes meaning to the elements of the behavior. Another analysis that can be applied, perhaps more fruitfully, to the accomplishments of the apes and, indeed, to natural animal and human communication has been described by Dennett (1978a,b); this structure is intentional analysis, deriving from analyses offered by Grice (1967) and others such as Bennett (1976, 1978). Philosophers’ use of “intentionality” is to be distinguished from the everyday term “intentionality,” the latter meaning “on purpose. An intentional or unintentional act can be subjected to the logical “intentional” analysis, which is a means of describing the degree of complexity of mental systems. A few examples of zero, first, second, and third order intentional analyses follow. ”

”

”

Zero order: A locomotes; A makes certain gestures under specifiable stimulus codilions. First order: A knows rhar p , where p is some proposition; A wants q .

LANGUAGE I N THE GREAT APES

151

Second order: A believes that B believes that p . Third order: A wants B to believe that A wants x .

Such an analysis is useful in suggesting experiments and in examining ape language experiments that have been conducted in order to better understand the level of mental complexity implied by various ape behaviors. It is beyond the scope of this article to attempt to apply an intentional analysis to the ape “language” work, but Dennett (1978b) has begun to do so. B.

CANAPESBE LINGUISTIC?

If the ability to be linguistic requires, literally, the ability to make vocal linguistic productions, then quite clearly the apes cannot be lingusitic. Work by Lieberman and others (Lieberman, 1968; Lieberman et al., 1972) indicates that the chimpanzee lacks the structures necessary to produce the variety of sounds that Homo sapiens can. However, this is not without some controversy since Lieberman et al. (1972) argue that the deficiency is due primarily to a difference in the vocal tract, while Steklis and Raleigh (1979) argue that it is caused mainly by the nonhuman primates’ lack of cortical association areas to support the speech motor mechanism. Lieberman (1975) argues further that speech as a means of communication may have developed fairly recently, less than ?4million years ago. In his view, reconstructions of the oral cavities of early hominids reveal that these protohumans did not have the vocal tract shape needed to produce the sound variations used in modem human languages. Furthermore, he argues, not only were the reconstructed vocal tracts of Australopithecus and Neanderthal man inadequate for speech as we know it, but they were also very similar to the structures of the modern chimpanzee. Of course, it can always be stated that, with soft structures like the tongue and larynx, modeling entails far more speculation than do reconstructions of foot bone. It can also be argued that anatomy does not necessarily predict behavior; witness the parrot who should not, by its anatomy, be able to speak (or more precisely, make speechlike sounds). That the chimpanzee can and does produce a variety of sounds in its natural communication is not disputable. It has not yet been demonstrated whether chimpanzees can learn to use, in language-like ways, some vocalizations that are easy for a chimpanzee to make or some nonvocal acoustic productions via a musical instrument or other sound-making device. At least one chimpanzee has been observed to incorporate noise-making devices into the natural communication system. Mike, of Tanzania’s Gombe Stream Preserve, banged on discarded kerosene cans, so improving the effectiveness of his charging display that he thereby enhanced his social status over that of previously dominant males (van Lawick-Goodall, 1971, pp. 112-1 17, as noted by Griffin, 1981). In any event, in view of failures to teach apes spoken language, it can be said that it is strongly

152

CAROLYN A . RISTAU A N D DONALD ROBBINS

doubtful that the chimpanzee has the ability to exercise the necessary voluntary vocal of the vocal apparatus to produce sounds similar to those of human speech. But almost no one requires that utterances be vocal in order to qualify as language. However, some linguists have suggested that the acoustic characteristics of speech bear a critical relation to the characteristics of human language (Lenneberg, 1967; Lieberman, 1967; Healy, 1973). Spoken language is necessarily sequential and each unit exists but briefly. In particular, linguists cite the need for a perceptual ability to analyze the flow of auditLiry stimuli into individual phonemes. Phonemes are the sound units which constitute spoken language. Any one language uses between about 35 and 50 phonemes, while humans in general seem to be able to produce and discriminate about 100 such sounds. The phonemes are combined according to rules to produce words, which in turn are combined according to syntactic rules to produce propositions or sentences. Healy (1973) notes that the signing projects (and those using artificial lexicons) were more successful than the Hayes’ initial attempts, perhaps because the recent projects utilized words rather than phonemes as the basic units. Premack and Schwartz (1966) and Premack (1971) described phonemic systems suitable for the chimpanzee, but to date there are no reported successful attempts to teach chimpanzees such systems. The signs of ASL have been analyzed into elements called cheremes, which Stokoe (1960) contends correspond to the phonemes of spoken language. “However, the cheremes of ASL, which specify the place, action, or configuration of the hands used in making a sign, are more closely analogous to distinctive features (Jakobson et a l . , 1969) of the gestural signs than to phonemes” (Healy, 1973, p. 142). If, the vocal issue notwithstanding, we wish to demonstrate linguistic competence in an ape, we must contend with the problem of translating various characteristics of language into testable notions and appropriately controlled scientific experiments. Lenneberg (1971) argued ihat five demonstrations are required in order to show that apes have acquired even rudimentary aspects of language. ( I ) The “language” itself must be evaluated in human-to-human communication to determine the capacities of the created communicating system. (2) If ( I ) is satisfied, then a similar test should be conducted between human and ape. He proposes the use of four human observers: 0, writes out questions; 0, would translate them to signs and sign to the ape; 0, watches 0, and writes down what O2 is signing; and 0, records the ape’s reactions (and does not know what O,, O,, or 0, did). Further, he notes that O4 should not see the ape until O2 is finished. (3) Test apes’ knowledge of the language by using two-choice (yesho) tests of questions about the language. (4) One must demonstrate that the basic aspects of every language are preserved in the “language” taught the apes. (5) A test must be made for productivity, i.e.: the ape must be able to answer any questions or execute any command that is possible for the language. Criteria (4) and (5) are somewhat vague, for the definition of “basic aspects of every language” is

LANGUAGE IN THE GREAT APES

I53

unclear and limits are necessary in reasonable tests of productivity. We can wonder whether all humans would pass these criteria. The criteria pose formidable demands upon the experimenter and the ape and tests have not yet been undertaken, although one might argue that some experiments meet some of the criteria. C.

I F NOT LINGUISTIC, WHATTHEN?

Let us consider the possibility that the ape’s performance in the language projects cannot be considered to be linguistic. What knowledge can we nevertheless gain from these endeavors? At least some of the projects, in particular that of Premack (1976). can possibly reveal some of the cognitive capacities of the chimpanzee. In fact, Premack considers that his plastic chip “words” are mapping preexisting cognitive capacities. Others (Terrace, 1979a) interpret the chimpanzee behavior as problem-solving abilities that have been developed through the specific experimental paradigms used. Whether the chimpanzees in their natural environment use some of these capacities, such as the concept of “same-different” or “name-of,” remains a moot point. Quite likely there are levels of understanding of these concepts such that a chimpanzee, or possibly even a young child, may use them appropriately in some circumstances. In other situations the immediacy of desires (e.g., food) or distracting stimuli may overwhelm the appropriate use of those concepts. Previous research by experimental psychologists on the cognitive abilities of apes and monkeys should be related to the ape language work to provide a fuller description of the apes’ abilities. Studying how the apes acquire their ‘‘linguistic” abilities might reveal cognitive processes, although none but recent research into word acquisition (Savage-Rumbaugh er ul., 1978a,b, 1980b) is specifically a study of such development. The Gardners, in their concentration on the acquisition of signs by chimpanzees, have suggested that just as children develop certain functional types of words before others (Nelson, 1973), so the young chimpanzee, exposed to a variety of signs, first learns certain word classes preferentially to others and then gradually expands the categories of words learned. It is, however, very difficult to determine whether, to the chimpanzee, a sign belongs to a specific semantic category (e.g., verb, noun) or what that “word” means to him. Some effort has been made on this task (R. A. Gardner and Gardner, 1969; B . T . Gardner and Gardner, 1975; Savage-Rumbaugh et ul., 1978b, 1980b), but not as much as is warranted at this basic level of understanding. A careful look at specific difficulties chimpanzees encounter with certain kinds of learning or training methods might be used to reveal possible species-specific proclivities for conceptual learning. For example, at least for their particular paradigms, Savage-Rumbaugh er al. note ape tendencies to associate lexigrams with location. Premack’s studies often indicate chimpanzee errors which may contain patterns, although they are not obvious ones.

154

CAROLYN A . RISTAU A N D DONALD ROBBINS

Some suggest that the ape language work may have implications for understanding human cognitive development and capacities (discussed further in Section IX). The most relevant areas would seem to be determining basic or “atomic” cognitive skills that may underlie linguistic skills (Premack, 1976), and becoming sensitive to ‘‘levels’’ of comprehension entailed, e.g., in the understanding of a word. One might then be able to determine if similar categories were appropriate in understanding the emergence of a young human’s cognitive capacities. Possibly the most important impact of the ape language studies has been in provoking research into the developmental characteristics of vocal and sign language as the young child develops these abilities. D.

PRAGMATIC APPLICATIONS

The computer-based Lana project was suggested as a potential model for teaching language to severely mentally retarded or aphasic human children (Parkel and Smith, 1979). Those techniques have borne fruitful results in at least some cases, although published details of the work are lacking. A project at the Georgia Retardation Center, Atlanta, Georgia, was developed by Dorothy Parkel, Royce White, and Caren Miller working with children of mental ages 3 years and less. The children were initially able to produce merely a few utterances decodable only by persons very familiar with them. Rumbaugh (1981) describes that five of nine such children could, through training, learn the meanings of arbitrary symbols and then use these symbols to engage in novel conversation with adults. The children’s general social skills improved greatly and behavioral problems decreased. It is unfortunate that such applications have not been conducted with proper control procedures. As a result, it is very difficult to determine which aspects of the procedure are effective or if, indeed, the extra individual attention given each child has had the predominant positive effect.

E. SUMMARY Definitions of language still remain ambiguous, although the capacity for a generative grammar and the meaning of a word and an utterance remain as central issues in most discussions. An important concept to be noted when determining the meaning of an utterance is the distinction between “knowing how” and “knowing that. ” Although the skills revealed in ape language projects should be considered with respect to human language, it is clear, from theoretical considerations, that the more general cognitive aspects of those skills are likewise of interest. The impact of the ape language research on investigations into children’s linguistic development is a valuable one. There may be useful pragmatic applications for aphasics of techniques utilized in the ape language projects.

LANGUAGE IN THE GREAT APES

155

IV. THESIGNING APES A.

COMPARISON OF THE SIGNING PROJECTS

I.

Subjecrs

The signing projects share a number of common properties, which are outlined in Table I. The apes were typically chimpanzees (B. T. Gardner and Gardner, 1971, 1975; R. A. Gardner and Gardner, 1969, 1975, 1980; Fouts, 1972, 1973; Terrace, 1979b; Terrace et a l . , 1979a,b), although a gorilla (Patterson, 1978a,b) and an orangutan (H. L. W. Miles, personal communication), have also been studied. Subjects were typically, but not always, female. 2.

Language Media

Major training usually began when the apes were about 1 year old, although the researchers recognized that even younger ages might have been more appropriate. In fact, Terrace’s and the Gardners’ most recent project did expose very young apes to signing. The investigators used the sign language, Ameslan (ASL), with local variations, and curtailed spoken English during the training sessions. There were two exceptions: Patterson purposely had the trainers do “simultaneous communication” using hand signs and spoken English with her gorillas, and Fouts (1974) used a similar method with one chimpanzee (Ally). 3 . Living and Training Environments Since the formality of the training sessions differed among the projects, exposure to spoken English and the variety of contexts in which sign language was used also differed. In all the projects, the living environment served as the training environment, and included, to varying degrees, access to the outdoors and special excursions. The Terrace project later emphasized classroom training. 4.

Problems in the Training of Ameslan

In much of this work, the initial trainers were novices at signing, resulting in inconsistent use of signs or signing sequences by an individual trainer and across trainers. The Gardners’ recent project overcame this difficulty with subsequent ape subjects by using fluent native ASL signers, and Patterson utilized some fluent ASL trainers throughout her project. In cases where nonfluent human signers were used, their lack of proficiency in the language makes it almost impossible to draw any conclusions about the sign sequences to which the subjects were exposed. It is therefore not clear whether signing sequences used by the apes are indicative of their potential grammatical capability. “Incorrect” orders may simply reflect what the apes saw humans signing. However, some of

TABLE I AN OVERVIEW OF RECENTAPE LANGUAGE PROJECTS: THE SIGNING APES Principal Investigator(s)“ Gardner and Gardner I

VI

OI

Gardner and Gardner I1

Terrace

Fouts

Patterson

Principal investigatotfs) ’ major interests regarding the project

Two-way communication between human and apes

Develop and describe “verbal” communication from birth on to determine limits and extent of chimpanzee linguistic capacities

Initially to amass corpus of “utterances” to evaluate linguistically. What features of a natural (human?) language can chimps master?

Communication between chimpanzees using signs. Can apes learn a language?

Investigation of linguistic capacities in the gorilla. Comparative cognitive abilities among apes and humans

Species

Pan troglodytes

Pan

Pan

Pan

Gorilla gorilla

Principal Washoe ( P ); subject(s)’ acquired and sex, age when training begun acquired, and when at 8-14 months training began old

Moja (9).Tatu ( 9 ) . and Nim (8); acquired and exposure to ASL Peli ( 6 )(died pnor to second birthday), and begun at 2 weeks old; Dar ( 6 ) ;all acquired formal classroom and exposure to ASL training from age of begun at 1-4 days 10 months old

Booee ( 6 ) . 3-year-old Koko ( P ), born in born in captivity; captivity: begun cerebral hemisat approx. 1 year of age. phere and spine surgically split Michael (d); begun except for medulla. at approx. 3.5 years old Ally ( d ) , 3 years old, born in captivity Bruno ( 6 ) . 32 months old, born in captivity; Lucy ( ? ), 7 years old; Cindy ( ? ), 45-51 months old, African born; Thelma ( ? ), 33-39 months old, African born Training begun at ages indicated

-

Living environment

House trailer with access to inside and fencedin yard

Home-reared

Home-reared and then homelike environment (with humans)

Training environment

Living environment

Living environment

Language medium

Manual signing

Manual signing

Living environment and small classroom at Columbia University Manual signing

A few principal trainers and student volunteers Most not fluent in ASL

A few project research assistants (at least six on any one day); total not specified Many deaf and fluent signers, research assistants, and others fluent in ASL

Trainers Number

VI 4

Qualifications

Major training methods

Molding, modeling, prompting, and observation of signing; no specific training for utterance size

Combinations of home-reared and partial home environment and caged at Institute for Primate Studies, University of Oklahoma Living environment

Zoo nursery, then five-room trailer; limited access to outside and some rooms; total access to kitchen and living room Living environment

Manual signing (Ally also heard spoken English)

Manual signing and spoken English ( “simultaneous communication”)

Sixty teachers, most Some principal of whom were “occasional trainers and other of brief durations playmates”; core group of eight teachers Many student volMost not fluent unteers; most not in ASL fluent in ASL; core group has I + years ASL training Molding, modeling, Molding, modeling, Molding, modeland prompting ing, and prompting prompting, and observation of signing; Used fluent native signers as human models

One primary trainer and 14 assistants Nine deaf or native signers

Molding, modeling, and prompting

Continued

TABLE 1 (Continued) Principal investigatofls)“ Gardner and Gardner I Rewards

Major testing methods

Other datacollection methods

Reliability tests

Social interactions, food, object, or event being taught was often “delivered” Vocabulary tests; “Wh” questions

Tape-recorded “comprehensive ” samples; videotape; written observations of sign and context Double-blind test of sign knowledge for sample of objects; use of naive (with regard to Washoe) native signers; interjudge reliability

Gardner and Gardner I1

Terrace

Fouts

Same as Gardner and Gardner I

Social. food

Social, food (a raisin was often given)

Formal vocabulary tests

“Wh” questions

Formal vocabulary tests

Diary records, Videotape; teacher reports; written inventory of phrases, samples of “verbal” observations of sign and context input and output; method of “obligatory contexts” Comparison of teacher Double-blind; independent data recorder reports: 77% and trainer for samples agreement before teachers consulted of interchanges; inone another, 94% tequage reliability after

Recording of outputs and contexts

Double-blind tests (83-1008. i=%%

agreement)

Patterson Social; object or event being taught or topic of conversation often * ’delivered ’ ’ “Wh” questions: sign alone, spoken English alone; infant intelligence test Videotape; daily diary; taperecorded sampling of signing and context Double-blind tests; Koko often ’‘avoided’ ’ task; overall performance = 60% correct

Criteria for acquisition of vocabulary item

jize of expressive vocabulary

c

4bilities and major results claimed by researchers

Three independent obserSame as Gardner Occur at least Identified by at vers note spontaneous and Gardner I once ”spontaneleast three of the occurrence; then sign ously” and approprifluent human must occur spontaneously ately on each of 15 companions, consecutive days; followed by 15 on 5 successive days; “spontaneous” = sign in “spontaneous” = consecutive days an appropriate context occur without prompt of spontaneous without aid or in response to appropriate usage What this? 132 signs after 125 signs after Approximately 40-70+ F i s t signs appear 51 months of at 3 months of age; approximately 45 months signs after 2-4+ training accepted “immature” of training years of training variants of signs; over 50 signs within-2 years Performance in Multisign utterances; Spontaneous Abstract use of no; vocabulary tests M.L.U. remains at 1.6 appropriate use f i s t 50-sign vocabranged from 26 signdutterance last of combinations ularies not different and phrases; from first 50-word 1.5 years; increase in to 90% correct; signs to self; vocabularies of length did not add communication overgeneralizaapproximately 2-yearinformation, e.g., Play me between signing old humans; new word tion to novel chimps; novel to Play me Nim; combinations“discourse” analysis objects and sitcombinations (all of videotape transe.g., Washoe: uations; word Moja), e.g., alka cripts showed signs water-bird creation seltzer = listendrink for swan often imitative of teacher’s prior sign

Insufficient information was made available to us to be able to include Lyn Miles’ work with an orangutan, Pongo pygmaeus bPersonal communication. Most recent published paper states over 200 signs after 3.5 years of training. =Determined from graph, p. 84, in Patterson and Linden (1981).

Recorded by two independent observers and be used spontaneously and appropriately on at least half the days of a given month 264 signs after 66 months of trainingb; approximately 185 signs by Gardners’ criteria‘ Engages in deceit; signs to self; overgeneralization to novel objects and situations; word creation; novel combinationse.g., cookie rock for stale roll

1 60

CAROLYN A. RISTAU A N D DONALD ROBBINS

the highly repetitive and peculiar sequences, e.g., those analyzed by Terrace et al. (1979a,b), seem unlike a sequence humans might use and more indicative of the apes’ own emphasis. Yet it is conceivable that during extensive drilling sessions, such as those of the Terrace (and other) projects, highly repetitive signing could be the norm even for humans. No one, including Terrace, has investigated this. During training, the apes observed the trainers’ signing, and the trainers molded and prompted signs by the apes. Molding means that a trainer took the hand of the ape and adjusted it to the correct form of the sign. Trainers prompted an ape by repeating a question several times or themselves making the sign to be imitated. Correct signing was treated by the trainer with social rewards, such as smiles, praises, hugs, and other exclamations of delight, by receipt of a requested item, and by food rewards. More recently Miles has been teaching signing to an orangutan without food reward, unless food is requested by the ape (H. L. W . Miles, personal communication). In general, there was no specific training aimed at increasing utterance size, and short utterances were as acceptable as long utterances. Although we recall no specific mention of this, it does seem likely that trainers, committed to encouraging ape linguistic skills, would unwittingly reward longer utterances by spontaneously expressing pleasure when such sequences occurred. It should be noted that a number of signs in Ameslan are iconic, i.e., the sign is similar in form to or imitates the object or action which is signed about. Some have criticized the iconicity of the ASL signs (Savage-Rumbaugh er a / . , 1978b), claiming that the apes’ use of iconic signs does not reveal referential ability as the use of more arbitrary signals would, but rather draws on the apes’ abilities to note perceptual similarities. Yet if we consider the so-called obvious iconicity of the sign for tree, which entails an upright forearm (the tree trunk) with fingers spread wide (the tree tops), the iconic aspects are obvious only after one is made aware of the meaning of that sign (Studdert-Kennedy, 1980). In fact, tests on the meaning of signs given to nonsigning humans reveal that such persons are generally unable to deduce the meaning of a sign from its form (Bellugi and Klima, 1976). Once the naive observers are told the meanings, however, they generally agree on the representational basis. Even if the chimpanzee is more adroit than humans at guessing the iconic referent for some sign (which is unlikely), one must keep certain points in mind: (1) Not all or even most ASL signs are icons; and (2) it is an impressive intellectual feat to comprehend the abstract relationship that, e.g., five open fingers represents the many leaves and limbs of a tree. Rather than being a denigration of their abilities, the apes’ capacity to appreciate the iconicity of signs should be considered instead an exciting cognitive skill to be investigated further. Others criticize the use of idiosyncratic sign dialects. Such dialects arise when trainers accept sign approximations from the ape subjects and, in fact, adopt

LANGUAGE IN THE GREAT APES

161

some of the compromises as part of the standard idiom. This is much as parents do when they accept approximate vocalizations from the developing child as adequate renditions of a word and when they use sounds like blankee for blanket in place of the more accurate counterparts. The disparagement of sign idiosyncracies is based upon the criticism that the time usually required by even a veteran signer to comprehend ape signing is due not to an adaptation to the ape’s dialect, but is due instead to the human observer becoming sensitive to nonverbal cues by which he can either guide the correct response of the chimp or come to understand what the chimp is signing about (Umiker-Sebeok and Sebeok, 1980). This is an invocation of the Clever Hans phenomenon, but to invoke Clever Hans errors is not to prove them. Postulating mysterious nonverbal cues offers little understanding of the ape’s signing or the Clever Hans phenomenon. We shall consider the phenomenon more fully in Section X .

5 . Differences in Training There appear to be at least two major differences in training between the signing projects. One is ?he number of trainers involved; Terrace’s project has had by far the greatest number of trainers, 60. However, even the Gardners (B. T. Gardner and Gardner, 1979) note that the chimpanzees in their recent project were exposed to at least six different humans during the course of a day. Detailed reports of number and turnover of trainerdcaregivers have not generally been published by the projects’ researchers. Yet a large number of trainers could undermine any sort of learning, if apes, like young children, need the emotional security of a stable caregivedtrainer relationship in order to exhibit their best learning. This issue is discussed further in the middle of Section IV,C. A second major training difference between the signing projects appears to be the emphasis placed on formal training sessions or “drills” as they are labeled by some (Fouts et a l . , in press). Terrace seems to have laid greatest emphasis on such vocabulary tests in which the human trainer typically asks the ape in signs, What this? or, occasionally, Who, Where, etc. However, all projects used these methods extensively, whether the animals were in an enclosed room, playing in a gym, or wandering outdoors. The emphasis given drills in training is an important issue, for some experimenters themselves have indicated the ape’s performance became more “perfunctory” (B. T. Gardner and Gardner, 1971) as sessions became more drill-like. It may be reasonable to expect less creativity and less spontaneity during drill sessions than during nonstructured time. Further, exposure to frequent drilling may result in general behavior that models that of drill sessions. In fact, we note from the field of early childhood education that children model rather precisely in their play the kinds of teacher-student interactions that they experience in school. See Section IV,C for further discussion of formality-informality in training.

162

CAROLYN A. RISTAU A N D DONALD ROBBINS

6 . Differences in Methods of Recording Data

Methods of data recording in the projects appear to differ primarily in the quantity of videotaping film. All the signing projects used at least some videotape recording, but Terrace’s reliance on this medium was by far the most extensive and consistent. Observers in other projects also used a daily diary or teacher reports, and tape-recorded observations of signing and the contexts of such signing, but again the Terrace project apparently had the most extensive description and subsequent analysis of the context. For example, Terrace et al. (1979a,b) analyzed a total of 3.5 hr of such videotape recordings, representing 9 different sessions, over an 18-month period (it is not clear whether this represented a sample of the videotaped sessions or all of them). Patterson (1978a, p. 76) has made 30-60 midmonth videotaped samples since the sixteenth month of her project, although we know of no published detailed analysis. Early in Project Washoe the Gardners (B. T. Gardner and Gardner, 1971, p. 142) indicated that neither films nor videotape was suitable because of the difficulty of keeping a camera in a position that permitted the signing to be intelligible. Later (B. T. Gardner and Gardner, 1980, p. 24), they point out that they have made film and videotapes of a “generous sample, although, again, in the same publication they indicate the difficulty in using film material (Footnote 3, p. 48). Thus, the Gardners’ use of film has been largely for demonstration purposes. In summary, even the motion picture data subjected to most detailed analysis (3.5 hr worth) are very limited for a project spanning several years. Yet the effort involved in such an undertaking is enormous. It is certainly not feasible to make and analyze continual filmed records, but more analysis of carefully selected samples should be done. Unbiased, unedited film records should also be available, at least to the scientific community. ”

B.

THEGARDNERS’ PROJECTS

Intrinsic to any discussion of the results of the signing project must be an understanding of the specific methods used to gather data and the criteria for acquisition of a vocabulary item. The Gardners developed methods and criteria which were adopted and modified by researchers in subsequent projects, so we shall focus upon the Gardners’ procedures. The Gardners (B. T. Gardner and Gardner, 197I , p. 140)reported spontaneous Occurrences of Washoe’s signs in the daily record. At first, productions were considered spontaneous if they occurred either without prompting or in response to questioning such as What is ir? or Whar do you want? A sign was considered to be introduced into the vocabulary after three different observers had reported its spontaneous and appropriate occurrence. Thereafter, at least one such occurrence each day for 15 consecutive days was required for the item to be considered part of Washoe’s vocabulary. Initially, it was feasible to keep an exhaustive account of the occurrence of all of Washoe’s signs; as her vocabulary expanded that method was changed. In-

LANGUAGE IN THE GREAT APES

163

stead, two different observers used a check list at least twice a day and recorded if any of the signs on the check list had occurred spontaneously and appropriately. The observers also noted the context, the quality of form, and the kind of prompting (if any). As the list grew in length, the amount of detail recorded on the check list was, of necessity, reduced (B. T. Gardner and Gardner, 1971, p. 140). As the vocabulary list lengthened at an accelerating rate, the contexts in which signs were produced broadened while the conditions for eliciting some of the signs, particularly the new ones, occurred quite rarely. The sessions therefore became more structured and observers deliberately introduced contexts appropriate to the signs that had not been observed (the method of “obligatory contexts”). With continuing increases in Washoe’s vocabulary, the sessions became longer and more like drill sessions, and Washoe’s behavior required more discipline and became more perfunctory. The use of more structured and inclusive procedures had at least two important impacts on the appropriate interpretation of Washoe’s signing behavior. One concerned the interpretation of the statistics for occurrence of signs; the drill procedure made certain that most signs on the list would occur at least once a day. Second, because of the drilling, the sense of the term “spontaneous” changed (B. T. Gardner and Gardner, 1971, p. 141) and was used in its broader sense for the duration (and major part) of the project. Colloquial use of the term “spontaneous” would probably refer only to occasions when the ape produced a sign with no explicit or implicit questioning by the human. For example, we might consider “spontaneous” only those times when Washoe signed cat for a cat, without her attention being drawn specifically to the cat, or when she signed cat to a picture of one in a book through which she was browsing, without being asked to name the objects in the book. However, Washoe’s replies to questions such as What is it? and to repeated pointing at an object were considered “spontaneous” by the Gardners, as was a cat sign made by Washoe to a picture of a cat shown her during a drill session. Signing cat to a cat seen on an excursion was also reported as ‘‘spontaneous” whether or not it had been necessary to point the cat out to her or to ask her to name it. Sometimes more informative prompting was required, such as modeling the correct sign for Washoe to imitate, indicating one of Washoe’s hands or the parts of her body that she should touch to make the sign, or actually guiding her hands through the correct motions. When prompting was informative and could have guided Washoe into making the correct sign, then Washoe’s sign was reported to have been prompted. In short, the distinction between “spontaneous” and “prompted” is not the usual one. In order to describe Washoe’s use of her repertoire of signs in her daily activities, the Gardners used a sampling technique. The procedure was similar to that used in obtaining samples of language for young children. The Gardners (B. T. Gardner and Gardner, 1971, p. 142) state that such data should indicate every-

164

CAROLYN A . RISTAU A N D DONALD ROBBINS

thing Washoe signed, everything signed to her, and the events surrounding each occurrence. Instead of using film or videotape the Gardners relied upon miniature recorder into which an observer whispered observations. Later in the project (month 29), they taped the supper meal for periods of 15-20 min as a set daily activity approximately once each month. The repeated sessions were intended to study the development of Washoe’s use of signs: the variety used, the length of utterances, and the amount of signing during those sessions. In other sessions, a variety of activities were explored in order to reveal Washoe’s choice of vocabulary in different contexts. From these records, the Gardners also accumulated examples of interchanges and of Washoe’s use of signs in combination in order to compare them with similar data from human children (B. T. Gardner and Gardner, 1971, p. 142). In the Gardners’ sequel to Project Washoe, they improved the procedure for their four subjects by exposing the apes to sign language beginning at birth, by using fluent ASL signers as adult models, and by strictly enforcing a rule of “Ameslan only” in the presence of the chimpanzees (R. A. Gardner and Gardner, 1978, p. 5 5 ) . This latter rule was imposed because the structure of Ameslan differs so markedly from the structure of English. Otherwise, the homelike environment and the continued exposure to sign language were similar to those of Washoe (B. T. Gardner and Gardner, 1979). Besides recording ongoing and prompted productions of the chimpanzees, the Gardners also used more formal vocabulary testing situations in both the Washoe and later projects. They also use a double-blind technique which was purported to demonstrate the high interobserver reliability of the data (B. T. Gardner and Gardner, 1974). Typically, the observers could not see one another, although they did have an unobstructed view of Washoe. Randomly ordered slides were presented to Washoe by one individual, while a second person asked her (in signs) what she saw. A third person (whom Washoe could not see) also recorded her signing behavior. Written reports of the latter two people are compared to determine interobserver reliability. In most tests, the observers were members of the staff. In two control tests, however, the person who was not interacting with Washoe was a deaf person fluent in ASL, but with little previous contact with Washoe. This condition, though not explicitly stated by the Gardners, was presumably to control for inadvertent social cuing (as in the Clever Hans phenomenon) and to determine if Washoe’s signs were reasonable approximations to ASL usage. On the average 69% agreement was found during the fkst test session between a deaf observer and the trained observer, and 89% during the second session. As the Gardners note, the results of the second test session are similar to the agreement found between trained observers. The Gardners conclude that Washoe’s signing was reasonably standard and could be learned readily by persons know-

LANGUAGE IN THE GREAT APES

165

ing Ameslan. In support of the Gardners, it can be said that even 69% interobserver agreement is a rather remarkable achievement, particularly when one recalls that 50% agreement is not at chance level. In fact, given that the number of items on these tests was typically about 20, chance, by any calculation, was less than 10%.However, one could also note that these are optimal conditions for verifying Washoe’s productions. Only one sign is studied at a time, the sign occurs during a trial period whose initiation is known to the observers, and by virtue of the slide technique used by the Gardners, the observers know the sign will be a noun. Furthermore, the observers have access to a list of Washoe’s signs from which the subset used in the test session is drawn. With such information, it is at least conceivable that even the “stranger” observers might be able to guess some of Washoe’s signs on the basis of her nonsigning reactions to the slide presented on a given trial, i.e., she might perhaps make excited food gestures or calls to a photo of a banana, and so forth. Assessing Washoe’s performance of spontaneous strings of utterances, an essential component in determining linguistic competence, is, of necessity, more difficult. It may, in fact, be almost impossible without filming or videotaping synchronously, from several vantage points. This would be a difficult procedure: the chimpanzee should be free to move about, be as relaxed as possible, and be habituated to the presence of the camera. Data analysis would be extremely tedious and the expense would be very great. Yet there is need for such a record in all the ape language projects. Miles (1978) used an intermethod reliability test in which she compared her written and oral transcriptions to a videotape of a chimpanzee signing, and found serious inaccuracies in the written and oral transcriptions. In particular, from two 10-min transcriptions, she found several signs and sign combinations missing and sign repetitions unrecorded; the sign order was incorrect in a few cases. To our knowledge, there is no other published intermethod reliability data. Miles decided to use a videotape record. The Gardners’ methods also included description of the developmental process of vocabulary acquisition and Comparison to human children. By 3 months of age, the chimpanzees were interested in pictures which facilitated the number of exemplars available to use in the sign teaching. Also by 3 months the first signs began to appear (R. A. Gardner and Gardner, 1975). The Gardners point out that some children make their first manual sign when they are but 5-6 months old. They pose the question of whether the human infant finds it easier to form signs than to form words (B. T. Gardner and Gardner, 1979). In their latest project using four very young chimpanzees, the Gardners report a striking similarity between the first 50 words used by a group of children (Nelson, 1973) and the first 50 signs of three of the four chimpanzees. Both the children and the chimpanzees were just under 2 years old when they reached the 50-item lower level vocabulary. Furthermore, the Gardners found they could classify the chimpanzees’ words into the same six categories devised by Nelson

166

CAROLYN A. RISTAU A N D DONALD ROBBINS

for the children she studied. These were nominals, both specific (proper names for children and adults) and general (e.g., dog, milk, and shoe), actions (go, see, u p ) , modifiers (states such as finished and hot, and the possessive, mine), personal-social (the social expressive terms no, please, refusal), and function terms, including the question word what and the questioning form of rhar. Of these, the nominals were the most numerous, accounting for half the vocabulary. Terms such as up or out, when used for actions such as pick me up or I want to go out, are considered an action term in this classification. Both Nelson and the Gardners noted that frequent exposure to references does not guarantee that they will be named; the references commonly named by children were “changeable, moveable, and manipulable” (B. T. Gardner and Gardner, 1979, p. 343). For example, household items such as sofas and tables were not named, but lights which children switch on and off were. The Gardners state that before the chimpanzees were 1 year old, they used sign combinations and, in particular, exhibited the basic sentence relations that children express in their early two-word occurrences: agent and action (Susan brush), action and object (Chase me, Gimme drink), demonstrative and entity (There drink, There diaper), and action and location (Tickle there, U p go). The Gardners (B. T. Gardner and Gardner, 1979, p. 346) reported that 74-90% of the chimpanzee’s two-sign utterances showed sentence relations characteristic of Stage 1 child speech (Brown, 1973) and child signing (Fischer, 1974; Klima and Bellugi, 1972), and also that Brown found this to be the case for 75% of the twoword utterances that he classified. The chimpanzees also occasionally used combinations with negative terms, which Brown (1973) regards as the most advanced form of simple combinations. The negative combinations included nonexistence such as Gum no, which the Gardners took to indicate “the nonexistence of gum at a customary location,” and refusals such as Hug no (B. T. Gardner and Gardner, 1979, p. 346). As with children, the negative term was sometimes combined with the item being negated, e.g., Hug no was used when the chimpanzee refused to hug. At other times the negative term was combined with a positive alternative, e.g., No, drink was used when refusing solid food (B. T. Gardner and Gardner, 1979, p. 346). Interrogatives are particularly interesting because different kinds of questions call for different types of reply. Children readily answer questions at age 21 months, and there is an order in their appropriate use of such questions. The question types that children answer early contain What, Where, and Who; replies to Whose and What do occur later, and replies to Why, How, and When occur even later (B. T. Gardner and Gardner, 1979, p. 348). Two of the young chimpanzees answered at least some types of the earliest interrogatives appropriately; Moja answered all the early ones correctly, as well as Whose and the causal question, Whatfor that? which interpreted as a varient of What do? For example, a whistle is for blow and shoes and hat are for G o out.

LANGUAGE IN THE GREAT APES

C.

167

THENIMPROJECT OF TERRACE A N D ASSOCIATES

The signing project by Terrace and his colleagues is outstanding primarily because of the extensive analyses done of Nim’s multisign utterances; more than 19,000 of Nim’s combinations were studied in some analyses and videotapes were investigated in others. To date, no comparable analysis exists for any other ape language project. A major purpose of the analyses was to determine whether Nim ’s utterances exhibited any grammatical rules. The subject in Terrace’s project was a male chimpanzee, Nim Chimpsky. Nim was formally exposed to manual signing in a home environment within 2 weeks after birth and began formal classroom training sessions in sign learning when he was approximately 10 months old. The project lasted until Nim was 44 months old, when he was returned to the University of Oklahoma, which houses several other chimpanzees who had been taught to sign. Some aspects of training were very similar to training in the other signing projects, such as the use of molding and guiding to form the chimpanzee’s hands into signs. Note that in Terrace’s project, as in the other signing projects, there was no explicit training for utterances longer than one sign. We can well expect, however, that trainers expressed delight when longer utterances were produced. Data collection techinques in the Terrace project were very different from those in the other signing projects. Once formal training of Nim began, data were gathered primarily during the 3- to 5-hr training sessions in the classroom at Columbia University, as compared to samples taken in other projects during meals and other parts of the chimpanzee’s day (R. A. Gardner and Gardner, 1978; B. T. Gardner and Gardner, 1979; Patterson, 1978a). We will next consider some of the results. The data were analyzed initially to determine whether Nim’s productions exhibited any lexical regularities. To answer this question, a large sample of the chimpanzee’s utterances from the eighteenth to the thirty-fifth month of training (roughly from the first third to the second third of the project)-over 19,000 utterances-was analyzed for lexical regularity. This is essentially a very simple question. Can one account for the word order in utterances of two or more signs simply by assuming independent position habits? The answer is no; the order in two-sign utterances is not accounted for by assuming independent position habits; there are lexical regularities. Yet there were not lexical regularities for three- and four-sign utterances, and there were insufficient data to analyze for such regularities in utterances of five or more signs. However, unlike longer utterances spoken by human children, Nim’s three- and four-sign utterances did not add new information; they were redundant and repetitious. For example, the most frequent two-sign utterances were Play me, M e Nim, Tickle me, and Eat Nirn. The most frequent three-sign utterances were Play me Nirn, Eat me Nim, Eat Nim eat, and Tickle me Nirn. The most frequent four-sign utterances were Eat drink eat drink and Eat

I68

CAROLYN A. RISTAU A N D DONALD ROBBINS

Nirn eat Nim (Terrace et a l . , 1979a,b). (See Section IV,F for examples of utterances that are far longer but mainly repetitious.) A lexical analysis, however, says nothing about whether regularities, such as those found in the two-sign utterances, are used in a meaningful way. It would be quite possible for an organism to use two-sign utterances whose order was not due to independent position habits, and yet the sequencing in the utterances might bear no relationship to the context i n which they were used. A semantic analysis is crucial if one wishes to suggest that even two-sign utterances follow grammatical rules. It is also not clear that an analysis for lexical regularity will reveal regularities that do exist. They may become apparent only if one does a semantic analysis as well. For example, the chimpanzee might possibly say Play me and Me play in the same or different proportions. He might say Play me 30, 50, or 70% of the times he uses those two signs. Yet whether he is using those signs grammatically depends upon the meaning he ascribes to the sign orders. It is possible, e.g., that Play me means “Play with me,” while Me play means something to the effect that “I am now playing.” The meaning and the existence or nonexistence of grammatical structure critically depends on understanding the context of the utterances, such as the behavior of the communicating chimpanzee, the recipient, and others about him before, during, and after the utterance, the general sequence of events occurring about the chimpanzee, the communicatory acts surrounding the signing, and so on. Terrace did do a semantic analysis, on data from videotape which we will discuss later. Terrace and his associates also performed a “discourse analysis” of videotape transcripts of Nim and his teachers’ signing in order to understand how Nim’s sign sequences were related to those of the teachers. These videotape transcripts totalled 3.5 hr, obtained from 9 sessions ranging between 20 and 30 min each, which were collected when Nim was between 26 and 44 months old (the last 1.5 years of the project). The video transcripts were apparently drawn largely from the formal traininddrill sessions, but also included some home sessions. The discourse analysis studied Nim’s adjacent utterances, defined as those that follow the trainer’s utterance without a definitive pause (Terrace ef a l . , 1979a, as based upon Bloom et a l . , 1976). Terrace et a l . note that approximately 87% of Nim’s utterances were categorized as adjacent (range: 58.7-90.9%). In contrast, a child about 21 months old in the first stage of linguistic development [mean length of utterance (MLU) = I .36] produces a lower proportion of adjacent utterances in conversation with an adult, namely 69.2% (range: 53-78%). Of Nim’s adjacent utterances, approximately 44% were imitations or slight reductions of the utterance just produced by the human trainer; in contrast, in children, imitations and reductions accounted for 18% of the adjacent utterances at Stage 1, decreasing to 2% of the utterances at Stage 5 (MLU = 3.91). Only 7% of Nim’s utterances during the videotape sessions were expansions of the trainers’ prior utterance, ii contrast to 21.2% of a child’s Stage I utterances. Expansions are those pro-

LANGUAGE IN THE GREAT APES

169

ductions that contain some of the lexical items of the adult’s utterance along with some new lexical items, and assume a larger proportion of the child’s utterances as the child gets older. Many of the child’s expansions are said to be systematic expansions of verb relations expressed in the adult’s prior utterance (Bloom et a l . , 1976), but Terrace et al. noted no such pattern in Nim’s expansions. In fact, preliminary analysis indicated that Nim’s expansions contain only a small number of additional signs such as me, Nim,you, hug, and eat. And these, as we previously noted, do not add new information to the teacher’s utterance. In addition, Nim signed simultaneously with the trainer 71% of the time and of these simultaneous signs (425 out of 585 signs made by Nirn), 70% began while the trainer was signing (or, in the interpretation of Terrace et al., were “intenuptions” of the teacher’s signing). Furthermore, Terrace et al. claim that analyses of films of other signing apes lead to similar conclusions. A number of criticisms have already been made of Terrace’s work. In general the critics disagree with Terrace’s conclusions about the lack of evidence for syntactic structure in the apes’ combined symbols to create new meaning. The critics attempt to discredit the data base and the data collection methods, arguing that if the data base is faulty, conclusions drawn from it must be as well. We will examine some of the major criticisms. A major concern is that Nirn had a very large number of trainers; he was exposed to 60 trainers in slightly less than 4 years. If, for chimpanzees, as apparently is true for children, language learning is optimal under the tutelage of a caregiver with whom a child has strong social bonds, then Nim is strongly handicapped because of his many trainers. At least superficially, it would seem that 60 trainers in 4 years, or 15 trainers per year yields an average of a new trainer every 3-3.5 weeks. It seems reasonable to suspect that a child who experienced a different caregiver every 3 or 4 weeks during this part of its life might suffer rather severe impacts on emotional development. We would surmise that an emotionally distraught child is unlikely to learn language o r anything else optimally. Data on the social development of wild chimpanzees (van LawickGoodall, 1968; Plooij, 1980) suggest that similar problems would arise for chimpanzees. There are certainly strong indications in the ape language literature (Terrace, 1979b; B. T. Gardner and Gardner, 1979) that the presence of a stranger as a visitor or trainer results in emotional disruption on the part of the apes, making them anxious and/or unwilling subjects. However, as indicated in the appendix to the book, “Nirn,” the durations trainerdcaregivers were associated with the Nim project are not so extreme as 3 weeks. In each year of the project, at least five or six trainers were involved and were present in both the home environment and the university. Nirn also had a core group of eight teachers, each of whom had at least 1 year’s training in ASL, while most of the other trainers, who were occasional playmates, typically were not as fluent in ASL as the core group. Anecdotes in the book suggest that Nim had developed

170

CAROLYN A. RISTAU A N D DONALD ROBBINS

some strong stable social relationships. Furthermore, as reported in an appendix to a recent publication by the Gardners (B. T. Gardner and Gardner, 1979), the chimpanzees in that project were regularly exposed to at least six different humans on any one day (three different trainers, a “chimpsitter,” a I-hr visitor, and an overnight visitor). In point of fact, most projects have not published detailed accounts of the number and turnover of trainershregivers. In summary, although the criticism of an overwhelming number of trainers for Nim remains a potential limit to Nim’s achievements, the criticism must be tempered by the realization that Nim did have some stable sociaYemotional relationships, and, furthermore, that chimpanzees in the other projects suffered some of these same difficulties as Nim, although apparently to a lesser degree. Terrace et al. were aware of this problem; they indicate that to attempt to go beyond Nim’s abilities, it is important that the ape be raised and taught by a small and stable group of trainers (Terrace et a l . , 1979a, p. 202, Footnote 67). The Nirn project is also criticized because there were very many formal training sessions or “drills” as some researchers have described them (Fouts et a l . . in press), although it is very difficult to evaluate this matter adequately for the nature of training is typically not clearly specified for other projects. For much of the project, Nim underwent formal training 3-5 hr a day, at least 5 days a week, as compared to chimpanzees in other signing projects who, according to Fouts, had 1 hr or less a day of similar formal training (Fouts et al., in press). Nim’s formal training was conducted for the most part in what has been described as the “structured environment” (Fouts et al., in press) of a Columbia University classroom, a room generally devoid of many objects and toys so as to avoid distraction. This was done, as Terrace notes, to permit greater concentration on training and ease in videotaping. However, as Terrace emphzsizes (personal communication), classroom sessions were fairly brief (30-45 min each) and other activities at Columbia included trips to a “gym,” walks around campus, and longer excursions to Riverside Park. Much of the classroom time was spent i n training Nim with new signs and in eliciting his knowledge of signing by asking What this? repeatedly. However, Nim’s teachers and caretakers attempted to teach him to sign during all of his waking hours, including the time he spent in his home environment. Chimpanzees in the other projects seem to have been trained for a greater proportion of time in apparently less structured environments such as the outdoors or a home, yet it must be realized that training sessions even in such environments also consisted of innumerable repetitions of the human trainer asking, What this? In fact, Miles (1978) notes of her work with Ally and Booee, two chimpanzees at the University of Oklahoma Institute for Primate Studies, that the most interesting and unusual sign language exchanges between a chimpanzee and a human occurred in the casual interaction between the apes and human trainers, rather than during training or experimental sessions. Thus, the time before and after sessions, during walks in the woods, car rides,

LANGUAGE IN THE GREAT APES

171

feeding, and general play tended to result in more spontaneous communication by the chimpanzees than did the formal sessions. A third criticism is concerned with the data on turn-taking: Terrace et al. report that Nim very frequently did not await his turn in the conversation but interrupted the ongoing signer. Wild chimpanzees, and various other primates, when aroused, sometimes join together in communal vocalizing; chimpanzee pant-hooting is a case in point. Similarly, young children must learn to develop the ability to take their turn in a spoken conversation rather than vocalizing whenever the mood strikes. Turn-taking is considered essential to conducting a spoken conversation (see Fouts et al., in press). As discussed by Fouts et al. and by others, there are three aspects of conversation that are important to us for our present purposes: to initiate, to continue, and to shift a conversation. As in the spoken language, signs are often emitted by the addressee while the speaker is still conversing and are not considered interruptions. These utterances are probably equivalent in the spoken language to the addressee nodding the head up and down and saying, Yes, Uh huh! Some recent work (Baker, 1975) indicates that roughly 30% of discourse can be characterized as continuation markers. Furthermore, in terms of sign language, the last sign is often held and that sign is not considered an interruption. Howeyer, Terrace has indicated that 70% of Nim’s interruptions were made while the teacher was still signing. Other data (about which we have great reservation) suggest that in human-ape signing interactions, humans were more interruptive than apes (oral report of M. E. Hannum, R. S. Fouts, and R. Ingersoll, 1980, as cited in Fouts et al., in press). The data were drawn from films of two chimpanzees, Nim and Mac, interacting with humans who were either signers or nonsigners. Nim interrupted the humans twice while the humans interrupted Nim seven times (56 total turn signals). Unfortunately, at least in the brief description in Fouts et al., the number of humans is not specified; it is not clear whether these data derive only from signing interactions; the human signers are not native signers nor is their level of expertise in ASL indicated; and the data base is small. We have more general concerns about analyzing turn-taking by signing apes. The apes’ length of utterance is simply not large enough to permit any detailed, sophisticated analysis of the apes’ turn-taking behavior. We need to ask to what were the apes exposed? Did they observe in the signing communications between humans and between humans and apes the turn-taking behavior to which we have alluded?

D. THEGORILLAKOKO, STUDIED BY

PATTERSON

The Patterson work (Patterson, 1978a,b, 1980; Patterson and Linden, 1981) is distinctive in that it is the only one in which the subject is a gorilla. Training

172

CAROLYN A. RISTAU A N D DONALD ROBBINS

began when Koko was about I year old and was conducted at a zoo nursery. After about 1 year, the project was moved to a house trailer where Koko has since lived and been trained. Koko had one major trainer and some 14 assistants, 9 of whom were deaf or native signers. As previously noted, Patterson used the method of “simultaneous communication,” so that Koko observed signing as well as being exposed to English speech. In this regard the Gardners (B. T. Gardner and Gardner, 1979) remind us that the structure of ASL and that of English is very different and thus speaking “good” English and simultaneously signing “good” Ameslan, would be extremely difficult. For example, Miles (1978) points out the sentence Let’s go tu school tomorrow. in ASL would be tomorrow you me go school, if we were emphasizing when we would go; however, to emphasize where we would go, it would be signed as school you me go tomorrow. As can be seen from Table I, Koko appears to have achieved a larger expressive vocabulary, and is credited with greater linguistic achievements, than most of the other apes (Patterson, 1978a,b, 1981 ; Patterson and Linden, I98 I ) . But Patterson indicates that some differences may not be significant, e.g., Washoe’s and Koko’s different rates of sign acquisition. Furthermore, she notes that gorillas should not be construed as being better at language than chimpanzees; more data from several gorillas and chimpanzees are needed (Patterson and Linden, 1981, p. 89). Patterson is currently training another gorilla, a young 3.5-year-old male, Michael, a prospective mate for Koko. As yet there is little published information about Michael’s abilities, although he is described as inquisitive, with good powers of concentration, and more tendency to spontaneous sign use than Koko had in the early days of the project. However, he is still far behind Koko in vocabulary size and “grasp” of sign use (Patterson and Linden, 1981, pp. 169-171). We will offer some commentary on Patterson’s projects when we reconsider all of the signing projects in Section IV,F. We would like to note that a very readable popular book has appeared, “The Education of Koko” (Patterson and Linden, 1981) with considerable important detail not previously available, but often lacking the information a scientific audience demands. We still await more data analysis from the Koko project.

E.

OTHERAPE PROJECTS

Fouts and his colleagues (1972, 1973; Fouts et al., 1976, in press) have been engaged in a number of studies, including evaluating novel combinations uttered by various chimpanzees, investigating communication between signing chimpanzees, and comparing various training methods. Miles’ (1978) thesis study illustrates the difficulties in data collection (pp. 28-39), and sign transcription (pp. 5 1-52), analysis, and interpretation often encountered in these studies.

LANGUAGE I N THE GREAT APES

F.

CONCLUDING COMMENTS ABOUT

THE

173

SIGNING PROJECTS

The signing projects share a number of methodological and interpretative problems. Some of the difficulties we shall consider relate to the process of acquiring data, such as the use of daily diaries, the procedures for conducting reliability tests, and anecdotal reporting of novel word use and combinations. Other difficulties refer to basic interpretive problems as well as definitional differences between the projects, e.g., the mean length of utterance (MLU), interpretive difficulties inherent in semantic analysis, and spontaneity of utterances which are thought to reflect the developmental process of sign acquisition.

I.

The Process of Acquiring Data

Considering first the daily diaries, we note that all the projects rely upon such diaries, though to varying extents. The diaries apparently are often prepared after the trainer completes a session; data collected in this manner are subject to the vagaries of the human memory system. Trainers in the recent Gardner project, e.g., are instructed to record chimpanzee signing behavior no more than 1 hr dfter it has occurred (B. T. Gardner and Gardner, 1979). Another problem common to many researchers of social interactions is faced by the ape language investigators as well-that of conducting reliability tests. We have already discussed the Gardners’ use of double-blind tests (Section 1V.B) and shall now consider those used by Terrace and Patterson, so as to reveal other classes of problems that beset researchers. The reliability tests of Terrace el a l . (1979a,b) consisted of the comparison of teachers’ reports. These reports were derived from transcripts of audio recordings of teachers’ ongoing verbal descriptions or from videotapes of the data collecting sessions. Seventy-seven percent agreement was found before teachers consulted with one another, and 94% agreement afterward. Social interactions between the teachers could have affected the interpretation of Nim’s utterances, i.e., on determining which signs were presumed to have been observed (a criticism often made of ape language projects by Sebeok). Yet even the initial 77% agreement is not poor performance in this difficult task. Patterson (1978b) utilized a form of double-blind testing. In her tests, one observer baited a box while a second (who did not know the contents) asked Koko in signs what she saw. Patterson reports that Koko often “avoided” the task by giving the same sign to all objects, not signing, or signing to have the box opened. Overall, her performance was approximately 60% correct. It is difficult to interpret these data; one might speculate that Koko did not like to take these tests, or that she had difficulty labeling objects apparently in her vocabulary. Clearly, there are numerous difficulties in so simple a task as determining the signs in an ape’s vocabulary. Assessing the ape’s performance of spontaneous strings of utterances, an essential component in determining linguistic competence, is, of necessity, more difficult.

174

CAROLYN A . RISTAU A N D DONALD ROBBINS

2. Signiftcance of the Mean Length of Utterance Data on the mean length of utterance (MLU) are considered to be important, because they presumably can help to establish parallels between the chimpanzee’s abilities and a comparable stage in the development of a child’s linguistic proficiency. Proper interpretation of the MLU data is an area of some controversy between various projects. Terrace et a l . (1979a) indicate that Nim’s MLU failed to increase during the last 19 months of the project, when Nirn was from 26 to 45 months old (MLU range: 1.1- 1.6). Patterson (1978a, p. 94, 1981, p. 86), however, notes a 33% increase in Koko’s MLU during a 12-month period when Koko was at an age similar to Nim’s (29-40 months old) and had experienced about the same duration of training as had Nim. Terrace et a l . (1981) respond by pointing out that Koko’s MLU did reach an asymptote of about 2 later in the project. Terrace and his colleagues then question whether MLU is positively related to an increased complexity in language use and cite the following as examples: Please milkplease me like drink apple bottle, which was signed by Koko when her MLU was 1.75 (Patterson, 1979, p. 3 4 3 , and Give orange me give eat orange me eat orange give me eat orange give me you, which is a 16-sign utterance made by Nim when his MLU was 1.6 (Terrace et a l . , 1979a). (Note that repetition cannot occur within an utterance if apes are using plastic chips or lexigrams. The ape is given only one chip of each sort. A lexigram key once depressed, remains depressed and cannot register repeated pressings, so data cannot be gathered on repetition.) Terrace states that productions such as these examples are uncharacteristic of the speech of hearing children or the signing of deaf children. Instead, he suggests, the apes have learned to produce contextually appropriate signs “until they got what they wanted” (Terrace et al., 1981, p. 87). His criticism seems reasonable, although not acceptable to all the ape language investigators. 3 . Interpretative DifJiculties Inherent in Semantic Analysis Lexical regularities, as observed in Nim’s two-sign utterances, do not in themselves yield information about the meaning of utterances. A semantic contextual analysis is needed to begin to understand the meanings of apes’ signs and to determine if apes use grammatical structure. Contextual analysis, as the term is used by linguists and by scientists in the field of animal communication (Smith, 1965, 1968, 1977; Marler, 1977a,b), requires investigation into the meaning of an utterance, as it depends upon the other utterances made, the nature of the interaction of which the utterance is a part, and, particularly as the ethologists and animal behaviorists have used the term, upon the context of that interaction such as the identity of the communicator and recipient, their sex and physiological state, the season of the year or time of day, and so forth. To cite one example, a contextual analysis would need to deal with the following: Suppose Nirn were inside a room, perhaps looking at the door; if Nim then

LANGUAGE IN THE GREAT APES

175

signed Me our to his trainer, one might attribute a certain meaning to it. In contrast, if Nim were walking in the woods with a trainer and signed Me out, we might consider that his signing referred to his thought “I am outside” or might be a repetition of signing that he or the trainer had repeatedly made before in that context. Without knowing this sort of detail about the context in which signing occurred, we cannot even begin to study the syntactic or grammatical structure of this signing. Note that a semantic analysis, which involves contextual analysis, requires observers to make judgments as to what each sign or word combination may mean. Terrace et al. (1979a) indicate that the method developed to use with human children requires an observer to relate certain aspects of the immediate context of an utterance to its contents; such a technique is called the method of “rich interpretation” (Bloom, 1973). A major problem with this method is to demonstrate the validity of the judgments of meaning rather than their reliability. When working with children, Bloom (1973) emphasized the need to search for independent evidence from the corpus of utterances. The following observations have been used as supporting evidence for the validity of meanings attributed to utterances of children: The child’s word order is usually the same as it would be if the meaning were being expressed in the standard adult form. Sometimes intonational differences are associated with different meanings (Bloom, 1973). As the child’s MLUs increase, one notes that relationships expressed in twoword utterances are the first to appear in three- and four-word utterances. Longer utterances seem to be composites of semantic relations found in shorter ones. New semantic relations are initially expressed in short utterances, and these are frequently imitations and/or reductions of previously made adult utterances. Many of these techniques are not available to validate interpretations of ape utterances. For example, ape sign combinations are simply not long enough to determine if the meanings of short utterances are the first to occur in longer utterances. (Recall that Nim’s MLU during the last 18 months of the project was 1.6 and Koko’s was about 2 . ) Due to the paucity of longer utterances, semantic analyses in the various ape signing projects are of two-sign combinations [B. T. Gardner and Gardner, 1971; Miles, 1978, Appendix G , pp. 198-200 (about chimpanzees); Patterson, 1978a, 1979, p. 159; Terrace et al., 1979al. We will consider Terrace et a1.k (1979a) analysis in some detail and relate it to some of the other research. Three of Nim’s teachers analyzed 1262 instances of Nim’s two-sign utterances by examining their own and the other teachers’ reports of a sample of sessions and then used the method of “rich interpretation.” After consultation they agreed upon the interpretation of 76% (967) of the combinations. All analyses were performed on these 967 two-sign utterances. Terrace et al. found 20 categories of semantic relations which accounted for 89% (852) of the usable two-sign combinations. In contrast, Brown (1970) found

176

CAROLYN A . RISTAU A N D DONALD ROBBINS

1 I categories to account for 75% of children’s combinations in his sample; 9 categories were found for 78% of a sample of Washoe’s two-sign utterances (B. T. Gardner and Gardner, 1971); and 11 categories for 75% of a sample of Koko’s two-sign combinations (Patterson, 1978a). The categories used by the various investigators are very similar; in fact, Terrace et al. ’s 20 categories could be collapsed into most of those used by the others with only a few remaining. The essential problem, as Terrace et a l . (1979b) indicate, is that the classifications are equivocal. For example, Nim has signed Nim banana in the context of a trainer about to give him a banana. There is no obvious way to distinguish between agent-object, beneficiary-object, and possessor-possessed object. In many other combinations of Nim or me with eat or drink, it was also difficult to determine if eat and drink referred to objects or to actions (or yet something else). Furthermore, as Terrace et al. note, the semantic analyses conducted by themselves and others do not take into account the highly imitative nature of chimpanzee signing; sign combinations that imitate those just made by the teacher may well differ in semantic relationships from more spontaneous utterances. In all, the difficulties inherent in validating interpretations for apes, and the paucity of utterances longer than two signs make the semantic analyses performed to date equivocal. It seems necessary, nevertheless, to do semantic analyses if we mean to talk in terms of “grammatical” relationships of signs and the meanings of ape multisign utterances. It would seem we must begin with more basic studies of meaning. (See discussion in Section VI on meaning, especially VI,A.) 4.

Spontaneity and Imitation of Utterances

One of the central points at issue is the spontaneous nature of the apes’ utterances. Spontaneity is considered a characteristic of human language, while the apes’ performance is sometimes criticized as being almost entirely prompted and/or imitative (Terrace et a l . , 1979a,b). The present confusion appears to concern use of the term “prompt.” R. A. Gardner and Gardner (1980) indicate that Terrace and colleagues (1979a) claim (1) any utterance of the teacher that preceded one of Washoe’s was a prompt; and (2) all of Washoe’s replies were considered to be prompted regardless of whether her signs were the same or different from those of the prior signs of the teacher. These statements seem to be at variance with the definitions actually used by Terrace et af. (1979a). An imitative sign is one which repeats the teacher’s immediately prior utterance. A spontaneous sign is one that has not occurred in the teacher’s immediately prior utterance. A prompted [italics in originall sign follows a teacher’s prompt. A prompt uses only part of the proper sign’s [italics added] configuration, movement, or location (Terrace et a l . . 1979a. p. 901, Footnote 25).

LANGUAGE IN THE GREAT APES

177

Since the prompt refers to some aspect of the sign, it would seem that a response by a chimpanzee to the teacher’s query, What this? as the teacher points to the object or to a picture of an object would not be considered prompted by Terrace et ul. (1979a,b). In this sense, Terrace et al. concur with the definition of “spontaneous” used by the Gardners (Section IV,B). In addition to their study of Nim’s signing, Terrace et al. (1979a,b) also investigated Washoe’s signing performance in the film, “Teaching Sign Language to the Chimpanzee: Washoe. ” Based on their analysis of Washoe’s multisign utterances (24 two-sign, 6 three-sign, and 5 four-sign sequences), Terrace et al. claimed that each of Washoe’s utterances in the film was preceded by a similar utterance or a prompt from her teacher. They concluded that Washoe’s utterances were adjacent (followed the teacher’s without a definite pause) and imitative of her teacher’s. The Gardners have noted that the film was intended to be a demonstration of some of Washoe’s signing vocabulary, that Washoe was not at ease in the presence of cameras, and that the film was not intended to be a representative sample of her normal signing activities. Patterson (1981) also disputes Terrace’s claims about the spontaneity of the utterances made by Koko, the gorilla. Patterson does not explicitly define spontaneity nor does she seem to disagree with the definition by Terrace et al. The controversy here appears to be quantitative. Patterson claims that 41% of Koko’s utterances are spontaneous, while Terrace et al. (1981) state that the figure is 28%, presumably using the same data. In any event, both figures are well above the spontaneous proportion of utterances (7%) attributed to Nim (Terrace et a l . , 1979a). Apparently, attempts to have a more formal, controlled training regimen may inhibit the apes’ lingusitic ability, while a more relaxed “homelike” environment may yield stiuations in which data-recording techniques are less than optimal. Terrace et a l . (1979a) concluded that their present evidence suggests that ape language learning is severely restricted. He and his colleagues consider that apes can learn many isolated symbols, but state that there is no unequivocal evidence that apes can master conversational, semantic, and syntactic aspects of language. We have no reason to disagree with Terrace’s conclusions. Realize, however, what this means. Methods of data collection and analysis to date do nor let us determine the limits of the apes’ ability, nor do they much help us to understand the meaning inherent in the apes’ productions. Concluding that there is no unequivocal evidence is an extremely conservative position to take. To be sure, there are problems with some of the methods and interpretations of projects by the Gardners, Fouts, and Patterson. These problems create enough uncertainty so that one is unwilling to take a position supporting some of their claims. On the other hand, there are some methodological and interpretive problems inherent in the Nim project-thus, the conclusions drawn from the project must also cause

I78

CAROLYN

A.

RISTAU A N D W N A L D ROBBINS

one some misgivings. We do not suggest rejecting all available data, but rather consider what each project has demonstrated in an effort to refine further the data collection techniques and our interpretative abilities. There are recent approaches which are beginning to deal with some of the concerns we have mentioned; these approaches are discussed in Section V1. Apes may be more proficient that Terrace intimates; the data do not at all dismiss that alternative.

V.

ARTIFICIAL LEXICONS

The only ape language projects to employ artificial lexicons rather than handsigning as the communication medium are the computer-based lexigram system used by Rumbaugh and Savage-Rumbaugh and their associates and the colored plastic chips used by Premack and his associates. An outline of salient characteristics of these projects may be found in Table 11. A.

THE L A N APROJECTOF RUMBAUCH A N D ASSOCIATES

The initial ape “language” program of Rumbaugh and his associates was the Lana project, which is an acronym for the Language Analogue Project and also the name of the subject, a female chimpanzee. The project received its primary impetus from a newly developed capability to record on computer all the lexigram productions made by Lana and her trainers. With such records Lana’s total “linguistic experience, including her developmental progress, was immediately available and could be stored indefinitely for future analysis. Since the lexigrams were printed, arbitrary combinations of geometric symbols, they were not subject to the interpretation of observers, as signing can be. Lana’s sequencing of lexigrams was automatically recorded on the computer and, therefore, the problem of incorrectly recorded word orders, a pitfall of the Gardners, was avoided. No other project to date has utilized a computerized record of all utterances. Lana lives in a large room on one side of which is a computer console with numerous keys on it. The keys are embossed with lexigrams in “Yerkish,” the language created by the investigators. Typically, active keys were brighter than inactive ones, and a key press made them yet brighter. In addition, at the top of the console the string generated by the key presses was reproduced on a screen in serial order. Strings generated by others were also displayed in this area. Lana could request food, drink, activities, the opportunity to leave the room, and the like. In short, she obtained her necessities and social interactions by communicating via the computer keyboard. Except for computer malfunctions, all of the strings produced by Lana and others, as well as their time of occurrence, are, theoretically, stored in the computer. ”

TABLE I1 AN OVERVIEW OF RECENT APE LANGUAGE PROJECTS: ARTIFICIAL LEXICONS Principal investigator(s)

Premack I

3

Premack I1

Savage-Rumbaugh and RumbaughO

Rumbaugh

Principal investigator(s)' major interests regarding the project

Intelligence; reveal atomic components of intelligence

Do apes have mental states such as belief? development of deception

Linguistic ability of apes; the nature of language; methods and tactics for language training of mentally retarded children

Development of a "word" in a chimpanzee; communication between chimpanzees

Species

Pan troglodyres

Pan

Pan

Pan

Principal subject(s)' sex, age when acquired, and when training began

Sarah(9);acquired at < 1 year old, African born; training begun at 5-6 years old

Lana ( 0 ); acquired at 1.5 years old, born at Yerkes Primate Center; training begun at 2.25 years old

Sherman (8).Austin (8); acquired at 2 and 3 years old, born at Yerkes Primate Center; training begun at 3.5 and 4.5 years old

Living environment

Standard caged laboratory environment In living environment Plastic chips arranged on the vertical on magnetic board; each chip was "word-like" in function

S a r a h ( P ) , B e r t ( & ) , Sadie('?) Luvie ( P ) , Jessie ( 8 ) ; acquired at 1-1.5 years old, African born; training begun at approximately 2-2.5 years old (except for Sarah) Same as Premack I

Large room with computer keyboard

Similar to Gardner and Gardner I and to Terrace, but apes not separated In living environment

Training environment Language medium

Same as Premack I

In living environment

Photographs, videotape, human actors

Computer lexigrams, color coded with regard to gross semantic classification

Computer lexigrams not color coded

Continued

TABLE I1 (Continued) Principal investlgator(s)

Premack I Trainers Number

Qualifications Major training methods

Rewards

Major testing methods

Three principal trainers and at least six who stayed briefly Unspecified Purely observational failed; noncorrection training for ”component skills”; trainers gradually increased utterance length If correct: food, verbal ( G o d girl), hand patting; if incorrect: verbal (No. you dummy), withdraw chips from board. trainer requests ape to give up chips Multiple-choice tests; homogeneous problem sets

Premack I1

Rumbaugh

Some principal trainers and others of brief duration

A few principal

Unspecified

Unspecified

Match-to-sample procedure requiring comprehension of problems faced by other individuals; learn how to acquire hidden foods in cooperative or competitive context If correct: food; if incorrect: no food and removal from room

“Labeling” by associating lexigram and object it denoted; “whole phrase” training

Forced choice photograph tests; production and comprehension tests

Savage-Rumbaugh and Rumbaugh“ Some principal trainers plus others

trainers plus others

Had taught ”language” to apes or intellectually impaired children; “functional” training for lexigram meaning; tool use and social cooperation; categorization training

Food, drink, social interactions, movies, and various events (when correct on tests or when ape generates “grammatically” correct ‘ ’sentences ”)

Food, social praise, tickling games, objects, exploring out-of-doors. etc.; reward sharing

Multiple-choice tests; relatively small problem sets; analysis of “conversations

Interanimal communication in cooperative tool-use paradigm

”

Other data collection methods

Some descriptions of her behavior

Videotape of a subset of trials

Computer has all productions

Interobserver reliability tests

“Dumb” trainer method illustrates reliability and evaluates role of “Clever Hans” cuing

Human observers not in same room or different trainers involved in baiting and testing

Criteria for acquisition of vocabulary item

Performance on multiple-choice tests

Not relevant

Humans not necessarily involved in many interactions with computer; however, human often present during “novel productions” Performance on multiplechoice tests

Size of expressive vocabulary Abilities and major results claimed by researchers

Approximately 130 items Hierarchally organized sentences, causal inference, conceptual classes, color of. name of3 shape of, size of, predicates, some quantifiers, logical connectives, negative article, wh?

Not relevant Ape understood actor’s intention or other mental state; chimpanzees able to convey and utilize accurate and misleading information; indicates capacity for intentional communication

Approximately 75 items Name of, relational concept, numerical concept, novel use of stock sentences, color identification, cross-modal matching

Some information about Savage-Rumbaugh and Rumbaugh work supplied via personal communication ( 1 981).

Computer has all productions; videotape sample of non“linguistic” productions “Blind” tests with ”dumb” trainer who cannot see ape give response

Performance on multiplechoice and tool-use tests; >95% correct on monthly tests of all vocabulary items in all semantic functions (receptive. labeling, statement. and request) Over 50 items Symbolic and communicative use of lexigrams as well as social cooperation (food sharing)

182

CAROLYN A . RISTAU A N D DONALD ROBBINS

Lana was taught to associate a specific lexigram with a particular consequence. Her training required that she pull down on the “go-bar,” which activated the system, and then press a key representing a particular consequence. For example, initially when the key for apple was depressed, a piece of apple was delivered. Lana was next required to precede her requests with a depression of the Please key and to terminate each request with a period, thereby instructing the computer that an expression had been generated. Then a whole-phrase training method was introduced. The keys representing the correct sequence of lexigrams for a request were placed in the same row and electrically connected so that pressing any of these keys resulted in them all lighting up. Gradually the object of the request was separated from the whole phrase, and eventually Lana was required to press each of the keys. At this time the order of the keys was scrambled, although they were still in the same row. After Lana mastered this phase of the program the keys were relocated anywhere on the console. Early in the project the keys were relocated often, at least once a day. Gradually, the keys were relocated less often and eventually relocation was stopped, because the investigators concluded that relocation had little effect on accuracy and only served to increase response time. It is not clear which keys were active for a given test or whether the keys were located in their “typical” positions. It should also be noted that early in training the trainer often entered Lana’s room and “molded” the correct behavior, taking her finger and pressing the appropriate key or pointing to the key. Verbal admonishment often followed erroneous key presses. These various guides would not be part of the computer record. In their early reports of this work (Rumbaugh, 1977), the researchers claimed that Lana had acquired grammatical abilities, could produce correctly ordered long sequences of lexigrams, and had learned the concepts “name-of, ” “same/ different, ” “more/less, ” as well as color identification and cross-modal matching. In subsequent publications, current project members have agreed that the claim for syntactic ability is not substantiated (Savage-Rumbaugh et a f . , 1980a), as we will discuss; in addition, there may also be experimental design problems and interpretive difficulties with some of the other apparent results (see later). In some respects, “Yerkish” language may have served to reinforce Lana’s acquisition of long lexigram strings and to reduce flexibility and novelty in string production. Because of restricted work space in the computer, sentence and phrase conjunction were not permitted. As a result, expressions such as Tim drink cofSee, Lana eat piece of chow, and Lana drink juice and banana could not be used (Rumbaugh, 1977, p. 116). In addition, since the dispenser automatically yielded food in prescribed shapes, only pieces of chow or fruits could be requested, while pieces of M&M (a brand of candy which consists of small pellets) could not. For example, it was illegal to request Please machine give piece of M & M or Please machine give chow (Rumbaugh, 1977, p. 117). There-

LANGUAGE IN THE GREAT APES

183

fore, M&M could not be preceded by piece of, while chow and fruit may have come to function as a single unit, e.g., piece ofchow. Constraints such as these may have caused the chimpanzee to reproduce phrases or sentences in toto and thereby both reduced attempts at novelty and caused diminished understanding of individual lexical items. Another kind of difficulty arises because semantic classes of the lexigram keys were related to colors of the keys. For example, keys representing people and animals as well as machine were violet, ingestibles were red, activities were blue, and so forth (Rumbaugh, 1977, p. 93). Coding semantic categories in this way could have caused Lana to learn sequential color codes rather than sequencing according to grammatical rules. Sequential color learning has, in fact, been demonstrated in pigeons (Straub et al., 1979). Experimental design problems may also exist. A series of experiments on cross-modal matching were interpreted to indicate that naming facilitates such matching (Rumbaugh, 1977, p. 190). Better matching occurred with “named” objects than with “unnamed” objects, 88 and 63% correct, respectively. However, the “named” objects were much more familiar than the “unnamed” objects. The “named” items were very familiar foods, such as bananas and apples, which were always delivered from the computer vending device. The “unnamed” items, however, were hand-fed supplements such as cabbage and carrots-these objects were familiar, but, the impression is given, they were not nearly so familiar as the “named” objects. Thus, both familiarity and food delivery method were confounded with the namedhnnamed variable. It seems reasonable to suspect that familiarity facilitates matching; in the least the confounding should be eliminated. As we have already noted (Ristau and Robbins, 1979), possibly the most serious design problem concerns Lana’s use of the period key. Recall that in order to communicate via the computer, Lana first must pull a go-bar which activates the system. That is followed by key presses indicating her request; finally, all sentences are ended by depressing the period key. The period key is used as a signal to the computer than an expression has been completed. Rumbaugh and his colleagues note that Lana spontaneously began to use the period key as an “electronic eraser.” Consequently, if she were producing a grammatically incorrect sentence, she could “erase” it by depressing the period key and starting again. It is conceivable that Lana meant “erase” or “I did not mean this,” “Ignore this,” when she depressed the key terminating a grammatically incorrect string. However, it is unclear how one can discriminate between this interpretation and one that simply argues that she was using the period key as she was trained to, namely, to terminate a request. It is not obvious how one might go about training a nonhuman subject to use one key to end a sentence (a period), and another key to mean erase, or ignore what I just did. Nevertheless, the gratuitous interpretation of Lana’s use of the period key as an

184

CAROLYN A. RISTAU A N D DONALD ROBBINS

“electronic eraser” leads to the experimenters ’ eliminating lexigram sequences that are inconsistent with a Yerkish grammatical rule. This procedure results in biased data analysis at the most basic level of analysis, for data are eliminated if they are inconsistent with the hypothesis that Lana has grammatical knowledge. This is an extremely serious criticism of the Lana project, which makes it almost impossible to determine Lana’s grammatical abilities. A final point we wish to make regarding the Lana project, previously noted by Ristau and Robbins (1979), concerns the often “enriched” interpretations of Lana’s utterances. With the exception of social interactions not mediated by the computer, there is available, in principle, every interchange between Lana and the computer and between Lana and the various trainers. The computer data could be analyzed in a variety of ways to assess the linguistic competence of the chimpanzee. Despite this, the investigators often use the “conversations” with Lana as evidence of a variety of linguistic abilities. Interpretations of these “conversations” actually reveal “enriched” or “loose” interpretations of the data by the investigators, rather than demonstrating very much about Lana’s lingusitic abilities. At times, the investigators’ interpretations of Lana’s productions involve assumptions about reasons for her errors that either the researchers cannot possibly know or that rely on information the researchers have not offered the reader. Consider, for example, the following conversation which occurred when the experimenter placed cabbage in the vending machine after Lana had requested juice and the experimenter had answered yes to her request. LANA: Please inathine give piece of cabbage. (twice) TIM: ?What in machine. LANA: Tim drink. TIM: ?What in machine. LANA: Tim swing. (This was probably a typing error) TIM: .?What in machine. LANA: Tiin put cabbage in machine. ?Tiin put juice in machine. (Gill, 1977, pp. 239-240).

Notice that when Tim, her primary trainer, asked Lana what was in the machine, she responded with Tim drink, which apparently did not refer to anything Tim did. The investigators ignore this statement. Moments later, in response to the same question, Lana replies, Tim swing, assumed to be a typing error. Tim put cabbage in machine is said to imply her understanding of the past tense. It is not obvious what the basis is for these interpretations. A final example calls attention to the kinds of “data” the investigators use as evidence for possible prevarication, or at the least, obstinacy on Lana’s part. Lana was shown a very familiar object, an orange box, which she had named hundreds of times previous to the incident. In response to repeatedly being asked What name-of this which-is orange?, Lana offered: Can, color, no name-($this

LANGUAGE IN T H E GREAT APES

185

which-is orange (Rumbaugh, 1977, p. 183). Is this evidence of prevarication? It might be, but it also might be recalcitrance, or perhaps Lana was not comprehending what was being asked of her. Recently, Thompson and Church (1980) suggested that Lana’s productions could be accounted for by two processes: paired associate rote learning of exemplars to a specific semantic class and conditional discrimination learning of six stock sentences with variable substitution. They analyzed over 14,000 of Lana’s productions and found that, when no experimenter was present, 91% of all of productions were one of the six stock sentences. When an experimenter was present, 66% were one of the stock sentences, 19% were errors, and 14% were nonstock sentences. They conclude that their nonlinguistic computer simulation model could generate the stock sentences but not the errors or novel productions. We need to keep in mind a few points: the principal investigators of the Lana project have themselves recently suggested that because of the particular training regimen and the constraints of the Yerkish language, Lana may have learned many sequences and engaged i n behavior similar to that of the Thompson and Church computer model. However, the fact that a computer model can generate sequences similar to those generated by Lana does not demonstrate that Lana was using the same processes as the computer to arrive at her productions. We must be cautious in attributing identical causes to what may appear to be similar behavioral patterns, i.e., similar behaviors do not necessarily imply the same underlying processes. It is also of interest to note the large number of novel productions made in the presence of an experimenter in contrast to when no experimenter was present. This difference may indicate the importance of a social interaction in chimpanzee language-like productions, or, alternatively, it may indicate the effects of subtle cuing. When we consider the design problems, the correlation between key colors and semantic class, the enriched and loose interpretations of data, and the use of the period key as an electronic eraser, we wonder what we can conclude from the Lana project. To be sure, it may be viewed as a pilot study-nevertheless, some striking claims have been made based upon the data it generated. Yet recently even the major investigator of the Lana project has reinterpreted Lana’s performance concluding that (1) with few exceptions, there have not even been attempts to determine if the individual signs or symbols actually refer to objects or even to states; (2) there have been no definitive demonstrations that symbols are used in a representational sense; (3) there is no evidence that symbolization has been achieved; and (4)scientists working with apes often interpret ape behavior too liberally (Savage-Rumbaugh et af. 1980a). However, we must note that two new techniques are considered by them to offer evidence of symbolic representation in apes (Savage-Rumbaugh ef af., 1978b, 1980a,b). These techniques are social cooperation in obtaining and using tools and a tool sorting and labeling task; we will comment upon these techniques in Section VI.

I86 B.

CAROLYN A . RISTAU A N D DONALD ROBBINS

THE PREMACK PROJECT

Initially Premack (1 972, 1976) sought to analyze language into its component skills, and then to teach these skills to chimpanzees. He next sought to examine the cognitive abilities or kind of intelligence that underlies the learning of language skills. For example, to learn to use the words same-differenr requires the cognitive ability to make judgments of whether items are the same or different. Some classes of “if-then’’ sentence construction require the ability to make causal analyses, and so forth. The focus of the Premack project, rather than being on language, is on intelligence. In order to study these problems, Premack devised a “language” consisting of plastic chips that were intended to function as words. The chips were pieces of colored plastic shapes with metal backing that adhered to a magnetic board. Sequences were written on the vertical. In the early training situation, the trainer sat beside the chimpanzee, patting and cajoling the chimpanzee to attend and to respond. As the chimpanzee grew older, the proximity of trainer and student decreased until the chimpanzee was on one side of the cage and the trainer on the other side. The major reason for these changes was concern for the trainer’s safety. Premack started with four African-born chimpanzees, although most of the data are from his most promising student, Sarah. The other subjects appeared to have suffered from a variety of apparently emotional difficulties, causing them to be inattentive and difficult subjects. Initially, Premack tried to rely primarily on an observational method to teach the plastic chip language to his chimpanzees, but the attempt failed. Subsequent training procedures entailed an errorless method, with only one plastic chip, the correct one, placed beside the chimpanzee on a table. The chimpanzee’s task was to place the chip at the appropriate place on the magnetized writing surface. After an accurate performance, the chimpanzee was told correct by means of the appropriate plastic word and given pats and pleasantries. The failure of observational learning points to an important difference between chimpanzee and man in the acquisition of language or language-related skills, assuming that human language is primarily a result of observational learning (Lenneberg, 1967), which it may not be. The reasons for the observational technique’s failing are not clear, particularly since field work strongly suggests an important role for observational learning for some chimpanzee skills, as when chimpanzees learn to fish for termites with twigs, utilizing techniques that differ in geographically separate groups (Sugiyama and Koman, 1979). The lack of success in learning the plastic chip “language” through observation may have occurred in part because the chimpanzee did not focus on those aspects of the situation that were of primary interest to the trainers. Furthermore, there seemed to be an inherent ambiguity in relating a word to a particular component of an ongoing behavioral sequence. For example, when an attempt was made to get Sarah to feed a bird and to learn various words associated with that activity,

LANGUAGE IN THE GREAT APES

187

Sarah instead tried to hit the bird. Similarly, in a milk-pouring episode, Sarah never learned the word pour, a term difficult to isolate, but may have associated the plastic chip with milk, the name of the object. Premack’s subjects did successfully master concepts such as sameldifferent, largelsmall, name-of, color-of, shape-of, and size-of, as well as the use of prepositions; the apes also demonstrated knowledge of some quantifiers, conceptual classes, the use of logical connectives, and what appeared to be knowledge of causal relations. Premack’s project, and the Lana project as well, have been criticized for using extensive multiple-choice tests which may merely require rote memory ability. For example, the Gardners (B. T. Gardner and Gardner, 1975) cite Farrer’s (1969) study as supportive evidence for such ability. Farrer presented chimpanzees with four choices of lighted panels with pictures, in all possible combinations (4!, or 24 arrays). The results indicated that the subjects readily learned the correct choice in each of the arrays. The Gardners argue that Farrer’s results could account for much of Sarah’s, as well as Lana’s, performance. In addition, the Gardners (R.A. Gardner and Gardner, 1978) note that test sophisticated rhesus monkeys show Harlow’s learning set phenomenon, namely large increases in problem-solving performance from trial 1 to trial 2 as the monkey solved more and more problems. In some cases performance increased from a chance level on trial 1 to almost 100% correct on trial 2. Therefore, they argue, only on trial 1 of transfer tests are we measuring transfer; after trial 1 new learning is being measured, not transfer. The Gardners then argue that, given Sarah’s and Lana’s extensive training experiences, they are certainly test “sophisticated. ” Analysis of trial 1 behavior would answer these objections. In fact, B. T. Gardner and Gardner (1979) argue that a learning-set-sophisticated rhesus monkey could pass all of Premack’s transfer tests at the average level he required. We know of no data that would support the Gardners’ statement, which may be an “enriched” interpretation of rhesus monkey behavior. Terrace (1979a) has made an extensive review of F’remack’s work and argues that much of it is problem solving. For example, tests of the use of prepositions never contrasted prepositions with one another, but rather, had the chimpanzee demonstrate objective knowledge of a specific preposition. To test understanding of on, Premack required Sarah to identify which object was on top. That is, she was presented with the plastic chips representing ?red on green or ?green on red and was required to answer yes or no. Furthermore, Terrace (Terrace, 1979a; Terrace and Bever, 1976) states that, with few exceptions, most of the problems presented during training sessions were of the same nature and, within a session, the number of alternatives were limited. Terrace argues not with Premack’s methodology but rather his interpretations. He raises the query, “Is Sarah showing discrimination learning rather than language?” There is a brief report available of work by Lenneberg (1975) indicating that

188

CAROLYN A . RISTAU A N D DONALD ROBBINS

high school students performed very well on a set of problems similar to those given by Premack to Sarah. The students, however, did not realize, despite their able performance, that the chips were supposed to represent “words” or that the sequences of chips were supposed to be linguistic. Yet the brevity of this posthumous report and the lack of details of experimental methodology lessens the impact of the results. The Premack project does offer evidence for the representational nature of the plastic chips, although that evidence is not incontrovertible. Sarah, e.g., described characteristics of an apple and a caramel when neither object was in her presence-only the plastic chips representing them were. Also questions concerning the color of objects (?brown color of chocolate; ?green color of grape) were answered correctly. These topics will be discussed further in Section VI. In brief, it is unfortunate that there were not more of these tests and that the tests were not elaborated further. Overall, the Premack project reveals creative and insightful tests of Sarah’s conceptual abilities, although more careful, complete, and precise description of the data is often sorely needed. In particular first trial data are often either lacking or not presented clearly and separately (for further discussion, see Section V1,E). Often, too, the experimental procedures, though conceptually elegant, need to be more thorough, tests need to be repeated more often, and more alternative choices should be offered to the ape. The issue of meaning or representational nature of the chips for the chimpanzee has not been addressed adequately, but this is a fault of most of the ape language work to date. Only recently have others utilized some tests for basic conceptual abilities as Premack has done and only recently has the issue of “meaning” been specifically addressed (e.g., SavageRumbaugh et al., 1980b).

VI.

INVESTIGATIONS INTO MEANING

We will consider the meaning of an utterance first as it refers to the individual items or “words” and then the meaning of the phrase or proposition as a whole. If we consider how researchers have delved into the meaning of individual items in the chimpanzee’s repertoire, we note that at least the following kinds of evidence have been offered as relevant to meaning for a chimpanzee: (a) labeling of an object; (b) novel uses of a word; (c) errors made in generalization; (d) novel word combinations to refer to previously unlabeled objects or situations; (e) feature analysis of an object and its label (Premack, 1976; (f) functional definitions of words: a look at apes’ internal images (Savage-Rumbaugh et a l . , 1978b); and (g) categorical sorting (Savage-Rumbaugh et a l . , 1980b).

LANGUAGE IN THE GREAT APES

189

Which of this information adequately reveals the meaning of a word to a chimpanzee? In the least, we must require that the chimpanzee can exhibit ( I ) generalization of the word to new objects, and (2) appropriate use of the word i n a wide variety of contexts. In any work that purports to demonstrate the chimpanzee acquiring a word, it would be very useful to look closely at the process of word acquisition or symbolization so as to reveal the complexity of the process, steps in the process, and possible similarities to humans’ process of word acquisition. As we consider the significance of these various kinds of evidence to meaning, we must realize that the definition of meaning is a murky issue, unresolved by philosophers and often not considered in its complexities by the ape language experimenters.

A.

LABELING

Training to name has been an integral part of all the ape projects. In the signing projects, labeling was routinely taught and tested by This-n statements and What-this? questions. It is not clear whether trainers also used new signs casually “in conversation” with the chimpanzees or whether all new signs were deliberately taught. Recall though that Premack’s formal attempts at observational learning in a nonsigning project using plastic chips met with little success (Section V,B). The issue of vocabulary acquisition in the signing projects was discussed more fully in Section IV. More formal testing of labels for objects was done using double-blind techniques (B. T. Gardner and Gardner, 1974; Patterson, 1978a). Some of the results from the original Lana project, in particular, serve to illustrate how some of the phrases routinely and correctly used by the chimpanzees to request desired items may be no more than a sequence of responses which lead to reward for the chimpanzee. Specific lexigrams, used many times by the chimpanzee, often apparently entailed little understanding of the individual items. One could describe her behavior as “knowing how” to obtain a reward, but not “knowing that” she was requesting a reward. For example, Lana required 1600 trials to learn the names of two items, banana and M&M candy. Previously she had requested these items hundreds of times, using such stock sentences as Please machine give M&M. The name learning occurred in a paradigm in which Lana was presented with either a banana or M&M on a tray, and was asked the question, ?What name-of this, on the experimenters’ keyboard. She was to respond via the keyboard, M&M name of this. Correct responses were rewarded with the object or amusement of her choice (Rumbaugh and Gill, 1977). The fact that Lana required so many trials to do this task suggests at least two possible interpretations: (1) It took that long for Lana to

190

CAROLYN A . RISTAU A N D DONALD ROBBINS

learn the nature of the naming task, and/or (2) Lana did not know the meaning of the individual lexical items when she had previously requested M&Ms and other objects. She was then given two transfer of training tests. In the first, she was presented with five favored incentives: pieces of apple, monkey chow, bread, a glass of juice, and a ball, in the manner just described. She had often obtained these items through the use of stock sentences. The question was whether she would be able to learn their specific names. Interestingly, she succeeded in naming bull on its first presentation. She rapidly mastered all the other names over the course of ten randomly sequenced trials. [It is unclear whether the 10 trials refer to the number of trials required to learn the remaining 4 items, or if each item required 10 trials for a total of 40 trials.] The second test exhausted the supply of items available at that time-a glass of milk and Lana’s blanket. The fact that Lana correctly named both of these items on their first and all following presentations is evidence, we believe, that she had gone beyond learning the names of the previous seven incentives and had mastered the abstract concept that things have names. From this point on, the mastery of new names has not been particularly difficult, and on occasion, in fact, Lana has apparently inferred some names herself (Rumbaugh and Gill, 1977, p. 171).

This was done by the trainer using new words, e.g., photo slide in a stock sentence, after which, on occasion, Lana spontaneously and appropriately commented slide name-of rhis. Lana’s initial lack of success in the labeling task and subsequent rapid labeling suggests that her difficulties lay in distinguishing the requirements of the labeling paradigm from those of the requesting paradigm she has used so many times before. In further research, Rumbaugh and his colleagues have sought to investigate more fully what “naming” might entail (Savage-Rumbaugh and Rumbaugh, 1978; Savage-Rumbaugh et a l . , 1978a,b, 1980a,b). (Discussion is Sections VI,F and G.) Premack’s work also necessarily dealt with labeling exemplars. The typical way of naming an item, e.g., figs, was to introduce them into a situation of the kind already “mapped” or well known to Sarah, such as giving. Then she would be induced to use jig in a sentence, the other “words” of which she knew. For example, the trainer, Mary, would place a fig before Sarah, give Sarah a pile of four plastic chips, namely, Mary, Surah,jig, and give, and induce her to write the sentence, Mary give Sarah fig. Premack interprets her behavior as learning the name forfig (Premack, 1976). In the least Sarah could substitute the plastic chip for jig appropriately in a variety of sequences. The concept “name of” was taught by means of two positive and two negative exemplars. The procedure was similar to that used to introduce concepts such as “same-different.” For example, the plastic chip for apple and an actual apple

LANGUAGE IN THE GREAT APES

191

were placed with a gap between them, and Sarah was given only one plastic chip, the new one intended to mean name of. By putting that plastic chip between the two items, she created a sequence, intended to be read as Apple nume-of objecr apple (Premack, 1976, p. 162). Nor-name of was formed by gluing the negative particle to the plastic chip for name of. Then the plastic chip for apple and the object banana were placed slightly apart from each other, with not-nume of the only chip available to place between them. She was then tested for her knowledge of name of and not-name of with a series of “wh-” questions and “yes-no” questions for which the answers were always one of two alternatives. Then the terms for name-of and nor-name of were used to introduce the plastic chips intended to represent new items, i.e., items not used in training, such as apricot. In these transfer tests, she was given questions similar in form to the training questions. On both the “wh-” and the “yes-no” forms, she made four errors in 21 trials with no errors on the first five trials (Premack, 1976, p. 164). On these as well as the original training tasks, she tended to make more errors on the negative cases as opposed to the positive. Two kinds of cases were omitted from the transfer test, because of the difficulties of instancing the relations. These were pairs such as red (the name), -red (the color), or more generally the relation between the name of the property and an example of it, and pairs consisting of the name of the action (a verb) and the action. Agents were also unintentionally omitted (Premack, 1976, p. 164). We can wonder if the chimpanzees were learning the “names” of items and were forming the concepts “name-of” and “not-name of.” This is a very general problem of determining the abilities underlying specific behaviors of an organism or particular aspects of a computer program attempting to model artificial intelligence. McDermott (1981) terms the tendency to overinterpret specific and limited capacities “wishful mnemonics. One should instead label a capacity neutrally, e.g., call it 6PX4 and then, by virtue of the actual behaviors of the organism, slowly build a notion of what 6PX4 might entail. The point is well taken. In the case of “naming,” e.g., we might better attempt to specify the attributes of “naming” and suggest the necessary experimental evidence to establish the occurrence of naming. As we use the term in everyday language, a “name” is a symbol. A symbol is many things. In part the symbolic use of a word involves “displacement,” the ability to speak of objects or events remote in space and time; this is one of Hockett’s “design features” for language (Section 111,A). It also involves freedom from task and context specificity so that the symbol can be used cognitively in a variety of ways. Linguists and philosophers have suggested still other characteristics which are beyond the scope of this article. ”

192

CAROLYN A. RISTAU A N D DONALD ROBBINS

B . NOVELUSESOF

A

WORD

The issue of novel word use becomes closely enmeshed with the issue of word generalization and overgeneralization. Novel word use is taken to mean that the chimpanzee has appropriately used the sign in situations or for objects for which neither he nor the trainers have used that sign before. Our designation of “appropriate novel use” often depends upon the chimpanzee’s abstracting features of the environment we consider outstanding and using a word we consider pertinent. Nevertheless, generalization and novelty can reveal a chimpanzee’s concept of the word he is using. To deal with the issue of novelty and its implications for meaning of a label, we need to have adequate sampling and quantitative measures of occurrence of labels and the environmental contexts of their use. The Gardners do mention examples of generalization although neither they, nor anyone else, give quantitative measures for use. For example, they note Washoe’s generalization of the sign for open to mean “for help or permission in opening various doors (house, room, car, cupboard, etc), various containers (bottles, jars, suitcases), and faucets” (B. T. Gardner and Gardner, 1971, p. 147) and for more to be used “for continuation or repetition of activities, and for further portions of food, etc. Often combined: ‘More go ,’ ‘Morefruit. ’ Fouts (1974) noted that Washoe used the word dirty metaphorically. when she insulted him by signing Dirty Roger after he did not grant her request. He also reports that Lucy has used dirry in a similar way (Fouts, 1974). The gorilla Koko responded quite similarly when told by Penny Patterson, I think Mike is smart, I s he smarter rhun you? (Mike is a young male gorilla living near Koko.) Koko, who has often indicated her jealousy toward him, replied Think. . .Koko know Mike toilet (Patterson and Linden, 1981, p. 148). Koko’s metaphoric tendencies were investigated by means of a test that had been used as well with human preschoolers and 7-year-olds. The test first establishes that a subject knows the difference between polar adjectives such as light and dark. Then the subject is confronted with decisions as to whether a certain color is happy or sad, hard or soft, and so forth. Although no details of Koko’s performance are given, her metaphoric senses of words tested were apparently strikingly similar to humans; 90% of her responses matched the test norms, while 82% of the sample human 7-year-olds matched those norms (Patterson and Linden, 1981, pp. 147-148). This kind of test is an intriguing approach to the nature of an ape’s mental images; refer also to Section VI,E for Premack’s work on feature analysis by a chimpanzee. ”

C.

ERRORS MADEI N GENERALIZATION

Some understanding of Washoe’s concept of the sign for flower can be achieved by taking very careful note of her uses and generalizations of the sign (R.A . Gardner and Gardner, 1969; B. T. Gardner and Gardner, 1975). Such

LANGUAGE I N THE GREAT APES

193

work underscores the tremendous difficulty inherent in determining the chimpanzee’s meaning for a sign that we may gloss quite differently with respect to referent and semantic class, i.e., whether the sign functions as a verb, noun, preposition, etc., to the ape. Washoe was introduced to the sign for flower by being presented with a real flower. However, she apparently interpreted this sign to refer to smell, so she generalized it to pipe tobacco and kitchen fumes. Since the sign for flower entails holding the fingertips below the nose and breathing in, an act similar to smelling the fingers, generalizing the sign to odor is not unlikely. Apparently the trainers often drew Washoe’s attention to odor by breathin deeply and obviously smelling (E. S . Savage-Rumbaugh, personal communication). Yet the fact remains, there are many features of the situation that could have been selected as outstanding to the ape and used when employing the sign in new contexts; the ape selected odor or the act of smelling. Patterson also describes Koko’s overgeneralizations. On the one hand, when Koko learned bean, she applied the sign to a myriad of objects-to cookies, shoes, artichokes, jello, a person clowning, and so forth. Patterson and Linden (198 1, p. 83) suggest that Koko enjoyed making the sign-grasping the tip of an index finger with the thumb and index finger of the other hand, then pulling. Later overgeneralizations suggest consistencies in her usage. For example, straw was introduced to mean drinking straw. Spontaneously, Koko made the sign as an apparent label for plastic tubing, clear plastic hose, cigarettes, a pen, and a car radio antenna-all objects for which she had no sign in her vocabulary, and all long and thin (Patterson and Linden, 1981, p. 94). Washoe’s and Koko’s behavior suggest that it might be of interest to present apes with one or a few instances for a sign or artificial lexigram and then note the natural generalization the apes make, thereby suggesting to us what the most salient characteristics of the situation are to the ape mind. D.

NOVEL WORD COMBINATIONS

Word combinations can reveal meaning as they are applied to new situations. Patterson notes that the gorilla, Koko, called a stale cake, cookie rock, and a face mask, eye hat, a Pinocchio doll, elephant baby, a ring, finger bracelet, and numerous other combinations. She has metaphorically termed herself a red mud gorilla and as red rotten mad (Patterson, 1978a,b; Patterson and Linden, 1981). Novel word combinations, all produced by one chimpanzee, Moja, were reported by the Gardners from their most recent project (1979). For instance, Alka Seltzer was dubbed listen-drink, cigarette holder was metal hot, while thermos bottle was called metal cup or drink coffee. Fouts (1974), and Linden (1974, pp. 105-109), more fully report novel word use that occurred in the context of the chimpanzee Lucy categorizing familiar fruits and vegetables, for most of which she did not know the sign. The signs she

I 94

CAROLYN A. RISTAU A N D DONALD ROBBINS

knew that were relevant to food were food,fruit, drink, candy, and banana. Results most pertinent to this discussion are that after her first bite of a radish she termed it cry hurt food and continued to use those signs although previously she had signed other combinations (e.g., that Lucy fruit) for radish. A piece of watermelon was referred to as drink, drink fruit, and candy drink. Though these particular combinations are very striking to us, she also made such juxtapositions as drink food for a radish before she had tasted it, smellpipe food for a stalk of celery, pipe candy and pipe smell for a small sweet pickle, andflowerfruit for a large sweet pickle. These latter sort of combinations may or may not be revealing the chimpanzee’s conceptions of the foods; it is impossible to evaluate for that task. Yet it is reasonable for someone to have undertaken such a procedure. Unfortunately, some reports of the work (e.g., Fouts, 1974) do not mention the less striking combinations, though they constitute the majority of sign combinations made. A now well-known sign combination occurred when Washoe, while on a boat ride, saw a swan in the water and signed water-bird: she did not at the time have a sign for swan in her repertoire. Thereafter she continued to use that sign for swan, even when given the proper ASL term to use (Fouts, 1974). But the details of this anecdote are not given. Was Washoe prompted by questions and pointing gestures such as What this? The interpretation is ambiguous. Was Washoe actually calling the swan a water-bird or was she saying that the bird was in the water? Even if her words merely described two prominent features of the environment, the trainer’s enthusiastic response to her production is likely to result in her thereby learning to call the swan a water-bird. In an extreme view, Washoe might have been severely limited by the words in her repertoire at the time, and might have been attempting to say blue sky but did not have the sign for either word and had only water and bird available. Probably the most important criticism to be made from this anecdote is that one needs a statistical analysis of word combinations and error frequencies drawn from an appropriate, unbiased sample of the ape’s productions in order to evaluate the significance of the “creative naming” (Desmond, 1979; Petitto and Seidenberg, 1979; Ristau, 1980; Terrace et al., 1979a.b). If, e.g., the majority of word combinations are nonsensical, then the few meaningful ones that are reported are completely inappropriate samples and should not be considered within the realm of scientific evidence. Apes have indeed been described as random sign generators (Petitto and Seidenberg, 19791, although mistakes in word combinations are rarely reported in the literature. It has been noted that Nim regularly combined one of his favorite foods and matters of conversation, the banana, “with all manners of words-sorry, drink, tickle, toothbrush, hat, and hand cream-it seems farfetched that on each occasion he was trying to say something profound” (Desmond, 1979, p. 39). Yet it is difficult to dismiss even preposterous combinations like banana toothbrush and banana toothbrush me

LANGUAGE IN THE GREAT APES

I 95

which do occur. As Desmond suggests, even context might not reveal the nonsensical basis, for the chimpanzee mighr have some thought of an absent toothbrush while eating a banana, and could have called for a banana along with a toothbrush so as to clean his teeth after eating. “The problems of interpretation are awesome” (Desmond, 1979, p. 39). We would like to add that playing with words is an important characteristic of children’s language acquisition and, in fact, the apes are sometimes slighted, particularly by linguists and developmental psychologists, for demonstrating little if any word play. It seems quite plausible that some word combinations, such as the banana combinations, are word play. This should not cause scientific despair. Rather, one needs to use an ethological approach to observe word use by both children and apes in an attempt to discern which activities indicate word play and which constitute novel examples of meaning. Perhaps word play is distinguished by many combinations made in a very short span of time or by facial expressions characteristic of other play situations. Most likely the boundaries between word play and novel meaning are blurred. Terrace (personal communication) notes that it is implausible for Nim to be requesting simultaneously a banana and a toothbrush for both were unlikely to be in view at the same time and Nim almost never signed for objects that were not physically in view. The latter fact, if applicable to all the signing apes, is an extremely important kind of information about their use of artificial language for it implies that the chimpanzees very rarely exhibit displacement, a criterion for a symbol. Clearly, however, there are at least limited circumstances in which apes can be required to exhibit displacement, e.g., Premack’s feature analysis (Section VI,E), and the tool use and exchange paradigm of Savage-Rumbauch et al. (Section V1,F). Another sort of novel word combination and word play is the ape’s joking with signs. Some of Koko’s jokes have recently been described, including contextual detail (Patterson and Linden, 1981) that had been lacking in previous reports. In one interchange Koko begins by signing Thirsty drink nose. The trainer asks Koko if her nose is thirsty: Koko says Thirsry. Having gotten some juice, the trainer asks Koko where she wants it. Koko replies, nose, then eye, then e a r , and, finally, signing drink, opens her mouth (Patterson and Linden, 1981, p. 142). Now we can suggest that Koko is stupid and doesn’t know what signs she is making, but her accurate past performance makes such an hypothesis unlikely. The telling piece of evidence in this anecdote is perhaps not emphasized sufficiently in the report. This is the fact that the gorilla laughed, described as “a chuckling sound that is like a suppressed heaving human laugh” (Patterson and Linden, 1981, p. 144). (We will not argue whether it is “like” a human laugh just now, supposing that Patterson has catalogued a number of such sounds made in contexts that lead one to the interpretation that the sounds are being used by the

196

CAROLYN A . RISTAU A N D DONALD ROBBINS

gorilla in ways somewhat similar to human laughs.) Given that we can accept the gorilla’s sound as a “laugh,” we then have independent evidence that by Koko’s interpretation, not just by our possibly ignorant human one, Koko’s signing was funny; i.e., the joke was intended. Furthermore, telling a joke presumes higher order intentionality. These issues are discussed further in Section VII. In all, despite the problems inherent in interpreting meaning of novel combinations, we nevertheless find the occurrence of creative names a most important part of the ape language data. We also strongly suggest the need for gathering the appropriate data to allow statistical analysis of such productions along with careful observations of context.

E.

FEATURE ANALYSIS OF

AN

OBJECTA N D ITS LABEL

Premack “intended to answer a question preliminary to that of symbolization” (Premack, 1976, p. 170). Is the chimpanzee able to analyze an object into its features? Can a match-to-sample technique provide an answer to the question? In Sarah’s first such test, the sample she saw was an actual apple. She was to “match” to this sample by choosing from pairs of alternatives that did and did not instance some feature of apple. The alternatives she saw were a sample of red vs green, a square with a stemlike projection vs a plain square, a plain square vs a circle, and a square with a stemlike projection vs a circle (Premack, 1976, p. 169). Sarah’s task, a familiar one, was to match, according to her own criteria, one from each pair of alternatives with the apple as a sample. None of her choices were rewarded in preference to others and all of her choices were acknowledged with approving tones. In order to maintain her interest in the task, she was occasionally allowed a choice of food rewards. Premack notes that her feature analysis of the apple accords well with human analysis. Her matching data from two tests of 20 and 22 trials in a twoalternative, forced-choice situation reveals that she consistently chose red over green, circle over square, square with stem over a plain square, but was inconsistent on her choice of roundness (7 choices out of 10) vs square with stem (3 choices out of 10). The test was then repeated except that the object apple was replaced with the word for apple, namely a small blue triangular piece of plastic. Her choices for this task were essentially the same, even making the same 7 to 3 split on round vs square with stem. She deviated only on one trial in two tests of 20 trials each when she chose green rather than red. Premack has been criticized for interpreting these results to mean that Sarah had a similar concept for the word apple as she did for the object apple (SavageRumbaugh et al., 1980a, p. 53). This criticism of Savage-Rumbaugh et al. may be justified, for chimpanzees have good memories, and Sarah could well be

LANGUAGE IN T H E GREAT APES

197

merely repeating the responses she made for the object apple when she was presented with the word apple. However, because Premack did the following experiment, the criticism is not substantiated. The critical experiment was to reverse the order of the task, i.e., present choices for the word for an object first and then present choices for the object itself. In such a test, Sarah was shown the word for a caramel and then was given four pairs of alternatives that were intended to instance the shape, color, size, and texture of a caramel, a brown cube of candy wrapped in cellophane. Again, the features she chose generally accorded with human analysis and were approximately the same for the word and the object. With the word as a sample, she deviated from the human analysis four times over 40 trials, while with an object as the sample she deviated five times. She was still 90% correct on the word, about 87% correct on the object, and she agreed about 94% between the features she ascribed to both the word and the object. Finally, she was given a feature analysis of the word apple, using words rather than objects as alternatives when possible. Her agreement between the analysis of the word and the analysis of the object was about 96%. In a different variation Sarah was asked to describe the attributes of the “word” for apple, and was given alternatives irrelevant to the object apple and relevant to the little blue plastic triangle, the name of apple. In this experiment, unlike the previous ones, Sarah’s incorrect choices were specifically disapproved of. Her alternatives were objects and plastic chip names of attributes of the blue triangle, such as round vs triangle. She erred once in the 15 pairs of alternatives. Premack interprets Sarah’s abilities as comparable to using a dictionary in both directions. Given the plastic chip name for apple, she can select the alternative pertinent to the object apple, and given the object, she can do the same for the word. Since both descriptions were made in the absence of the item that was described, “a need for memory or internal representation” was introduced (Premack, 1976, p. 172). Premack later studied the attributes of apes’ internal images with an ingenious variety of experiments. Certain points should be noted about these experiments. No first trial data were reported for these or indeed for almost all of Premack’s experiments with Sarah and the other apes. Since monkeys and apes have shown very rapid learning, even from one exposure, it is a fair criticism that data from trials after the first can reveal new learning rather than exhibiting, e.g., Sarah’s mental image of apple or caramel (Petitto and Seidenberg, 1979; Ristau and Robbins, 1979; Savage-Rumbaugh et al., 1980a; Seidenberg and Petitto, 1979; Terrace, 1979a). However, it is also clear that for certain classes of experiments, such as those on feature analyses with Sarah as a subject, the results were often consistent and first-trial data were often the same as results for later trials. For clarity’s sake, Premack should have presented his first-trial data separately. Premack’s work is open to criticism for lack of such data in some, but by no means all, of his

198

CAROLYN A. RISTAU A N D DONALD ROBBINS

experimentation, and some reviewers have simply been incorrect in asserting that the data were not available (Savage-Rumbaugh et al., 1980a). It is also the case that the experimenters were sometimes not trying to reward specific choices, but instead were attempting to discern Sarah’s conception of an item. Indeed, some of the ape’s alternatives in the feature analysis of the apple presented two positive features, in particular, square with stemlike protuberance vs a circle. In these cases, Sarah’s choice revealed the weight she accorded certain features. However, in the tests with the caramel, the experimenters intended to pit a positive feature against a negative feature, so that Sarah’s choices were either accurate or errors. For these experiments, the trainers gave “approving tones” for all choices. However, it is quite plausible that such “approving tones” might have been unintentionally “more approving” for the more humanlike analyses than for other choices. In all, Premack’s feature analysis approach is a fascinating and useful one. It is unfortunate that neither he nor other researchers have, to our knowledge, extended this technique to a broader range of concepts and with yet more creative sets of alternatives for the chimpanzee to associate with the exemplars. F.

FUNCTIONAL DEFINITIONS OF WORDS:A LOOKAT APES’ INTERNAL IMAGES

The meaning of labels or “names” to the chimpanzee has not yet received enough research interest. The Gardners have catalogued the uses of the signs made by their chimpanzees and have commented that meaning ascribed to such signs by humans and chimpanzees may differ greatly (B. T. Gardner and Gardner, 1971). Study of the development of word meaning has been undertaken by SavageRumbaugh, Rumbaugh, and their colleagues (Savage-Rumbaugh and Rumbaugh, 1978; Savage-Rumbaugh et al., 1978b). They suggest that words at first appear to be closely linked to function. This conclusion derives from their endeavors to train chimpanzees to use graphic symbols to request a tool needed to get food and also from research training chimpanzees to distinguish between edible and inedible objects (Section VI,G). In both sets of experiments, the researchers were attempting to teach the name of an object or class of objects. Yet the chimpanzees’ meaning for a word appeared to arise from the whole context with which the lexigram is associated, in particular with the location of the object and with the activity of the chimpanzees with the object. SavageRumbaugh and Rumbaugh argue that training the meaning for an object’s name through functional use of the object is a far more effective means than a labeling procedure which is the more typical technique in the ape language projects. In a series of experiments Savage-Rumbaugh and Rumbaugh (1 978) investigated whether the symbols used by the chimpanzees were indeed wordlike in

LANGUAGE IN THE GREAT APES

I99

function, or whether such “words” were mere motor patterns which had become associated with various contexts. In addition to Sherman and Austin, who were to become the subjects of the tool-use studies, two other chimpanzee subjects participated in these experiments. The computer-based lexigram system of the Lana project served as the artificial medium of communication. The studies are interesting primarily because they suggest which methods do nor lead to the use of symbols in a wordlike manner. The design of the experiment was roughly the following: Initial training required the ape to press a lighted lexigram key when the experimenter held up an object (later a food); if done, the ape received food reward. The apes failed even after 3000 trials. The apes were next taught to press a key (e.g., the lexigram for M d i M ) of their choice which led to the delivery of the chosen food through a vending machine. In another procedure, an experimenter held up a piece of food, and the ape was expected to push the relevant lexigram key and then receive the food from the experimenter; this “labeling” procedure required the apes to attend to which food was held up-the apes failed. Next Sherman was taught to make the two-unit combinations (verb and object), Give orange and Pour Coke, in order to obtain food. If Sherman had learned anything about the semantic context of give and pour, he should not have used both verbs in one sentence, yet he did so 16% of the time. When he was given transfer tests with milk and banana loaded in the vending machine, objects he had previously requested via single key depressions, he correctly requested any one of them (Give banana or Pour milk) only 31% of the time. As the authors conclude (Savage-Rumbaugh and Rumbaugh, 1978, p. 283), even if an animal reliably depresses a key, or uses a sign or plastic chip to obtain food, he may not understand that such a lexigram, etc., serves as a symbol to represent the food. Another major result occurred during the next phase in which a different vending dispenser was used for each food. When the apes could see which dispenser was being filled, then the subject learned to request only the edible which had just been loaded. Only after such experience were the apes able to learn to request foods which came from the same vending device. In other words, as we have suggested in other parts of this article, location of a food or object appears to be a salient feature to the chimpanzee and the chimpanzee tends to associate a lexigram with that feature although he can be taught to do otherwise. Finally, after all these experiences the apes were able to generalize correct use of two-lexigram sequences to various foods. The authors conclude that the apes’ semantic use of verbs had begun to emerge. These experiments are, of course, highly relevant to the discussion of labeling (Section VI,A) and are the preliminary stages for studying acquisition of word meaning in the cooperative tool use and exchange experiments which we shall describe next.

200

CAROLYN A . RISTAU A N D DONALD ROBBINS

To understand what claims may justifiably be made from the tool use and exchange experiments, we must examine the methods used. The experiments had three phases: I, Learning to request tools; 11, Separating tool from its function; and 111, Inter-animal communication. Each phase is so distinctive from the others that we will describe and comment on each, and then on the experiment as a whole. In Phase I , tool names were taught to the apes by making acquisition of food dependent upon using the tool which they obtained by a request to the trainer. Initially the apes, Sherman and Austin, watched a human bait a site with food, and then use a tool to get the food. The chimpanzees quickly learned appropriate tool use themselves. Although aspects of this procedure involved observational learning, the chimpanzees requested aid by gestures and vocalizations, and apparently received aid in order to learn the function of each tool. The human then labeled the tool with the proper lexigram, e.g., This sponge, and the ape then had to use the correct lexigram to request the tool from the human. Finally, in phase 1, the ape himself had to determine which tool was needed and then to request and use it. In order to compare the ease of name-learning in the functional and naming procedures, the chimpanzees were also taught to label several items: box, bowl, key, blanket, and lock by a typical naming procedure. These items were apparently encountered daily by the animal and so were already familiar to them before the name training began. Apparently, the humans, at some point, labeled an item, allowed the ape to play with the item, and rewarded him with food or social praise if he correctly named the item when asked, although the procedures are not described in sufficient detail to determine this point. The critical result is that after hundreds of trials, neither animal had learned the label for an item in the naming paradigm to the criterion of nine consecutive correct responses. [All object names in both paradigms, however, would have met the Gardners’ (B. T. Gardner and Gardner, I97 1) weaker criterion of one spontaneous usage per day for 15 consecutive days.] The differences between the naming and functional paradigms are manifold. In the naming procedure, the objects were already very familiar to the ape and thus not interesting due to novelty. Also, the trainer indicated no distinctive activities to associate with them. “The motivation and attention of the animals [in the object naming paradigm] were poor” (Savage-Rumbaugh et a l . , 1978b. p. 541). In the functional paradigm, most of the tools were apparently novel items and therefore attracted more attention. They were initially introduced at different sites, with different devices containing food, and the trainer demonstrated distinctive activities to be associated with each tool. Simply put, even for familiar tools, the number of distinctive cues associated with the object and its label was far greater in the functional than in the naming paradigm. However, we do recognize that learning may also be facilitated by the ape’s active involvement with the tools and his ability to obtain a desirable reward

L A N G U A G E IN T H E GREAT APES

20 1

(food) through his own efforts. Such traits of independence are prominent in the linguistically developing human toddler with his cries of I wanna do i f . Early childhood education programs recognize the value to learning of a child doing a task himself, rather than merely observing it. Research in other fields of psychology, in particular, perception, has indicated the importance of active, selfinitiated sensorimotor involvement i n the development of, e.g., limb-eye coordination (Held and Hein, 1963). As we have previously stated (Section IV,B), Nelson (1973) and the Gardners (B. T . Gardner and Gardner, 1979) note that frequent exposure to a referent does not guarantee learning to name the item. Referents commonly named were “changeable, movable, and manipulable” (B. T. Gardner and Gardner, 1979, p. 343), such as lights children switch on and off. Likewise, the chimpanzee‘s more active involvement with tools in the functional paradigm than in the naming paradigm may be the critical factor in the much more rapid learning in the former paradigm, but the present Savage-Rumbaugh reports have no data on that issue. The work is, however, a useful beginning. In Phase I1 of the experiments the name of the tool was separated from its function. This was accomplished by the following procedure: after naming the tool displayed by the experimenter or selecting a tool requested by the experimenter, the chimpanzees did not use the tool to obtain food. The rewards instead were social praise or a token to be used to obtain food on the vending machine. The purpose of this training was to insure that the label, e.g., stick, was associated with the object stick and not with the activity and context of using the stick. If it is functioning as a label, the lexigram for stick should be separable from a specific need-related context. It is important to note the process of acquisition, because it may reveal the cognitive tendencies of these animals. When training began, the chimpanzees’ vocabulary consisted of food and drink names and a few simple verbs (give, tickle, etc.) which were used by the chimpanzees typically to request tickling and other rewards. When the first tool word, key, was learned, it was used whenever the chimpanzee wanted to obtain food that was not held by the experimenter or otherwise directly accessible. With the introduction of the second tool word, stick, the chimpanzees frequently used the wrong tool label for a tool, but very seldom used a tool label for food. These errors did not seem to indicate that the chimpanzees were making distinctions of category. Rather, the chimpanzees were confusing new vocabulary items, all names of tools and all of which they had learned to use in the same new situation, that of inaccessible food. Another particularly interesting common error was a tendency to associate a tool name with a location, rather than with the tool itself or the specific activity involved with the tool. If a chimpanzee consistently tends to focus on location rather than other aspects of a whole context, that could begin to reveal interesting proclivities of a chimpanzee’s ways of thinking. In this particular instance, even

202

CAROLYN A . RISTAU A N D DONALD ROBBINS

when the chimpanzee could correctly request all his tools he could not name them. “In fact, he gave no indication of understanding what the request to name his tools meant” (Savage-Rumbaugh et al., 1978b, p. 543). This tendency towards emphasizing location was also noted when a second tool site and tool function were introduced. When, e.g., a key was used to open a box instead of a door, the chimpanzees could correctly choose the required tool and even use it; but when they had to request the needed tool from the experiment, they could not name key at the keyboard. Instead, key continued to be associated with a specific location for food. As more tool sites and tool functions were introduced, the chimpanzee began to focus on a label being the name for a tool, rather than for location or other aspects of the context. Lana, however, did not experience these difficulties. She already had extensive experience with naming objects and immediately transferred tool requests to other sites, including sites which had to be recalled from memory while the request was made. She was also able to name all her tools correctly when first asked What this? There were striking differences in the abilities of the other animals as well. Austin was quickly very proficient (89% correct) in a naming task (experimenter asks, What this?) and Sherman much less so (70% correct). Similarly, Austin was very accurate (81% correct) in a receptive task (experimenter asks, Give sponge?), while Sherman achieved only 33% correct. After more training his performance rose to 79% correct. Such large differences between chimpanzees who have much the same training experience suggests, not surprisingly, that chimpanzees can differ greatly in cognitive abilities. It may also be that the potential limit of chimpanzees’ accomplishments has not been reached in any current ape language project, and awaits the successful meeting of a brilliant chimpanzee and a training procedure refined by experienced researchers. In Phase 111 of their experiment, the researchers studied interanimal communication, attempting to deal with the problem of intentionality in communication. The critical features of this phase were the following: The animals faced a task which required cooperation. To achieve cooperation, it was necessary to communicate to the other chimpanzee what was needed; that chimpanzee then had to respond appropriately. When an appropriate level of “teamwork” had been achieved, the design called for the chimpanzees to switch roles of communicator and recipient. Essentially, in this paradigm, one chimpanzee saw a site baited with food and could determine which tool was needed to reach the food. By means of lexigrams, he requested the needed tool from the other chimpanzee who could not see the tool site. This other chimpanzee had access to the tools from which he selected the appropriate one and gave it to the first chimpanzee. The food from the site was then obtained and shared between both chimpanzees. In this phase the difficulties encountered seem to center on the chimpanzees becoming distracted, so the experimenter initially needed to coordinate their

LANGUAGE IN THE GREAT APES

203

actions and attention, to remind them of the task at hand, and to correct occasional inaccurate requests or selections of the tool from the tool box. The experimenter also had to enforce the food-sharing after the task had been completed. The chimpanzees did come to share food by themselves, but apparently under threat of intervention by the experimenter. Over the course of 5 or 6 days of experimentation both animals participated in 30-60 trials per day (total of 257 trials); their performance gradually improved with occasional decreases. The scores ranged from both animals being correct 68% of the time on day 1 (individual scores of 77 and 91%) to both animals correct 92% of the time on day 5 (individual scores of 95 and 97%) (Savage-Rumbaugh er al., 1978b, pp. 549550). Several points should be made. Several commentators (e.g., Bernstein, 1978) have noted that cooperative food-sharing done by these chimpanzees seems quite remarkable in light of the infrequency of food-sharing in the wild. But then few occasions arise in the course of field work in which chimpanzees are so critically dependent upon each other in order to obtain food. Furthermore, the chimpanzees in the Savage-Rumbaugh et al. research did not engage in spontaneous food-sharing; humans guided the development of sharing and enforced it. The control condition for nonverbal cuing is not sufficient to demonstrate that nonverbal cues could not be used instead of lexigrams to guide behavior or that nonverbal cues and the lexigrams did not actually guide behavior. The control condition merely revealed that abrupt cessation of lexigram cues disrupted the chimpanzees’ performance. The control and experimental conditions are not equated in terms of duration of training. The chimpanzees had a long time to develop communication via lexigrams. In the control condition they had only 30 trials to indicate their proficiency with or to develop their ability to use nonverbal cues. If the chimpanzees were using nonverbal cues during the experimental conditions, the control condition probably would not reveal that, because trials with a deactivated keyboard were interspersed with experimental trials. Deactivation thus functioned like a faulty keyboard which occasionally did not work. The chimpanzees were disoriented by the “faulty” operation of the lexigram keyboard, but quickly gave correct responses when the keyboard was reactivated. On trials with a deactivated keyboard, the chimpanzees adopted a strategy of either repeatedly offering the same tool or else cycling through all the tools. They did not appear to be innovating gestures, although the authors suggest the chimpanzees might have done so over a longer period of time. Iconic gestures by both experimenters and chimpanzees had apparently been devised in the past, often as a link between symbols and event. From this control condition, one can surmise that the lexigrams were a far more effective cue than any nonverbal cues; this seems to be the point the researchers wish to make. Yet given the same kinds of guidance by the experimenter and time to develop gestures, it is not at all clear that nonverbal cues would not have functioned just as efficiently.

204

CAROLYN A . RISTAU A N D DONALD ROBBINS

The cooperative tool use and food-sharing experiment, just described in detail, was subjected to peer commentary by philosophers, psychologists, anthropologists, and biologists in a 1978 issue of The Behavioral and Brain Sciences. Comments ran the gamut from positive to negative and are too diverse to be summarized here; we highly recommend this issue, including the other articles in it, to anyone seriously concerned with the issue of mental continuity. We find that some major impacts of the tool-use experiments by SavageRumbaugh et d.do not seem to pertain to human language behavior per se. In large part, the experimenters had to deal with the general problem faced in experimental psychological research, in animal training, or in animal circus acts, that of inducing the subject to figure out the nature of the task confronting him. The cooperative nature of the chimpanzees’ task is also somewhat ambiguous, because it is coerced “cooperation,” in that the experimenter modeled the behavior and apparently saw to it that no one got food if i t were not shared; cooperation, as we humans use the term, is supposed to be more spontaneous. The tendency of the chimpanzees to restrict the meaning of the tool labels to certain sites strongly suggests that, if we do wish to compare the chimpanzees’ use of the tool lexigrams with human language, we need to test their comprehension of tool names in yet more contexts (i.e., ability to generalize) and we should look for spontaneous, non-food-related use of the tool names in a large variety of contexts. Only then do we approach closer to the level of “wordness” that humans seem to have, even for simple tool names. Yet this work, or at least the authors’ interpretation of it, is not without intrigue. Recall, e.g., the functional learning vs labeling distinction. By labeling methods, the chimpanzees often required hundreds of trials to learn a label-and sometimes did not learn the label even then-while the word was learned by “functional” methods in tens of trials. Such an overwhelming difference in speed of learning may well reveal vastly different cognitive capacities and kinds of word understanding that are involved in the two training techniques. The functional-labeling difference also suggests an interpretation similar to those offered by Piaget and Bruner (Piaget and Inhelder, 1969; Bruner, 1966) to explain early concept/word learning by a young child. Piaget argues that a child, in the sensorimotor stage of development, incorporates his bodily actions, i.e., his sensorimotor involvements, into the schema or, loosely interpreted, “mental image” of the concept. Thus the meaning of Q hole is “to dig.” Many early childhood curricula are based upon such a “hands on” philosophy, in which children are specifically encouraged to work with their hands and bodies in order better to understand concepts of volume, number, and even objects. Yet the functional labeling difference may simply reflect that an ape or any other organism will respond as it has been taught to respond. If a trainedteacher asks the subject/pupil a question in a manner that precisely imitates what it has been taught, the subject may appear learned. If the question is asked differently, the subject may seem to have little knowledge. Specifically, the chimpanzees had

LANGUAGE IN THE GREAT APES

205

observed a human model both requesting and selecting tools and had done both themselves. In the labeling procedure, the subjects had been shown a label for specific tools but had not been guided to press the appropriate lexigram key when a tool was held aloft. The vastly different results could be that simple-the chimpanzee has to be helped to understand these kinds of tasks. G . CATEGORICAL SORTING

More recent work by Savage-Rumbaugh e l al. (1980b) attempts to deal with whether lexigrams used by the chimpanzees were functioning at a representational (or symbolic) level or whether the lexigrams were merely paired associates with the items they “labeled.” To answer this question the chimpanzees were trained to label the names of three inedibles (sticks, key, and money) as tools, and the names of three edibles (beancake, orange, and bread) as foods. The chimpanzees’ ability to generalize these categories was tested by presenting them with the names of 17 other foods and tools and asking them to label these additional names as foods or tools. “In order for the chimpanzees to make a categorical judgment of this sort on the first trial, it was necessary for them to recall some representation of the actual object” (Savage-Rumbaugh et al., 1980b, p. 922). The subjects were Lana, Sherman, and Austin, all of whom had extensive experience with the Yerkes computer-based language training system. Their training, however, differed in significant ways. Recall that Lana’s initial training emphasized object naming and appropriate sequencing of the lexigrams in an attempt to teach her grammatical structure, while Sherman and Austin’s training emphasized the functions of lexigrams, specifically of tools, and encouraged interchimpanzee communication. The foods and tools used initially by the chimpanzees in this study were those whose lexigrams they knew well; indeed, they scored 100% in blind tests of request, labeling, and receptive skills. The animals learned categorical sorting of the foods and tools by being required to sort the three foods (orange, bread, and beancake) into one bin and the three tools (sticks, key, and money) into another. When they could sort 90% or better correctly, the lexigrams for food and tools were introduced. The chimpanzees were now required to sort a food or tool into its proper bin and then to select the lexigram representing either food or tool. When consistently correct with the training items, additional food and tools were presented, each of whose lexigrams the chimpanzees knew. Each item was presented once, interspersed with trials of training items. The lexigram keys of these “new” exemplars were deactivated but the other 50 lexigram keys remained on. Results indicated that Austin correctly categorized all 10 items on trial 1; Sherman correctly categorized 9 of the 10, missing sponge, the tool which he occasionally nibbles as he uses it; and Lana was correct on only 3 items. The

206

CAROLYN A. RISTAU A N D DONALD ROBBINS

authors interpret these results to indicate that Sherman’s and Austin’s concept of food and tool was “functionally based, generalized and symbolically encoded” (Savage-Rumbaugh et al., 1980b, p. 923). Lana, however, had learned only the ‘‘specific paired associative responses required by her training” (SavageRumbaugh et d.,1980b, p. 923). Yet Lana could correctly sort the items; she simply could not associate the appropriate class lexigram, namely food and tool, with the items or the sorting procedure. The Rumbaughs interpret this inability as a failure to encode symbolically the different functional relations of food and tool. Next photographs, rather than objects, were labeled. Initially, photos of the objects were taped to the objects themselves, and the chimpanzees were required to label the composites. Then merely photographs were used and finally “novel” photographs were used. (The authors do not explain exactly what is meant by “novel” in this situation.) The final and most critical phase of this study was labeling lexigrams. In this phase, the lexigrams for the original training foods and tools were taped to photos of the items; eventually the photographs were removed and the chimpanzees saw only the printed lexigrams. When proficient at labeling the lexigrams with the correct lexigram category, i.e., pushing the key for food or for tool, the chimpanzees were tested on lexigrams for items most of which they had not encountered in this particular experiment. In the least, the chimpanzees had never been asked to make a categorical assessment of the symbols for the items. Austin categorized all lexigrarns correctly on 17/17 trial 1 presentations; Sherman was correct on 15/16 trial I presentations, again emng with the sponge. The authors conclude that an ability to organize perceptual aspects of the world along the dimension of functional similarity does not necessarily give rise to a similar symbolic organization. Lana, e.g., could sort foods and tools, even novel exemplars, but could not transfer the food and tool labels to novel exemplars. The authors suggest that Sherman and Austin could perform these tasks because of their functional training. Savage-Rumbaugh et al. question whether any other apes have reached the level of symbolic functioning achieved by Sherman and Austin. In their view, training of other apes (Nim, Washoe, Sarah, Lana, etc.) “has emphasized only the skills of associative labeling and combining, and these skills alone do not require either semantic comprehension or representational symbolic ability” (Savage-Rumbaugh et al., 1980b, p. 924). In analyzing the significance of this experiment, it seems important to try to understand the nature of the task from the point of view of the chimpanzee. For one, it is not at all clear that the categorical lexigrams intended to mean “food” and “tool” had the meanings we ascribe to them. To begin to determine that issue, the animals would have to sort a variety of obejcts or lexigrams, some of which are properly categorized as food and tool and some of which cannot be so categorized. Perhaps the chimpanzees were instead sorting something of the kind

LANGUAGE IN THE GREAT APES

207

“edible” vs “everything else.” It may be interesting to note that training the chimpanzees to make other categorical classifications, in particular, “items found inside” vs “items found outdoors” met with failure (E. S. SavageRumbaugh and D. M. Rumbaugh, 1980, personal communication). Second, the reason for Lana’s failure in this paradigm may be less a matter of the most appropriate way to teach symbolic understanding to a chimpanzee than it is a matter of two different training techniques, in particular, overtraining on a particular technique, leading to tendencies to behave as one was trained to behave. In point of fact, Lana had been trained over thousands of trials to label an item by pressing the appropriate lexigram key on the computer, when presented with that item. Suddenly, she is asked not to so label the item but to give it a new lexigram label, what we are calling food or roof. In fact we have no data on her tendencies to label the items with their actual “names” because the lexigram keys for the items presented to the chimpanzee were deactivated when the test for categorical knowledge of lexigrams was made. We do not therefore know if Lana repeatedly kept hitting the lexigram key as she had been trained so consistently to do. Rather than reflecting an inability to use lexigrams symbolically, Lana’s “symbolic” behavior may be limited by the context. It may well be that Lana, if placed in a different context, in which the lexigram for the name of the object were not available rather than merely deactivated, or if the entire keyboard were not available and Lana had some other representation of the lexigrams to deal with, she might well have been able to reveal her “symbolic” knowledge. The training experienced by Sherman and Austin also differed from Lana’s in that tools were apparently all initially introduced by being used in a context of inaccessible food. That is, the chimpanzees did not learn the functional use of tools in as large a variety of contexts as a child might. For example, a child often watches an adult use a tool, uses a toy model of a tool from a play kit at home, uses another tool to open a cookie box in the kitchen, uses tools in the nursery school, and also watches someone open a door with a key on television, and so on. Instead, tools were introduced to Sherman and Austin in the context of an experiment about cooperative tool use in which a tool could reach food otherwise inaccessible. The chimpanzee need only recall that context common to certain items in order for him to sort all the items correctly. All our criticisms notwithstanding, the more recent work by SavageRumbaugh ef al. is a notable attempt to come to grips with the unwieldy issue of meaning. Studies expanding the number of categories as well as the specific exemplars used, including extracategorical exemplars, would shed further light on this issue.

VII.

INVESTIGATIONSINTO MENTALSTATES

There has been a recent reawakening of interest in animal mental states and awareness, as evidenced perhaps by the publication of The Behavioral and Brain

208

CAROLYN A . RISTAU AND DONALD ROBBINS

Sciences and by Griffin’s book, “The Question of Animal Awareness” (1976, 1981). At the present, Premack and Woodruff (1978a) are the only ape language researchers explicitly dealing with the issue of mental states of the apes, although Savage-Rumbaugh el al. (1978b) do claim their work with chimpanzee tool use and exchange bears on the issue of intentional behavior. Premack and Woodruff are inquiring whether the chimpanzee has a “theory of mind,” which they define to mean that the chimpanzee imputes mental states to himself and to others. The researchers note that humans regularly impute mental states. Humans infer purpose or intention most widely, but are also likely to infer knowledge, belief, thinking, doubt, guessing, pretending, liking, and so forth. The researchers raise the possibility that the chimpanzee may also infer similar states. Inquiry into the existence of such states raises the question of consciousness in the chimpanzee. Premack and Woodruff, perhaps not wishing to deal with the unwieldy problem of defining consciousness, sidestep the issue by noting that they do not mean to imply consciousness by use of terms like “theory of mind,” but other suitable terms seem to be lacking (Prernack and Woodruff, 1978a, p. 625). Neither do Premack and Woodruff mean to indicate that the chimpanzee is not conscious. Dennett (1978b) suggests that we may deal with the issue of a chimpanzee’s theory of mind by offering an intentional account of these apes, i.e., by attributing to them beliefs and desires o r belieflike and desirelike states (Dennett, 1978a). He asks whether it might be useful to attribute to apes second-order beliefs and desires-beliefs and desires about the beliefs and desires of others. “If so, then chimpanzees have a theory of mind in the requisite sense, for they use the concepts of belief and desire (or concepts importantly analogous) in their own action governance” (Dennett, 1978b, p. 569). If they have humanlike theories of mind, he adds, they would use even higher orders of intentionality. Dennett emphasizes that experiments must demonstrate that imputing a theory of mind to apes is richly predictive of a multitude of results, otherwise requiring adhoc hypotheses. Dennett’s approach to apes’ theory of mind seems very useful and we recommend readers to his commentary in The Behavioral and Brain Sciences (Dennett, 1978b) and await any further analyses of the ape “language” literature. Two lines of research on mental states have been described by Premack and Woodruff. One paradigm requires problem comprehension by the chimpanzee. If the chimpanzee understands the problem faced by another individual, he is said to be infering at least some aspect of the mental state of another individual. The other paradigm deals with deception by the chimpanzee, presumably a form of intentional communication. Chimpanzees apparently learn to deceive a competitive nonsharing human about the location of hidden food, but they are “truthful” to a cooperative humall. In the first paradigm, problem comprehension, Sarah typically watches 30 sec

LANGUAGE IN THE GREAT APES

209

of a videotape of another individual, usually a human, facing a problem or choice. The last 5 sec of the videotape are then repeated over and over, while Sarah is given a choice of two photographs, sometimes three, from which she must select one. The experimenter is out of the room when Sarah first sees the photographs and while she makes her choice, so as to control inadvertent cuing. Sarah’s task in these experiments is actually far more complicated than Premack and Woodruff suggest. Sarah has to deduce that she is supposed to infer that the agent is facing a difficulty andthat the photographs she is given to select pertain to that difficulty. That is, Sarah must realize that the object in the photo can be used by the agent in some way to solve his difficulty, or the photo is a prediction of a possible future event and so forth. Some of the experiments involved problem comprehension in which a desired object, specifically food, was inaccessible, as in the original Kohler problems (Kohler, 1925). For example, a human in a cage struggles to get bananas which were out of his reach vertically or horizontally or behind a box. In another group of experiments a broader scope of problems was encountered. A human actor was attempting to escape from a locked cage, or was shivering and kicking a nonfunctioning unlit heater. The person was distressed by a phonograph that he was unable to play, because it was unplugged. A human was unable to wash down a very dirty floor, because the hose was not attached to the faucet. The pairs of photographs given to Sarah to choose between included one which represented a solution to the problem. The solutions were photographs of a key, of a lit cone of paper (similar to those she had seen used to light a heater), of an electric cord plugged into a socket, and of an attached hose. She made no errors. Then in a second set of trials, the alternatives were refined such that they consisted of a twisted, a broken, and a whole key, or an unlit cone, a burnt out cone, and a lit cone, and so forth. In this series, Sarah made 1 error out of 12 choices; she once chose the twisted key, due, Premack suggests, to the poor quality of the small photograph used. Unfortunately in the refined condition, the initial choice Sarah made in the simpler paradigm was still the correct solution, so Sarah could well be remembering that, rather than exhibiting more discriminating knowledge. More importantly it would seem that we need not infer that Sarah had a belief about what the agent should or would choose to use in some way to solve his difficulty. Her choice merely indicates that she comprehends the problem inherent in the videotape situation, or that she knows what she would do in that circumstance. That entails considerable comprehension, but does not seem to indicate much about a theory of mind, as Premack and Woodruff are using the term. Critics of these studies (Savage-Rumbaugh and Rumbaugh, 1979) have claimed that Sarah could have solved the tasks by physical match-to-sample and associative matching strategies in which she is well-trained. Such strategies would require that she match the photograph of an object either to a similar image

210

CAROLYN A . RISTAU A N D DONALD ROBBINS

in the terminal videotape scene or to an item in the scene with which the photographed object has been frequently associated. However, these criticisms are unfounded, as Premack and Woodruff ( 1 978b) note, essentially because such matching is a necessary, but not a sufficient condition for choosing correctly. Elements in several photographed alternatives can be matched or associated with a videotaped image; Sarah must make the correct match. Furthermore, Premack indicates that stronger evidence that Sarah understood the videotape sequences can be found in the experiments with a trainer that Sarah liked and one that she disliked. Her choices of correct solutions for the human she liked and mishaps for the disliked human reflected her preferences, not matching behavior. It seems reasonable to us that simple, behavioristic notions are not the best explanations of Sarah’s behavior, but neither can we simply ignore them until we have evidence, as we do here, that they alone are inadequate. Premack and Woodruff have also proposed a very interesting experiment, the embedded videotape paradigm, to demonstrate Sarah’s belief about another’s belief. Since some of the concerns about this research are generally applicable to the other paradigms, we shall consider the embedded videotape paradigm in detail. In the embedded videotape situation, Sarah would watch a videotape of another chimpanzee, the observer 0, watching a videotape of still another chimpanzee, participant P, facing some dire problems. As an example, participant P might be locked in a cage confronted with a lion. On the videotape, observer 0 would have to make a choice between a photograph of P successfully finding a correct key and escaping the lion and the photograph of P being mauled by the lion. Sarah is not informed of 0 ’ s choice. Instead Sarah is given two photographs. One photograph is of 0 choosing a photograph of P escaping and the other photograph Sarah sees is of 0 choosing a photograph of P being mutilated by the lion. The critical feature is that 0 either likes or dislikes P, and Sarah is said to be made aware of this. The suggested interpretation is as follows: If Sarah chooses a photograph of 0 choosing a photograph of P being eaten by the lion, then Sarah has a belief about 0 ’ s belief about P, namely that Sarah believes that 0 dislikes P as indicated by the behavior 0 exhibits. One could offer different interpretations of this situation. The central difficulty seems to us to be that Sarah’s belief, which we infer, need not be about the other chimpanzee’s belief (Ristau, 1979). In the given example, Sarah, by choosing the gruesome photograph could be expressing her own attitude toward the participant P trying to escape from the lion. Sarah might not like P. Sarah could have adopted this attitude by observing behavioral interactions between 0 and P, perhaps hostile, that were meant to indicate 0 ’ s attitude towards P. The central issue, however, is to determine if one can design an experiment with chimpanzees to demonstrate a belief about a belief. All the paradigms we have yet encountered can be interpreted as demonstrating a chimpanzee’s belief about a behavior. That is, we infer Sarah’s belief from her behavior. We might

LANGUAGE IN THE GREAT APES

21 1

be willing to do this, because we have seen that Sarah exhibits a variety of intelligent behaviors in other situations, and therefore we might be willing to attribute beliefs to her. We note that Sarah acts differently depending on the behavior of another organism. Presumably her belief is based on the behavior of the other organism. She will, if necessary, predict the behavior of the other organism. But Sarah’s behavior does not seem to depend on the organism having a belief or on Sarah believing it has a belief. Sarah’s belief about another’s behavior will do. Bennett (1978) and other philosophers developing the ideas of Lewis (1969) have suggested the following experiment to demonstrate beliefs about belief. First, one should give a chimpanzee, say Sarah, considerable information about another organism, the agent A, about whom Sarah will presumably form beliefs. Sarah should have evidence that agent A is cooperative and seems to have a similar motivational system. Then one can construct coordination problems so that Sarah can realize that what it is prudent for her to do depends upon what A is in fact going to do. The problems should be such that Sarah is very unlikely to be successful unless she does rely on predictions of A’s actions. Much natural social behavior requires just such predictions. Yet such a task, as Premack also notes, does not require Sarah to have a belief about A’s belief; Sarah need only predict A’s behavior and act accordingly. Other interpretations are also possible. In general, Premack and Woodruff’s use of terms like “a theory of mind’’ seems excessively complex; undoubtedly they are attempting to investigate aspects of chimpanzees’ mental states and have devised ingenious, though sometimes tortuous, experimental situations. They are to be commended, though neither they, nor anyone else to our knowledge, have proposed a viable “theory of mind.” In a second group of experiments, Woodruff and Premack (1979) investigate intentional communication by the chimpanzee by exploring the chimpanzee’s ability to deceive and to resist being deceived. In these experiments a human and a chimpanzee communicated about the location of hidden food. No artificial plastic chips were used; instead, the human and the chimpanzee used and developed nonverbal signals. Each member of the human-chimpanzee dyad served alternately as “sender” and “receiver” of the information. In the initial phase, the chimpanzees were required only to produce behavioral signals for a cooperative and competitive trainer; they did not have to decode accurate or misleading information from a human trainer until the next phase. When the human cooperated with the chimpanzee and gave all the hidden food, if found, to the chimpanzee, the chimpanzee was, from the beginning, successful both at producing and comprehending behavioral signals about the food’s location. When the human and chimpanzee competed for the goal and the human kept all the food for himself, the chimpanzee learned to withhold behavioral information about the location of food, to mislead the competitive receiver, and to discount the sender’s

212

CAROLYN A. RISTAU A N D DONALD ROBBINS

misleading behavioral cues. Woodruff and Premack interpret the chimpanzee’s ability to send and to receive both accurate and misleading information, and to do so appropriately depending upon the context, (in this case the nature of the interactor), as evidence of a capacity for intentional communication. The fact that the chimpanzee can perform both the role of sender and recipient of misleading information is critical to the notion of intentionality. As Woodruff and Premack note, an assumption of truth is an important component of any conversation (Grice, 1967). When this assumption is violated, intentionality may be revealed by the interactants’ ability to suppress information or otherwise adjust their responses to information. It is this voluntary aspect of the behavioral adjustment, dependent upon context, that is central to an assumption of “intentionality. ” There are field examples of deceptive behavior in a variety of species, e.g., the familiar one of ground-nesting birds feigning injury when offspring are in the presence of a predator (Simmons, 1955). Some critics claim the behavior is not intentional, not purposeful, and in the language of the classical ethologists, is “triggered” by the predator’s presence. Yet in the case of the birds, the behavior is dependent on context, although it is not always performed when a predator is near. Focused studies of deception in the natural environment are obviously needed, including careful description of context and preferably experimental manipulation of context. There have been other examples of apparently deceptive behavior by primates (e.g., Kohler, 1925; Menzel, 1974; van Lawick-Goodall, 19711, as Woodruff and Premack themselves note. Menzel’s work is in fact rather similar to Woodruff and Premack’s in that a chimpanzee gradually learns to deceive other chimpanzees about the location of hidden food through altering his behavioral cues. Menzel’s work differs in that the sender of misleading cues does not also play the role of recipient and learn when to use and to discount behavioral signals such as eye gaze and direction and speed of locomotion that can serve as cues to the hidden goal. The experiments using a tool use and exchange paradigm (SavageRumbaugh et d.,1978a) also contained information relevant to deceptive behavior by a chimpanzee. The authors introduced a variant in which the chimpanzees Sherman and Austin requested food from each other, rather than requesting tools used to obtain food. With initial and occasional continued encouragement by the trainers, the chimpanzee recipient of a food request generally complied. The recipient “erred” most often when a highly preferred food, such as chocolate, was requested. At such times, the researchers note, the recipient seemed to ignore the request or to act as though not understanding it, and was willing to give a piece of monkey chow instead of chocolate. Social hierarchy influenced results in that Austin, the lower ranking animal, almost always complied with Sherman’s request, but human trainers (presumably dominant to Sherman) needed to encourage Sherman to comply with Austin’s requests an unspecified proportion of the time.

LANGUAGE IN THE GREAT APES

213

In the course of the Woodruff and Premack studies, a very interesting behavior emerged, that of the spontaneous development of pointing by the chimpanzees. The pointing, an outstretched arm or leg, emerged quickly in one animal and more gradually over the course of more than 100 trials, in the other three. A series of photographs (Premack and Woodruff, 1978b) illustrates a chimpanzee, Sadie, lying to a selfish trainer. First, the chimpanzee extends a limb towards the unbaited container while making eye contact with the competitive “bandito” trainer, and then she looks and points at the unbaited container. Finally, after the trainer lifts the container and discovers no food, “Sadie’s head snaps in the direction of the other container, which she knew contained the food” (Premack and Woodruff, 1978b, p. 619). The emergence of pointing is all the more interesting, because chimpanzees, in the lab or field situation, have never or rarely been observed to point; but chimpanzees do comprehend pointing by humans. As Premack and Woodruff note, the emergence of pointing is not easily explained in the traditional terms of the experimental psychologist. Pointing does not result from shaping or guidance, nor does it depend upon differential reward. It is not explicable in terms of least effort. By the time pointing emerges, the chimpanzee has been successful at obtaining food merely through the trainer seeing her glance in the appropriate direction. It is interesting to note the process whereby behavioral cues were suppressed to the competitive trainer. Approach and “pointing” behaviors dropped out first, while eye movements decreased very slowly; such eye motions consisted of glances toward a container and gazes to the trainer when the body was oriented to a container. All orientational responses decreased, while during some portions of the experiment, subjects tended to orient to the same location, irrespective of the actual location of the baited box. Orientational cues were thus uninformative. Other responses increased, e.g., some that competed with orientational cues, such as climbing the wire mesh and clinging to the aide. Woodruff and Premack suggest that pointing emerged in the chimpanzees because it is easier to control a limb than it is to control cuing through eye gaze; that seems reasonable. It also seems to us that many of the chimpanzees’ responses in the competitive situation that served to suppress information were in fact anxious reactions, e.g., clinging, climbing, and staring at one location. This in turn suggests that deceptive cuing can involve several levels: ( 1 ) anxious behaviors confuse the significance of other cues; (2) correct orientational cues are suppressed, first the voluntary (pointing and approach) and then the involuntary (eye movements); and finally (3) actual voluntary misdirected orienting occurs, as by pointing. It is fine to see investigation underway of problems like deception and to note careful attention to behavioral details, a method actually akin to an ethological tradition. Of course, there are limitations to this initial work: (1) Only one type of

214

CAROLYN A . RISTAU A N D DONALD ROBBINS

trainer (cooperative or competitive) was present during each session so the ape’s facility with deception was not much strained. (2) The methods used to determine interobserver agreement were somewhat peculiar in that the chimpanzees’ response topographies were recorded from videotapes only after discussion and agreement on response type and orientation by two or more of three observers; these observers were familiar with the chimpanzees, but had not participated in the experiments. Interobserver agreement was then apparently calculated on the basis of the “agreed upon” responses; it was 85% o r more for each chimpanzee (Woodruff and Premack, 1979, p. 341). This method does not utilize independent observations since the observers discussed their decision before making it, and inflates interobserver agreement since apparently only ’‘agreed upon ” responses were used in the calculations. (3) The experimental situation is unlike a normal interaction in that the human did not try to “outwit” the deceitful chimpanzee as a normal human or, one would suppose, a normal, hungry chimpanzee might well do. (4) The overriding concern is whether the chimpanzees are indeed practicing deception or whether they have simply learned effective means of obtaining food in different situations. The problem is essentially the one raised by philosophers and even Premack himself, the distinction between “knowing how” and “knowing that.” The chimpanzees in these experiments “know how” to engage in behaviors that we term “deceptive,” but do they “know that” they are being deceptive? The facts ( 1 ) that a new voluntary behavior, pointing, emerged and was used advantageously in two different contexts, to point to the correct or the incorrect container, (2) that a greater number of “anxious” behaviors occurred when the chimpanzees were lying, and in particular, (3) the sequence of actions illustrated in the photographs of a lying chimpanzee (in particular Sadie’s head “snapping” back to the baited container when it was “safe” to look there) seem to us to be strongly suggestive evidence that the chimpanzees are knowingly deceptive. The evidence would be even stronger if the “competitive” trainer were a less anxiety-provoking individual and were as friendly as the cooperative human, though it may be more difficult to conduct the experiment under these conditions. A better method to test the intentionality of the chimpanzees’ pointing might be to conduct an experiment suggested by Dennett and Savage-Rumbaugh (Seyfarth et a l . , 1981). The paradigm is essentially the same as Woodruff and Premack’s (1979), except that on some of the trials, food is put into a translucent container, so that it is plainly visible to the apes and even to the “competitive” trainer. If the chimpanzees were intending to deceive that trainer by pointing to the unbaited container, they should not point on such trials, for their pointing would be “pointless. ” The “competitive” trainer would already know which container was baited. The apes, if they were intending to deceive in the sense of affecting the knowledge o r belief of that trainer, presumably should be aware that the trainer knew which translucent box contained the food. If we were to make an

LANGUAGE IN THE GREAT APES

215

intentional analysis of this situation (see Section III,A), at least the third order of intentionality is required, namely, “The ape wants the trainer to believe that the ape knows which box has food.” If the apes continued to point on trials in which food was visible, we would question the “point” of pointing. Firmer evidence that the chimpanzees are knowingly deceptive would lie in demonstrating that they can form a concept of “truth” vs “deception or falsehood.” Furthermore, since deceit about the location of food is a simple lie, it would be worthwhile to demonstrate the chimpanzee’s ability to engage in a more complex lie. One such lie might require the chimpanzee to take the role of the other in order to determine the accuracy of information presented to him and to know precisely how to lie. An example follows in which a young child, learning how to lie, does not understand the roles possible for the “other”: The child, when asked if she wet her diaper, replied that her daddy did it. We know, but she doesn’t, that Daddy couldn’t have. Possible experimental paradigms to begin to deal with these issues are discussed in Section X. Finally, there is the subject of consciousness. Despite attempts to avoid this unwieldy issue, the problem of consciousness continues to arise from various facets of the ape language research. As previously mentioned, although some human communication is unwitting and presumably some animal communication is as well, much of human linguistic communication is intentional. In this context, intentional can have two meanings, an “everyday ”notion of “conscious purposefulness” and a “philosophical” sense of attributing belief states to the communicator and recipient. Consider first the everyday notion. When we attempt to convince another of the reasonableness of our arguments, we are aware that this is our purpose and we are aware we are communicating about it (note even here our concern with another’s beliefs). Simultaneously we may be communicating other matters of which we are not necessarily conscious. For example, we may be nonverbally communicating either fear or assurance which the recipient understands. If we wish to understand the similarities between animal and human communication, and between the language and communication of the developing child as contrasted to that of the human adults, we should look for evidence of intentionality of both sorts in communication, including both the natural and artificial communications of the nonhuman primates. At some point we should look for evidence pertaining to level of intentionality as well (Section 111,A; Ristau and Robbins, 1982; Dennett, 1978a,b). It behooves us, too, to be aware that even we rational adult human conversers engage in both conscious and unconscious communication. In the course of understanding the ape language projects the issue of consciousness also arises from Cartesian concerns. What distinguishes man from beast; what is it like to be human, to be one of the beasts? Nagel (1974) raises this matter in an article entitled, “What is it like to be a bat?” He considers that conscious experience is widespread, occurring in many levels of animal life, and

216

CAROLYN A . RISTAU A N D DONALD ROBBINS

in countless forms. The central point he raises is that “fundamentally an organism has conscious mental states if and only if there is something that it is like to be thar organism-something it is like for the organism” (Nagel, 1974, p. 436). He adds that even to form a conception of what it is to be a bat, (or an ape for that matter), one must have the point of view of the bat or ape. But to have such a point of view may entail facts which humans never will be able to represent or comprehend, “facts that do not consist in the truth of propositions expressible in a human language” (Nagel, 1974, p. 441). If one can take up such a point of view just partially then one’s conception will also be partial (Nagel, 1974). And Nagel, as we and doubtless others agree, notes that it is very difficult to say what provides evidence of consciousness. Part of the problem is of course not having an adequate definition of consciousness. Some distinctions, however, have been made that appear to be reasonable to note and may finally help us to determine what sort of evidence we should seek to indicate the presence of a form of consciousness in a given species. Natsoulas (1978a,b) distinguishes between “awareness,” (Consciousness,), “direct awareness” (C4),and “personal unity” (Cs), and several other types of consciousness. “Awareness” is the state of being mentally conscious or aware of anything, e.g., facts, external objects, mental occurrences. When we study nonhuman species, we might want to consider “propositional awareness, or awareness of facts, statements, or propositions as a separate matter. C4 is recognition by a thinking subject of its own acts, affections or mental episodes. “[Bleing C4 includes a representation of oneself qua conscious being” (1978b, p. 141). C4 is self-reflexive; i.e., an individual is aware that he is communicating, or seeing a certain color, or intending to ask for a tool. Csis the whole set of one’s mental episodes; it is close to Nagel’s sense of consciousness as what it is like to be a bat or any other conscious being. It is not clear that we can progress very far by avoiding the issue of consciousness. And we do, even in the existing ape language projects and other related research, have at least shreds of evidence that begin to deal with issues of awareness, and, perhaps, of consciousness, too. We are nibbling at the edges of the problem when we consider Premack’s work investigating Sarah’s concept of “apple. That work differs from a concept-formation task per se in that Sarah is being asked to reveal, in a limited way, what it is about the concept of appleness that is important to her. This certainly begins to deal with the issue of awareness. Homing in towards the issue of intention are the experiments on deception that Woodruff and Premack have done, although we still hope for more demonstrations of deception and of more complex forms of deception. It would be interesting to study precursors to deception perhaps by study of other primates and also nonprimates in the lab and in the field, and by looking at the growth of deception in young human children. Premack, e.g., notes that false negatives ”

”

LANGUAGE IN THE GREAT APES

217

precede false positives. In the development of a chimpanzee’s learning to deceive about location of food, at first some of the chimpanzees were more proficient at making false negatives, i.e., leading the human to believe that a baited box was not filled with food (Premack and Woodruff, 1978a, p. 524). Kummer (1981) reports observations he made of hiding behavior by baboons in the field and in a captive colony that suggest to us “knowing” deception by these animals. A female hamadryas baboon in estrus, living under natural conditions, moved away from her adult male leader and copulated repeatedly with a juvenile male. The adult male could not see them for they were behind a rock, although the female did look out at the leader. Sometimes she even approached the leader, presented herself to him (a “reassuring” gesture) and then left to copulate again with the juvenile. In fact on one occasion, a female spent 20 min edging herself into a position, behind a rock, from which she could groom a juvenile male, while the onlooking adult male could see only the top of her head and back, missing her attentions to the juvenile. Based upon field observations such as these, Kummer conducted experiments with captive gelada baboons. In one experiment, each of several females was placed together with a male other than her mate; her mate was behind a wall out of sight but within earshot. When the females began to groom the “wrong male,” the male suppressed a loud yawn typically made in this species at such times. As Kummer notes, “The male’s assessment of the situation seemed to agree with that of the experimenter. ” The more complicated lie we would wish to see demonstrated is of especial interest, because it requires an understanding of the capacities of the “other” and roles possible for the “other.” Such a lie can, in fact, require predictions about the other concerning behaviors never directly observed by the lying chimpanzee. This is not the same as understanding the sense of “self” of the other, but these kinds of knowledge do seem to begin to deal with the edges of the problem. If a chimpanzee can form a notion of an “other,” it may be reasonable to speculate that she can likewise form a notion of her own “self.” [These are not new concerns; anthropologists have long been concerned with the evolution of sense of self (Hallowell, 1955). However, anthropologists tend to assume that only human beings have a sense of “self” and then ask where in human evolution such consciousness appeared.] Possible examples of self-description may also reveal a sense of consciousness which is something other than mere awareness. There are now, within the ape language literature, examples of self-description which do not seem to require an understanding of self as a mind, but let us note a few of them. Washoe, while tumbling about and putting her head into a hat, has been seen spontaneously to sign in or head in both before and after that act (from a film of Washoe, as reported by Premack, 1976, p. 91). One of Premack’s chimpanzees, Elizabeth, was involved in a fairly openended situation with such objects as an artificial apple, a knife, a container, and a

218

CAROLYN A. RISTAU A N D DONALD ROBBINS

bowl of water, as well as a small pile of relevant plastic chips. The chips, however, did not contain the plastic pieces necessary to request food to eat. On 16 occasions, Elizabeth cut the artificial apple with a knife and, on 6 of those occasions, she displayed the chips representing Elizabeth apple cut. She received bits of apple whenever she made a complete three-item sentence (whether accurate or not). These are intriguing examples, though we recognize that they are vulnerable to the same flaws of interpretation that have been made regarding other facets of the ape language projects and which we have criticized. The examples are few; we tend to add much of our own meaning that we give to words and various classes of acts to meaning we attribute to the chimpanzees. Yet these examples should not be ignored. With effort we may possibly accrue more examples and devise other demonstrations of self-description. Taking our cue from personality tests applied to humans, we might even devise a multiple choice “questionnaire” containing descriptive, “if-then’’ statements or other statements requiring causal analysis. In these various statements, the chimpanzee could be represented by a symbol, and then would be required to choose from a variety of alternatives, which describe what she does, what her feelings state is, her preferences, and so forth. There is already a set of experiments which deal with the issue of selfrecognition; this is work done by Gallup involving chimpanzees as well as monkeys and nonprimate species (Gallup, 1977). When chimpanzees are first shown a mirror, they will gaze at themselves and engage in other-directed behavior, such as grimacing, until about the third day. At that time self-directed behavior sharply increases. Examples of such behavior are grooming parts of the body, especially those parts not normally seen without the aid of a mirror. Chimpanzees with a previous history of social isolation as well as all of the many species of monkeys and nonprimates tested do not engage in such self-directed behavior, nor do they exhibit self awareness of the sort illustrated by Gallup’s well-known experiment. Gallup based his experimental paradigm upon the observations just described. He painted red dye on the upper eyebrow ridge and opposite upper ear of an anesthetized chimpanzee. When dry, the dye did not rub off on one’s finger and was, presumably, tasteless and not felt. The chimpanzee was allowed to recover from anesthesia and was placed in front of a mirror with which he was familiar. The chimpanzees then proceeded to investigate and groom the places painted with dye by looking in the mirror. The chimpanzee then typically sniffed at the fingers that had touched the dye; no monkeys did any of this. Note that these experiments do not require an interpretation that the chimpanzee has an awareness of self as a mind; an awareness of self as a body will suffice. How does one demonstrate, in a convincing way, an awareness of self as mind, the existence in a chimpanzee of a belief about a belief, or the ability of a

LANGUAGE IN THE GREAT APES

219

chimpanzee to prevaricate? Premack suggests we need many different tests, all converging toward the same conclusion, namely that a chimpanzee has a belief or, we might add, that he “knows that” he is deceiving. However, a basic difficulty in interpretation still seems to remain. For those who believe simplicity is a virtue, it is not clear which is the simpler interpretation, that Sarah has a belief about another’s belief or that Sarah has a belief about another’s behavior. We humans, even as children, easily speak in terms of mental states of others, using them so broadly that states like “liking” or “happy” are attributed by children even to toys. It is difficult, however, to be a behaviorist and seems to require long years of training to make predictions in terms of specific behaviors, rather than in terms of general mental tendencies. Which then is primary for a creature as complex as a chimpanzee, to have a belief about a belief or to be a strict behaviorist? Given the familiarity we do have with chimpanzees, we, the authors, feel intuitively comfortable with the hypothesis that a chimpanzee could have some notion, perhaps of special relationships between individuals, that chimpanzee Adam likes chimpanzee Eve or that I like my child. Are there some kinds of evidence to which we as scientists and philosophers are resistant to which we need to have access in order to deal with phenomena such as awareness and mental states, phenomena we do deal with in our everyday life?

VIII.

RELATIONTO ANIMAL COGNITION AND NATURAL ANIMAL COMMUNICATION

In this section we will attempt to relate the apes’ artificial language abilities to cognitive abilities studied in the laboratory and to some abilities apparently exhibited in animals’ natural communication. We will be concerned with what animals naturally communicate; what aspects of the apes’ abilities, as revealed by the ape language research, are used by the ape in the natural environment; the possibility of demonstrating linguistic-like abilities in the laboratory for species other than apes; the evolutionary forces that may have led to linguistic-like abilities, if and insofar as they exist for the ape; and, finally, the implications of natural animal communication for improving investigations into artificial ape language, ape mental abilities and states. A.

RELATION TO A N I M A COGNITION L

A number of critics of the ape language research have been concerned that the apes’ productions (and comprehensions) have not been linguistic. The critics do, however, agree for the most part that the behaviors exhibited by the apes reflect problem-solving and memorial abilities. The present ape language work can in this way be related to the classic work of Kohler (1925) and of Yerkes and

220

CAROLYN A. RISTAU A N D DONALD ROBBINS

Yerkes (1929), who investigated cognition in apes some 50 years ago. Kohler, e.g., studied “insight” learning and used problem-solving tests such as those requiring apes to put two sticks together or to stand on a box to reach otherwise inaccessible food. Premack and Woodruff ( 1978a,b) adapted these paradigms in videotape sequences used to investigate an ape’s attribution of mental states. The Yerkes specifically studied memory. Memory is a prerequisite for the design feature, displacement, the ability to communicate about an object or event in the absence of the object denoted. Since the ape language work could reveal important characteristics of an ape’s cognitive processing system, it is reasonable to look for related contemporary research in experimental animal cognition. However, experimental psychologists have traditionally been divided into human and animal “camps”-with the problems of each often quite disparate. Further, within the area of animal learning a division often exists between nonprimate and primate research, and still further between studies of great apes vs other primates. Thus, different methodologies, techniques, stimuli, issues, and even theories have often emerged from these presumably divergent lines of research. Recently, there have been attempts to synthesize the concepts applied in these subsets of experimental psychology. The work of Premack and his associates (Premack, 1976; Premack and Woodruff, 1978a,b) and the recent Savage-Rumbaugh ef al. (1980b) research deal with the representational capacity of the apes. Some of the ape abilities that have been revealed or at least suggested by this work are indicated in Table 11. Earlier research has shown that apes can form categorical representations of fairly traditional laboratory stimulus dimensions-shape, size, color, and the like (Robbins, 1977). Savage-Rumbaugh et al. (1980b) have shown that apes can categorize food and nonfood objects quite well. Such capacities lead one to inquire about the nature of the representation that must persist in memory (Robbins, 1977), so that the ape can perform memory tasks as varied as a delayed matching to sample and a delayed discrimination task (Medin et al., 1976; Robbins, 1977). Most recently, Premack and his associates have sought to investigate reasoning abilities in apes, including analogical reasoning, transitive inference, and primitive mathematical concepts (Gillan et a l . , 1981; Gillan, 198 1 ; Woodruff and Premack, 198 I ) . In future work they plan to investigate the impact of linguistic training on the ape’s reasoning capacities. A cognitive “revolution” has taken place over the last 15 years in the animal learning and memory area. Instead of the former emphasis on stimulus-response interpretations, the cognitive view has made the field more receptive to ape language research. Thus, although controversy exists regarding the nature of the apes’ linguistic abilities and achievements, it seems that most accept the idea of studying cognition and cognitive processes in nonhuman animals, and, in particular, great apes.

LANGUAGE IN THE GREAT APES

B.

22 1

WHAT Do A N I M A LNATURALLY S COMMUNICATE?

The traditional interpretation held that animals communicated only about their emotional states and immediate needs. The classical ethologists, in fact, considered there to be a reflex-like, lock and key mechanism whereby a signal (the Sign stimulus) from the environment or from a conspecific or predator released (via the Innate Releasing Mechanism) a response (the Fixed Action Pattern) which could in turn be a communicative signal to release the next response from the recipient of that signal. Potentially a long chain of communicative behavior could arise in this way, as instanced by the ritual courtship of the stickleback mating ”dance. ” More recently the reflexive aspects of this interpretation have been deemphasized. This seems necessary, especially as one attempts to understand the communications of “higher” animals and/or as one looks more closely at the apparently reflexive sequences of at least some of the “lower” animals. The role of context has come to be a more important consideration in understanding communication. W . John Smith (1965, 1968, 1977) has proposed that there exists a limited set of messages which are used at least in the vertebrate animal communication systems. He distinguishes between the message of a particular signal and the meaning a specific recipient ascribes to that signal, such meaning being dependent upon context. For example, in the early spring, the song of a territorial male bird may be “interpreted” by an unmated female as something akin to “I am an unmated male of species x; approach for possible mating,” while an adult male bird may interpret that same song or call as *‘I am a male with territory; I have a high probability of fighting with another male who comes onto my territory.” If we observed the responses of these animals, we could probably gather evidence that both meanings existed. [Smith’s ideas and their relation to ape artificial language skills are discussed further in Ristau and Robbins (1979).] Smith’s ideas are particularly important in that they attempt to propose a system for investigating and explaining animal communication that can potentially relate animal communication to linguists’ systems for human language. One of the vagaries of Smith’s system lies in the difficulty of defining context. In an effort to avoid these difficulties, Marler (1 977a) and Green (1 975) propose a different interpretation. Green (1975), based upon his work on the vocal communication system of Japanese macaques, suggests that the vocalizer selects a set of addressees, i.e., potential recipients specified perhaps as to sex, status, age, mating condition, and so forth. In this way, some of the aspects of Smith’s term “context” are accounted for in a different way. Griffin (1976, 1981) expands the range of potential topics in animal communication, by suggesting that animals might communicate about a large variety of intentions, directions, and semantic referents. In particular, he emphasizes the waggle dance of the honeybee as a symbolic achievement by which the honeybee

222

CAROLYN A . RISTAU A N D DONALD ROBBINS

can communicate the distance and direction of a food source to its hive members, as well as indicate the overall desirability of the food, based, e.g., upon the richness of the supply. The bees can also use the dance to communicate about matters other than food, such as the location of water, or even no doubt about subjects never before encountered in the lifetimes of those particular bees, namely the desirability and location of a potential new cavity for the hive. At present, then, researchers in animal communication are expanding their ideas about what it is animals do communicate. If we believe that devising an artificial language for apes may help us understand man’s language and that apes may have precursors to human language, then it is reasonable to look for aspects of natural animal communication that are similar to man’s. Some similarities are obvious, particularly in man’s nonverbal language. The traditional topics of mating, aggression, and flight are clearly expressed nonverbally by man. More recent suggestions by Smith of messages of “association ” are also expressed by man nonverbally, linguistically, and in ritual. Other aspects of human communication such as greetings and leavetaking ceremonies (e.g., goodbye, see you later), passing by each other along a narrow way or through a door as befits our respective statuses (Goffman, 1963) find parallels to similar communications in animals. The attention-drawing capabilities of verbal and nonverbal human communication likewise find their parallels in animal communication. Some examples are people watching an event draw other people to watch; a glance directs the attention of a recipient; group members gather and leave in concert, often through nonverbal signals. The social status of a particular individual can often determine the attention paid to its communications; and the attention structure of a group, determined through analysis of patterns of communication, can be a useful way to understand the group’s social relationship (Barkow, 1976). Furthermore, whether animal communication can be informative as well as manipulative is a concern not only of the philosophers we have mentioned but of biologists as well (Dawkins and Krebs, 1978) (although the sense of “manipulative” differs for the philosophers and sociobiologists). A very important distinction to make when attempting to compare human language and animal communication is to distinguish between the functions of language and the specific images or referents communicated. For example, sitting with a small group of friends before the fireplace sipping brandy and talking about the cold weather, a book, or old times is probably less importantly “about” the weather or other referential information, and more importantly concerned with bond maintenance. In that way it may profitably be compared to a small group of gelada baboons sitting together in the fading sun softly grunting in their various ways. In fact the gelada baboons in that situation may be communicating about matters which we have as yet no means to investigate. With this illustration of the gelada baboons, we raise another issue, namely, the

LANGUAGE I N THE GREAT APES

223

presently available methods and concepts we have to obtain data about animal communication. The methods and points of view or “stances” (Dennett, 1978a) we now use may not allow answers very different from those we already have. For instance we typically record, on audio tape or even film, the behavior and communication of the communicator, the general environmental and social context, and the response of the recipient to the communication. The communication, often vocal, may have occurred naturally or have been provoked by playback over a tape recorder. At present, we do not typically allow very many possible choices for the nature of the recipient’s response. The alternatives are either precise objective descriptions of some behaviors, or broader interpretive descriptions like maintain association, leave, agress, and so forth. Griffin indeed suggests there may be a variety of other answers, such as intentions (1976, 1981). Yet, too, as Smith suggests (1977), the effect of an utterance may be to store information. We may see no nonvocal response of the grunting gelada baboons or the mumbling humans, though we assume that the humans, at least, are communicating. The implications we can draw from natural animal communication for the ape language research or human linguistic communication are severely limited by our methods and “stances” in animal communication studies. Taking an intentional stance, and framing observations and field and lab experiments in terms of levels of intentionality (Dennett, 1978a) in order to appreciate the mental complexity of the presumed communicatory acts, should be an exciting and extraordinarily fruitful “next step” in our endeavors.

c.

WHAT ASPECTS OF T H E APES’ “LINGUISTIC” ABILITIES ARE USED I N T H E NATURALENVIRONMENT?

Premack considers that his experiments in ape “language, ” and presumably the work of other researchers as well, “map” or reveal preexisting underlying cognitive capacities (Premack, 1976). Presumably these capacities should exist in the natural environment. Ristau and Robbins (1979) speculate that there are chimpanzee behaviors already described from field research (e.g., by van Lawick-Goodall, 1968, 1971, 1973) that reflect many of the chimpanzees’ cognitive abilities revealed in the lab. Some behaviors could even be viewed as precursors to some simple grammatical structures. To use a grammatical structure implies, at least, a knowledge of certain kinds of relationships. Subject-verb-object structures can be related to the categories of actors, acting, things acted on. Chimpanzees, known to be capable of tool construction and use, might be presumed to have an appreciation of those categories of relationships. An understanding of the preposition “in” and the grammatical structures in which it is involved could also be revealed by chimpanzee tool use, e.g., by the act of poking twigs into tunnels in termite mounds so as to collect termites. An understanding of sequencing of

224

CAROLYN A . RISTAU A N D DONALD ROBBINS

activities and implicit “if-then’’ relations would seem to be required by the cooperative hunting activities of chimpanzees which often entail the roles of an “enticer” and a “grabber” of the prey. Presumably the grabber must know what to do “then,” “if” the enticer does one thing or another. Such cooperative hunting bears at least some resemblance to the coordination type problems proposed by Bennett (1978) and by Lewis (1969) to reveal a chimpanzee’s understanding of the knowledge or belief (or predictable behaviors) of another chimpanzee. Insofar as hunting entails a division of roles, and if it can be shown that chimpanzees often switch roles, we are gathering at least suggestive evidence that bears on the notion of a role, or “self” and “other” (see Section X,C). Existent data from recent field work with adult and young chimpanzees have been interpreted using some linguistic concepts, in particular “perlocution” and “illocution” (Plooij, 1978). As defined by Bates et a l . (1973, “perlocution” refers to a signal issued by one person which must have some effect, intentional or unintentional, on a recipient, while “illocution” is the intentional use of a conventional signal to carry out some socially recognized function such as commanding or indicating the presence of events or objects. The young human infant’s preverbal communications are seen as developing from perlocutions to illocutions, the critical changes, being the infant’s new-found knowledge of the meaning of his acts to his mother and his intentional (“on purpose”) use of his acts in accordance with that knowledge. (Note that such knowledge and usage seem to require higher order “intentionality,” e.g., the infant must be able to cognize: I think that mama thinks I want to p l a y or I think that mama thinks I will cry.)

As an example of the chimpanzee’s similar development, Plooij suggests the following: When the mother chimpanzee is playing and tickling her infant, say in the neck, he will “defend” himself by bringing his hands up over his shoulder, trying to push mama away. This produces a characteristic posture. At an age of about 11 months, infant chimpanzees begin to initiate play sessions by directing such a posture to mother or others. Thus, says Plooij, the infant is using the signal to control his mother’s behavior; “the border between perlocution and illocution . . . seems to have been crossed” (Plooij, 1978, p. 117). Initiations of grooming and various other social interactions develop in similar ways. Plooij makes good common sense, though we might want to see more evidence that the signal usage is intentional through data that indicate the young ape becomes upset if the mother does not respond or does not do so appropriately; and in such instances, the ape tries a variety of means to achieve his presumed goal. Plooij also cites incidents which suggest that the adult chimpanzee has an ability to combine gestures, sometimes using an apparent attention-drawing vocalization in these combinations. Furthermore, he states that adults can use a gesture in strikingly different contexts. These abilities are, you will recall, important in Thorpe’s and others’ criteria for language.

LANGUAGE IN THE GREAT APES

225

Another ability, the capacity to form concepts or categories, as revealed by the chimpanzees’ abilities to label and to categorize (Savage-Rumbaugh et a l . , 1980b), would seem, prima facie, to be useful to the chimpanzees in a real world where they distinguish between edibles and inedibles and, indeed, between the subsets of categories of good-tasting berries and not-so-preferred berries. Semanticity or actual labeling of objects in the real world has been suggested by field work not only with apes, but with monkeys and ground squirrels. Sherman’s ( I 977) work with ground squirrels indicates the use of different vocal alarm calls to different predators. Kavanagh’s (1980) field work with monkeys includes slightly suggestive evidence for plasticity in the African monkey Cercopithecus aethiops tantalus, formerly a savannah dweller, which had invaded the forest. In the original savannah habitat, C. a . tantalus typically responds to the presence of a dog by climbing into a tree and making specific loud alarm calls, which, however, render the monkey conspicuous. In the forest, where dogs are often accompanied by hunters with rifles who can thereby attack even monkeys in trees, the monkeys apparently inhibit their vocalizing and make themselves as inconspicuous as possible (Kavanagh, 1980). The most convincing evidence for semanticity derives from experimental manipulations done in the field by Seyfarth and Cheney (Seyfarth et al., 1980a,b; Seyfarth and Cheney, 1980). Their work is based on Struhsaker’s (1967) observation from extensive field work with the vervet monkey that the monkeys have different alarm calls made in the presence of different predators, and, in fact, the response to each alarm call also differs in a manner so as to provide optimal protection from that class of predators. Seyfarth and Cheney played back tape-recorded alarm calls of individually known companions to vervet monkeys. The tape recorder was hidden behind obstructions such as a bush, where the individual monkey might likely be. Recordings were played only of vocalizations from monkeys not in sight of the listening monkeys. This was done to preclude confounding the results by monkeys being able to identify the vocalizer individually on the basis of the sound of the alarm call. The behaviors of the respondent monkeys were filmed before, during, and after the vocal playback. The researchers did indeed find that monkeys responded differently and appropriately to the various classes of vocal alarms. For example, the monkeys looked upward and then dove for cover in the bush when the “eagle” alarm was played, and climbed high into trees when the call for baboon or “ground predator” was played. The monkeys were also seen to behave appropriately to poisonous vs nonpoisonous snake alarm calls. Another particular interesting aspect of this work (Seyfarth and Cheney, 1980) was the ontogeny of vocalization use by vervet monkeys. At first young monkeys used the “eagle” alarm call quite broadly; when very young, they might so vocalize to a falling leaf. As they grew older, their use of the call became more restricted and precise. The call might, early on, be given to a

226

CAROLYN A. RlSTAU AND W N A L D ROBBINS

variety of birds flying overhead, then to the general class of raptors (birds of prey), then to raptors which resemble the martial eagle flying overhead, and finally to the martial eagle itself, apparently the main flying predator of vervet monkeys. Just how the vervet monkeys learn to make their calls more precise is of course not clear from this initial work. “Search image” or innate template notions might be relevant in explaining the monkeys’ developing ability to focus on the object appropriate for making an eagle alarm call. Attention to and learning from vocalizations made by older, more experienced troop members is another reasonable explanation that may also be involved. The latter kind of plasticity is most directly relevant to the ape language learning work, although observetional learning by itself is apparently not adequate for developing the apes’ proficiency with artificial lexicons (Premack, 1976) or signs. Natural proclivities exhibited by the vervet monkey (recall the example of young monkeys vocalizing eagle alarms to a falling leaf but not to a ground predator) that may direct their learning, may also operate in the development of ape’s natural communication and in the artificial language learning of apes. Recall Washoe’s generalizing the sign for “flower” to other odors rather than to flowers; we may also note that in the failures of observational learning, Sarah tended to focus on matters other than those to which the experimenters wished to draw her attention (Premack, 1976, pp. 53-57). Evidence for semanticity, such as that from the vervet monkeys, has been criticized by Bennett (1978) as insufficient to distinguish between “statements” which aim to produce belief, awareness, or knowledge (e.g., I see a martial eagle or Thur is a marrial eagle) and ‘‘injunctions. ” which aim to elicit behavior from the other party (e.g., Flee! Jump into a bush!). Bennett’s criticisms are well taken, yet as Griffin (1981) notes, gathering evidence for the existence of a specific injunction is a difficult enough first step; at some later point we may be ready to worry about how to gather evidence about the making of a statement. Suggestive evidence for the emergence of abstract signaling and the possibility of the signal becoming “disengaged from context” (Bronowski and Bellugi, 1970) does exist. Male pygmy chimpanzees, observed in captivity, use gestures to direct the position of the female into a variety of different postures for copulation (Savage-Rumbaugh et a l . , 1977). Some of these gestures involve pushing or touching her body to move it. but others are made from a distance, indicating which part of her body is to be moved and the position it should assume. This method of communicating is somewhat more sophisticated than pointing. Once the hand motion no longer involves touching the body, that signal would seem likely to be available for use in other contexts. It might be fruitful to investigate such use, preferably in the field. The existence of complex sequences, particularly in animal vocalization, has also been noted in the field; this would be, of course, a precursor to any possible grammatical structure. Marler (1977b) notes that combinations made in the voc-

LANGUAGE I N THE GREAT APES

227

alizing of nonhuman primates have meanings different from that when a segment of the vocalization is made in isolation. For example, an alarm vocalization may be part of a more complex sequence, in which case the primate does not respond to the alarm segment as it typically does. Moynihan claims a primitive grammar for neotropical monkeys (Moynihan, 1976), but that claim is controversial. The awesome, lengthy, complex songs of the humpback whale are composed of themes in a fixed sequence. While the order of sounds in their themes progressively changes, new themes are sometimes added, and others may drop out-still, the order of themes remains invariant (K. Payne, P. Tyack, and R. Payne, personal communication, 1981). Certain species of birds, e.g., some wrens, have songs constructed of well over 100 song types. Transitions between song types are highly predictable, thereby at least raising the possibility that information is available in the ordering of song types or deviations from it (Verner, 1976; Kroodsma, 1980). Other sequences also often occur in bird vocalizing, but whether these complex sequences have meaning due to the particular sequence, as opposed to the elements in that sequence, is very much an open issue. D.

ARTIFICIAL“LANGUAGE” PROJECTS I N OTHERSPECIES

Dolphins have been chosen for study primarily because of their intelligence. Furthermore, since dolphins’ natural communication system includes a great variety of pulses and tonal sounds and they appear to be facile imitators of certain acoustic variations, some experimenters have attempted to teach dolphins human vocal language. In an excellent review, Herman (1980) notes that dolphins were unsuccessful in producing such sounds, although receptive vocabulary training has been successful. Herman states that recent work (which is still underway) indicates that dolphins can respond to simple “language” components composed of gestures or computer-generated, whistle-like sounds and can likewise correctly respond to combinations of two or three components. In addition, dolphins have also been studied following procedures based on Premack’s work using plastic chips to represent objects and actions (Langbauer, 1982). These dolphins appear capable of receptive and productive use of single components, and can respond appropriately to two-unit combinations but cannot produce such combinations reliably. Projects are also underway with “talking” birds such as the African gray parrot (Pepperberg, 1981), and the mynah bird (Turney, 1980), the latter work deriving from previous research by Ginsberg (1963). One hopes these researchers understand the pitfalls of the ape language projects. One also hopes that these projects will not engender a rush towards investigating quantities of species of animals in order to teach them “language. ” At least one hopes for a delay until some very well designed and carefully executed experimental programs can be utilized by researchers with a good understanding of the species

228

CAROLYN A . RISTAU A N D DONALD ROBBINS

they are working with, including knowledge of the species’ natural communication system, and of the problems and accomplishments of the ape language research. The investigators should as well have an appreciation of experimental psychological research design and some sensitivity to the nature of human Ianguage and developmental aspects of children’s language acquisition. That would be no minor achievement. E.

EVOLUTIONARY FORCES THATMAY HAVELEDTO LINGUISTIC-LIKE ABILITIES IN THE APESOR EARLYMAN

This issue, though a fascinating one, is generally beyond the scope of this paper. Gestural theories (Hewes, 1977) and vocal-onomatopoetic theories have been offered for the origin of human language. A variety of evolutionary forces have been suggested, such as tool use, or, perhaps even more plausibly, the directing force has been seen in the emerging complex social relationships and group structures of early anthropoids or protohomonids. [See Crook (1980) and Humphrey (1980) for discussions of the evolutionary development of intellect in social animals.] One might also suggest that language may have arisen to communicate information about location and other data in spatial maps. Indeed, primates and other species seem to have fairly complex sorts of spatial maps. And the bee dance, a tantalizing possible example of naturally occumng symbolic communication, yields primarily information about location. (Desirability, in terms of the current needs of the hive members, is also indicated.)

F.

IMPLICATIONSO F APE NATURALCOMMUNICATION AND BEHAVIOR FOR IMPROVING INVESTIGATION INTO ARTIFICIAL APE LANGUAGE A N D COGNITION

Observations of chimpanzees in their natural environments suggest that artificial ape languages should be constructed so as to have apes communicate about social relations. This suggestion derives in part from the fact that young apes apparently spend far less time than children do manipulating physical objects in their play, but instead focus on social relationships. Yet much of the ape language work, particularly Premack’s, deals with physical objects and descriptions of them. The Lana project and the recent work of Savage-Rumbaugh, Rumbaugh, and colleagues have dealt largely with names of objects, especially foods and tools to get food, while the signing projects have stressed names of animals, human trainers, and objects in the experimental environment. It is true that Patterson claims the gorilla Koko can use signs for feelings such as sorry. Given, however, the great difficulty we have in determining the nature of a gorilla’s concept of ‘‘sorry” and its relation to a human concept of “sorry” and the cultural expectations of behaviors to accompany “sorry,” it is difficult to

LANGUAGE IN THE GREAT APES

229

evaluate her work. Despite the difficulty, it could be worthwhile to attempt teaching a chimpanzee to associate signs with such mental states as aggressive, frightened, in pain, hungry, thirsty, or wanting to play. In this regard, it is interesting to note that one of the first signs Washoe learned was fickle, which she used to request a playful tickle from a human trainer. It might also be reasonable to attempt to teach the chimpanzee terms designating status, such as boss or alpha, dominant over you, submissive to you, male, female, and so forth. In analyzing the grammar displayed by Washoe, McNeill (1974) suggests that the word order reflects predominantly “addressee-nonaddressee. ” Since such a word order would often be the same as “agent-object,’’ at least in the simple word combinations used by the signing apes, such sequencing by the ape accords well with our structural analysis of “agent-object. ” An example might be an ape signing Ann give milk. Apart from addressee and nonaddressee, McNeill further suggests, word order probably depends on the prominence of the things referred to.He also suggests, as have Terrace et al. ( 1979a), that words were often added to the strings not to convey new information, but to make the message of the string more emphatic and intense. McNeill gives as an example, Please open hurry (McNeill, 1974). Many examples of Nim’s signing are filled with repetitions of a preferred item of food (Terrace et al., 1979a). McNeill likewise recognizes that social hierarchy is important in Washoe’s signing and suggests that it is important for a chimpanzee to sign its intentions clearly, whether they be subservient or aggressive and the degree of such. He notes that Washoe’s signing yields many examples of appeasement and begging to the human trainers. For example, in the signs for please come-gimme, the ASL sign for come gimme, at least in the form used in the Washoe project, is very similar to the natural chimpanzee begging gesture. McNeill (1974, p. 160) further suggests that typical messages of the chimpanzee will probably consist of “entreaties, demands, mollifications and declarations of ownership and location.” How different, we wonder, are these presumed messages from those ethologists have long been suggesting are the messages of animal communication? Some examples from monkey communication may also be relevant. Recall the gelada baboons grunting together in the setting sun; other species of monkeys similarly make soft, complexly varying vocalizations in apparently peaceable social interactions. We could speculate a bit about what these monkeys may be communicating, and try very carefully to introduce signs to apes to signal calmness, pleasure, and so forth. Grooming and being groomed may also be reasonable matters to sign about. It may be of significance that the signs for gimme fickle, perhaps akin to grooming or to chimpanzee play, are easily learned signs. By utilizing our knowledge of primates’ natural behavior, we may begin to formulate an artificial ape language that will indeed serve as a bridge between

230

CAROLYN A . RISTAU A N D DONALD ROBBINS

natural ape communication and cognitive abilities and human language and cognitive abilities.

IX.

IMPLICATIONSFOR HUMANLANGUAGE A N D COGNITIVE DEVELOPMENT

In this section, we will reconsider the various criteria for human language that were offered in Section 111, and suggest which criteria the apes have met in their artificial language productions. We will also consider the deficits of the apes’ productions. Then we will make the assumption, at least for purposes of discussion, that apes do exhibit linguistic-like behavior. We will then examine whether it is by homology or analogy that they exhibit such capacities and consider the evidence for homology. Finally, we will briefly consider the implications of the ape language research for studying the development of children’s communication, both signing and verbal. A.

HAVE “LINGUISTIC” APESMET THE LANGUAGE?

CRITERIA FOR

HUMAN

Since we have at this point reviewed the ape language projects, we might now want to reconsider the application to the chimpanzee’s abilities of the 16 design features proposed by Hockett and amplified by him and Altmann (Hockett, 1958; Hockett and Altmann, 1968; Thorpe, 1974). Only human verbal language is presumed to have all these design features (see Section 111). The features are ( 1 ) vocal auditory channel, (2) broad transmission and directional reception, (3) rapid fading, (4) interchangeability, (5) complete feedback, (6) specialization, (7) semanticity, (8) arbitrariness, (9) discreteness, (10) displacement, ( 1 1) openness, (12) tradition, (13) duality of patterning, (14) prevarication, (15) reflectiveness, and (16) learnability (Thorpe,, 1974). At this point of experimentation, the only characteristics that we can definitely claim are not a characteristic of the chimpanzee’s abilities with artificial language are (1) use of a vocal auditory channel and (2) rapid fading. The first denial is made because the current ape language projects utilize signing or artificial visual lexicons and previous attempts to teach apes to use vocal words were generally failures (e.g., Furness, 1916; C. Hayes, 1951; K. Hayes and Hayes, 1951; Kellogg and Kellogg, 1933; Laidler, 1978). Furthermore, research, though controversial, does suggest the ape supralaryngeal structures will not support the controlled vocalizing needed. Rapid fading is also not a characteristic, since both signs and artificial lexicons can be displayed for a long period. Clear yeses can be given to the presence of specialization (6), discreteness (9), and learnability (16). Interchangeability (5) of the roles of transmitters and re-

LANGUAGE IN THE GREAT APES

23 1

ceivers has been demonstrated between chimpanzees and humans in all the projects and between two chimpanzees in the Savage-Rumbaugh et al. toolexchange paradigm. Ape-to-ape signing has been reported (Fouts, 1974; R. A. Gardner and Gardner, 1978), but details of these interchanges are lacking, so we unfortunately cannot draw any major conclusions from them. Fouts indicates he has observed ape-ape signing, though it is not certain whether this occurred in situations with no cuing from humans. The equivalence of the chimpanzee’s proficiency in both roles is certainly challenged by work such as Terrace’s which emphasizes the high proportion of imitative responses and the low proportion of spontaneous utterances by the chimpanzee acting as transmitter in a humanchimpanzee dialogue (Terrace et a l . , 1979a,b). Complete feedback ( 5 ) certainly occurs for the chimpanzees using artificial lexicons. Signing chimpanzees also receive complete feedback, excpet insofar as their own vantage point differs from the recipient of their signs. Other nonverbal behavior, which probably does not afford complete feedback, accompanies ape “linguistic” behavior just as it does with humans. The importance of these nonverbal cues in making clear, on the one hand, the meaning of the human’s linguistic utterance and, on the other hand, the presumed meaning of the chimpanzee’s signing should not be underestimated. Such cues are the basis of the Sebeok assertions (Umiker-Sebeok and Sebeok, 1980) that the ape language projects reflect little more than the Clever Hans phenomenon. (Further discussion in Section X , A . ) The arbitrariness (8) of some of the ASL signs and some of the chimpanzees’ variants of those signs has been challenged (Savage-Rumbaugh et a f . , 1978b), since some of the signs are said to be icons of the object denoted, while others are very similar to natural chimpanzee gestures used in similar circumstances with presumably similar “meaning. ” As we have previously discussed (Section IV,A), it is difficult to recognize the icon represented by a sign, without already knowing the meaning of the sign. Furthermore, it is clear that a very large number of the ASL signs are neither icons nor chimpanzee natural gestures. Evidence seems to have accumulated for the property of displacement (10). Premack’s work with Sarah, e.g., reveals her abilities to answer questions of the sort ?What color of chocolate when only the plastic chip for chocolate was present and not the candy itself. Evidence also derives from signing dependent upon the memory of an object or event, so long as the signing is not imitative of the teacher’s prior signs. Evidence of prevarication (14) and of reflectiveness (15) is more equivocal, but again seems to be tending towards the positive. Woodruff and Premack’s experiments (1979) in which the chimpanzees lied about hidden food is the best evidence for prevarication or some precursor of it. Results from some of the signing projects, in particular Patterson’s, indicate anecdotal evidence for prevarication. It is, however, difficult to evaluate her line of evidence due to a lack of

232

CAROLYN A . RISTAU A N D DONALD ROBBINS

data about the statistical occurrences of inappropriate or seemingly nonsensical chimpanzee utterances, as well as to some seemingly overgenerous interpretations of chimpanzee abilities. Among the various projects, prevarication in the sense of denial, a false negative, does seem to have occurred; evidence for false positives is much less clear. Reflectiveness ( 15) is the ability to communicate about the communication system itself. Reflectiveness is at least suggested by the chimpanzee’s use of metalinguistic concepts such as name of which seems to have been exhibited by apes in several of the projects. (See Section VI,A for detailed discussion of problems inherent in determining chimpanzee’s meaning of signs such as name of.) Other suggested occurrences derive from tasks such as those in which Premack’s chimpanzee Sarah described the characteristics of the plastic chip representing caramel and also the characteristics of a caramel itself. For the ape languages to be passed on by tradition (12) would require that one chimpanzee spontaneously teach the communication system to others, most probably offspring. Fouts has arranged for Washoe to have an infant chimpanzee in her care and thereby a potential student of signing. Although we have seen no published data on this matter, information from an oral description of the work indicates the infant has done some signing (Fouts rt al., 1979). We do not know if the conditions of observation precluded cuing by observers, nor do we know about the contexts in which signs were used, nor whether they were imitations of preceding signs by Washoe. At least in the design of the experiment, no human should sign in front of him, so any signs used by the infant should have been learned from Washoe. Furthermore, although some traditional skills may be passed on through merely observational learning, as in humans and quite probably when animals pass on traditional migratory routes, it would be of interest to note whether Washoe specifically attempted to teach signing to the young infant in her care. The existence of semanticity (7), openness (1 l ) , and duality of patterning ( 13) remains, in the opinion of many researchers, highly controversial, perhaps especially because linguists (e.g., von Glasersfeld, 1977) specifically mentioned these features as essential conditions for language. Semanticity would seem to be an accomplishment of the chimpanzees; evidence derives from the ability of the chimpanzees to name objects and events of the external world and in particular from the recent Savage-Rumbaugh et al. experiments (1978b, 1980b). Semanticity is, however, a very general term, and useful avenues of research still lie in exploring the development of word meaning with comparisons between the chimpanzees’ achievements and the richness of a word in human language. Openness in the sense of novel word combinations may have been exhibited, if one accepts anecdotes such as chimpanzees signing cry hurt jbod for radish (Fouts, 1974). Because these anecdotes are open to criticisms previously made, it is unclear if one can assume the criteria for openness have been met. Openness

LANGUAGE IN THE GREAT APES

233

through novel grammatical structure has not been demonstrated, yet the length of utterances made by most signing apes and by Premack’s chimpanzee Sarah affords little opportunity for creative use of grammar. Determining what is novel is not simple either, for we apply the term “novel” to appropriate new utterances and “error” to utterances that seem inappropriate to us, but might at least in some instances, be attempts by the ape to invent a new “word.” One might possibly include as examples of some degree of “openness” concepts (e.g., on, in, under) in transfer tests involving “novel” objects and locations. By “novel” we now simply mean items not involved in training use of prepositions. Experiments do indicate that some chimpanzees have such abilities (e.g., an oral report by R. S . Fouts, W . Chown, G . Kimball, and J . Couch, 1976, as cited in Fouts, 1977; Premack, 1976). Concerns with these experiments center on whether apes can use the prepositions spontaneously, but accurately, and in a large variety of contexts other than the experimental testing of subsets of prepositions. Duality of patterning can have at least two meanings. It can refer to the combination of meaningless elements such as phonemes in human spoken language, to form meaningful patterns, namely words. If taken to mean, in signing, a combination of cheremes, i.e., finger positions and hand placements and actions needed to produce the pattern that is the sign for a word, then duality of patterning has been exhibited by the apes. Most usually, however, (e.g., Healy, 1973) the ape language studies are considered to use a meaningful element, the word, as the basic unit. (Discussion in Section 111,B.)If duality of patterning is also taken to refer to structural rules, such as those in grammar, then there is no convincing evidence that ape “artificial language” has this property. lmportant characteristics of human language are lacking among the design features. Human language is intentional communication (Bennett, 1976; Grice, 1967). A proposition within that language is considered to be an attempt by the communicator to change the belief or knowledge state of the other, not merely to change the other’s behavior so as to administer rewards to the communicator. Clear demonstrations that the apes distinguish between “knowing how” and “knowing that” are still needed. The use of rewards by experimenters to their communicating apes often muddies the interpretation we are able to make of this distinction in the apes’ productions. Language must be used in a variety of contexts, and the user must have the capacity to emit, i.e., to show spontaneous use of a word in an appropriate situation (Lenneberg, 1971). Chomsky reminds us of other uses of language “telling a story, requesting information merely to enhance understanding, expressing an opinion or a wish (as distinct from an instrumental request), monologue, casual conversation” (Chomsky, 1979, p. 39). There is no evidence to indicate that apes use language in these ways, and it is also unclear exactly how to gather convincing evidence. There are some suggestive anecdotes: Is

234

CAROLYN A . RISTAU A N D D O N A L D ROBBINS

Washoe or Koko, “browsing” through a picture book, signing the names of the objects to herself, engaging in a monologue? (R. A. Gardner and Gardner, 1974; Patterson, 1978a). In fact, Washoe also corrected herself during this activity (B. T. Gardner and Gardner, 1974). She also signed to herself when playing alone, sitting up in a tree, or lolling in bed, but often stopped when she noticed she was being observed. Another anecdote-are not the signs made by Lyn Miles and a chimpanzee before or after an actual experimental session, when the situation is very relaxed, akin to causal conversation? (Miles, 1978). Finally there is the richness of meaning in human words and poetry, in metaphor, in words and language as they are a part of culture and reflect that culture. The apes’ utterances clearly differ from human language, yet an understanding of the nature of the differences and similarities can better reveal the complexities of our human language and point to possible precursors in apes’ natural communication and in their cognitive abilities. B.

BEHAVIORS ANALOGOUS ARET H E APES’“LINGUISTIC-LIKE” HOMOLOGOUS TO MAN’S?

OR

Let us, for the point of argument, suppose that apes exhibit language-like behavior. By making this supposition, we are sidestepping the difficulties in determining exactly when behavior is linguistic. We may still question whether such language-like behavior is the consequence of neuraucognitive systems similar to those underlying human language behavior and due to a common origin, or whether the similarities are superficial structural ones dependent largely on training procedures (Malmi, 1976). Our discussion shall focus on this issue. “In biological terms, the issue is one of homology vs. analogy (Mayr, 1969; Simpson, 1961)” (Malmi, 1976, 1980, p. 192). Yet “traits of common origin may diverge markedly over a short time, whereas traits of different origins may converge so as to appear virtually identica1:‘homologous structures may be extensively similar or dissimilar. Therefore, similarity is to be considered something quite apart from considerations of homology’ (Campbell and Hodos, 1970, p. 101)” (Malmi, 1976, 1980). The great genetic similarity between man and chimpanzee has often been taken as evidence for behavioral homology between the two species (King and Wilson, 1975). Even great genetic similarity does not imply behavioralkognitive similarity, nor does the absence of such genetic similarity necessarily imply the lack of intellectual comparability. For example, the phylogenetically distant man and dolphin might share more intellectual abilities then do man and ape. In short, genetic similarity is an overly simple and inadequate explanation. (This issue is discussed further in Ristau and Robbins, 1979.) In the case of the chimpanzee, it may be that some of the skills underlying the apes’ performance in the “language” projects may reflect capacities common to

LANGUAGE IN THE GREAT APES

235

the extinct ancestor of today’s man and today’s chimpanzee, and that such abilities might be particularly relevant to understanding human language behavior. This is not a simple comparison to make for we are studying contemporary great apes (Passingham, 1979). Inasmuch as the current hominids diverged some 20 million years ago from the great apes (Poirier, 1973) and that modem apes have been evolving during this time (as have we), modem apes are not representative of the past. This is not to argue that we may not find evidence of mental continuity, but to alert us to the fact that precursors to language that may have existed 20 million years ago may have been eliminated in great apes, while at the same time the precursors to language may have been selectively enhanced in hominids, particularly during the last 0.5 million years. With this warning in mind, if we still wish to search for evolutionary precursors to language in the ape, we must ask how chimpanzees naturally employ such skills in their freeranging behavior, and attempt as well to judge the homologous nature of those skills. Several kinds of data are considered as evidence for homology (Malmi, 1976, 1980). ( 1 ) A fossil record which of course can exist only rarely for behavior as in animal tracks. (2) Traits that share a multiplicity of similarities, especially if such similarities are not functionally related to one another. A roster of design features would be an example of such an approach. However, Malmi contends that the design features exhibited by the “linguistic” apes are related to each other, because some features depend on others and nearly all the features result from training. (3) Similar behaviors that share a similar ontogenesis. (4) Minuteness of similarity, for evolution is more conservative in micro- than in macrostructure. The bat is morphologically convergent with birds because of its aerial niche, the porpoise with fish because of its aquatic habitat. In each case, mammalian affinities are revealed by examination of blood, skeleton, eyes, brains, and other features less subject to environmental features than fins, wings, and snouts. Thus, traits should be examined for minuteness of similariries (Malmi, 1976, 1980, p. 194).

Malmi further suggests that in dealing with behavioral comparisons one must examine biological substrates underlying performance. This is because behavior is much more plastic than its physiological basis. Among mammals, and especially the higher primates, learning plays an important role and behavior is very dependent on experience. Humans take advantage of this in training animals, so that apes have been taught to ride bicycles, have tea parties, smoke cigarettes, drink from a glass. and use “language.” All but the last can have played no role in chimpanzee evolution, so that success in such activities stems from modification of skills that are present in the chimpanzee repertoire for other purposes (Malmi, 1976, 1980, p. 195).

These thoughts imply that if we expect to find evidence of homologous development between chimpanzees’ and children’s ‘linguistic skills it would be

236

CAROLYN A . RISTAU A N D DONALD ROBBINS

reasonable to look at the minuteness of similarity between those skills. We might guess that some areas of similarities could be turn-taking, the nature of grammatical structure, the process of word acquisition, and the sequence of skill development, although at least the last two skills are dependent upon the specific teaching techniques used. The search for biological substrates that might indicate homology is more perplexing. Are the ape skills sufficiently complex so as to be comparable to human linguistic skills that are under the control of the left hemisphere; are there comparable tendencies for control by the left hemisphere for some of apes’ skills, perhaps in other cognitive areas’? We seem to have a sensitive or critical period for language learning. Work done with birds, which have a sensitive period for song learning that is hormonally dependent and apparently related to puberty (Nottebohm, 1969), might lead us to speculate that the end of the human “sensitive” period is related to puberty as well. We might ask if a similar period also exists for chimpanzees either in terms of their natural communication systems or for their artificial language learning. The latter is complicated by the fact that older chimpanzees are simply not so pliant as young ones about participating in any kind of experiment. In short, to assume that any ape finguistic-like accomplishments are homologous to man’s, rather than analogous, is not without difficulties. We do not mean to imply that the approach to search in apes for possible precursors to man’s language is a mistaken one; only that one must conduct that search with care and caution. C.

FOR HUMAN CHILDREN’S COMMUNICATION IMPLICATIONS

Comparisons of the ape artificial language use to language use by human children are extraordinarily difficult to make although some have been attempted (e.g., McNeill, 1974; Bronowski and Bellugi, 1970; Brown, 1970). For one, the signing apes should properly be compared to signing children, not to verbal children. There are almost no data on the corpus or development of signs in human children, although data on these issues are beginning to accumulate (Goldin-Meadow and Feldman, 1977; Klima and Bellugi, 1972; Klima et a f . , 1979). If, in the absence of such data, we look to studies of the development of verbal abilities in human children, we are again faced with a lack of data to use in making the comparisons, although the lack is by no means so severe. We do have data on the corpus of some groups of small children (e.g., Nelson, 1973). To begin to understand the “universal, ” characteristically human aspects of language development, we need to have corpora from more disparate groups of children, in particular, those with non-Western first languages. We know of no quantitative data pertaining to the occurrence of novel productions such as word combinations, which we and other critics have sought to request from the ape language researchers. Children do, of course, create novel word combinations in forming new sentences. We wonder, though, how often

LANGUAGE I N THE GREAT APES

237

children, if not provoked by What this questions, create novel word combinations, such as cry hurtfood for radish. They do engage in playful, nonsensical babbling. However, cursory, personal observations suggests such novel combinations are very rare. Other issues arise from findings of the ape language projects or criticisms of them. Is the functional use/labeling distinction helpful in understanding how yound children first learn words? The ape language researchers may offer “gratuitous interpretations” (Ristau and Robbins, 1979) of the apes’ productions, but then so do enthusiastic parents. Such interpretations may not provide useful scientific descriptions, but may be an important part of the language acquisition process, bridging the gap between children’s attempts and conventional forms. As we and others (e.g.. Bindra, 1981) have suggested, nonverbal cuing and the Clever Hans phenomenon are present in human adult and child “linguistic” conversation. “Clever Hans” must also doubtlessly affect researchers’ interpretations of child linguistic competence of an autistic child shunning human contacts and forceably taught a language for a few hours each day (presumably analogous to the apes typically reared away from conspecifics). We need to know yet more about differences between an adult’s and child’s language, and we need to be able to compare proficiencies of an ape and of a child under comparable conditions. Most of these questions are potentially answerable but have not been as yet (Bindra, 1981). It is as an impetus for research into children’s language acquisition, that a major contribution of the ape language projects may lie.

x.

PROBLEMS RAISED BY THE LANGUAGE RESEARCH AND SUGGESTIONS FOR FUTURE RESEARCH

The ape language research has raised a number of issues. (A) The Clever Hans phenomenon and the growing awareness of its impact on the ape language research and other areas of scientific endeavor and personal interactions; (B) Has the response to ape language research been “overly” critical? An associated issue is the problem faced by researchers of changing criteria for language acquisition. Finally, we will consider (C) suggestions for future research into ape language, cognition, and mental states, emphasizing the importance of applying understandings gathered from natural animal behaviorkommunication to the design of the most efficacious paradigms.

A.

THE CLEVER HANSPHENOMENON

The role of the Clever Hans phenomenon in science in general, and ape language research in particular, has been emphasized strongly by Umiker-Sebeok and Sebeok (1980) and was the focus of a recent meeting of the New York

238

CAROLYN A . RISTAU AND DONALD ROBBINS

Academy of Sciences (May 1980) (Sebeok and Rosenthal, 1981). In that session Sebeok tried to discredit most ape language researchers, because, in his view, they ignore the prevalence of the Clever Hans phenomenon and have results explicable by inadvertent cuing and expectancy effects. Sebeok’s arguments derive from the classic case of the clever horse, Hans, who could apparently perform a variety of mental feats such as counting and arithmetic. A psychologist, Hungst, noted that the horse’s performance disintegrated, if, e.g., the examiner did not know the answer to a question proposed to Hans. Pfungst determined that nonverbal cues, such as slight relaxations and inclinations of the body, provided the information to Hans to stop tapping his hoof, i.e., to stop “counting. ” It is cues such as these that the Sebeoks seem to consider sufficient to cue the signing and other “linguistic ” chimpanzees. In delineating their views, Umiker-Sebeok and Sebeok consider the kinds of relationships man may have with an animal, in particular, when man is training an animal. They distinguish between “apprentissage, ” loosely rendered as “scientific training” presumably with little emotional involvement between man and animal and “dressage” or circus training, with rich emotional involvement and much concomitant interpersonal signaling (Hediger, 1968, reported in UmikerSebeok and Sebeok, 1980). Man-ape relationships in the ape language projects are all considered to lie somewhere between apprentissage and dressage. For example, all the projects have noted the usefulness of the congenial social atmosphere on the apes’ performances and disruptive effects from the presence of observers who are strangers (e.g., even in the structured laboratory situation used by Premack, 1976). Furthermore, the researchers in the ape language projects are subject to expectancy effects, much like those reported from a long series of experiments by Rosenthal and other social scientists (Rosenthal and Rubin, 1978). Such expectancies, state Umiker-Sebeok and Sebeok, will lead the researchers to observe and/or record ape behaviors inaccurately, to overinterpret ape performances, or unintentionally to modify the ape’s behavior in the direction of the results desired (Umiker-Sebeok and Sebeok, 1980). To be sure there are observational and recording inaccuracies, for instance, as indicated by Lyn Miles’ measurement of intermethod agreement (see Section IV,B). Some interobserver agreement measures used by Terrace entail recording signs after discussion of those signs by observers. Inaccuracies also often result from omissions of signs, when records are dependent on written or verbal transcriptions rather than filming or videotape. Overinterpretation does occur as we have discussed. For the most part, however, Sebeok asserts rather than proves effects due to Clever Hans errors. Inadvertent cuing through nonverbal signals can reasonably explain how a clever horse can know when to inhibit tapping his hoof, but contributes little towards understanding how an ape knows which particular

LANGUAGE I N T H E GREAT APES

239

finger and hand configuration to use in making a sign. More recent experimental work by Premack and Woodruff (1 978a; Woodruff and Premack, 198 1) and by Savage-Rumbaugh and Rumbaugh and their associates (e.g., 1978b, 1980b) have utilized, though to varying degrees, much stricter “blind” experimental designs in which the experimenter is out of the room during testing. In some experimental paradigms some nonverbal cuing still remains possible. We also wonder, though, if the chimpanzees are clever enough to discern such subtle cues, might they also not be capable of doing the experiment? All in all the effects of Clever Hans errors are powerful and the impetus to improving experimental design is a laudable one. It seems, though, that attributing the results of the ape language research to Clever Hans errors is unwarranted.

B . OVERLY CRITICAL ANALYSIS OF THE APE LANGUAGE R E S E A RC ti‘?

Some have claimed that critics of the ape language projects hunt for flaws in the research rather than engage in a thoughtful, considered evaluation of the work. Once such flaws are remedied, critics search for yet more. It can be argued that all scientific research has uncontrolled variables and faults; the problem always is, based upon our understanding of the field and training as scientists, to determine which uncontrolled variables should be controlled in further experimentation. Scientists generally err on the conservative side. Even so, sometimes we guess wrong. The problem is compounded in the ape language research, primarily due to the difficulty of replicating any studies with great apes, especially the extremely costly and time consuming ape language work. Since the possibility is almost nonexistent of collecting comparable data in one’s own laboratory, scientists must ask for as clear a reporting of methodology and as complete a presentation of data as possible. Transcripts of apes’ and humans’ signing together and unedited research films or videotapes of the apes’ performance, particularly the signing apes, should be easily and inexpensively available to the scientific community. Scientists can then begin to analyze and interpret the data from their own viewpoints, possibly a very useful addition to this field. Terrace has begun to analyze the small amount of film from the Gardners’ chimpanzee Washoe that was available to him and also what was actually a minute amount of filmed performance by Koko; he came to conclusions different from those of the respective researchers (Terrace et al., 1979a, 1981). (Some discussion of this is in Section IV,F.) Thompson and Church (1980) have developed a computer simulation model and applied it to unpublished data from protocols of Lana, suggesting interpretations different from those originally offered by the researchers (Rumbaugh, 1977; Gill, 1977). (For further discussion, see Section V,A.) More such analyses are needed.

240

CAROLYN A. RISTAU A N D DONALD ROBBINS

Also in lieu of the possibility of most scientists replicating the ape language work, critics must continue to indicate flaws in methods and interpretations, so the ape language researchers can further develop their investigations. Another issue raised about criticisms of the ape language research is that there are changing criteria for the acquisition of language. When apes seem to satisfy previously stated criteria, new and more demanding ones are erected. This is a real problem and yet understandable. As we have frequently mentioned, no one has yet proposed a satisfactory definition of language. In the absence of our knowledge, linguists and scientists observe human language and try to abstract the relevant and critical features. Then when presented with a research design that purports to incorporate these criteria, and with results that often seem at variance with our every day and intuitive understanding of language, we may be resistant to accept that research. Often our intuitive appreciation leads us to realize something is amiss, before our verbal representations and analytic powers can provide intellectually and scientifically satisfying explanations. Yes, there are changing criteria for the acquisition of language and this continual refinement is necessary until we better understand the nature of language and cognition.

C.

SUGGESTIONS FOR FUTURERESEARCH

In this section, we will reiterate suggestions already made and offer additional approaches. I.

What Is Acceptable Scientific Evidence?

Research into apes’ “linguistic” and cognitive abilities and their mental states has certainly raised this formidable issue, which can only be noted here and not discussed in any detail. Attributes of mental states and of consciousness are made by us in our everyday lives and used to pragmatic advantage (Section VII). Yet scientific study of these matters is extraordinarily difficult. It may be a reasonable approach to attempt to analyze the kinds of evidence we use in our everyday judgments, as some philosophers have been doing for a long time, and to try then to apply these to approaches useful in scientific inquiry. Another issue to be raised is the need for careful reporting of anecdotes within the scientific literature. It is indeed unfortunate that, at the present, the popular magazines such as National Geographic or Time must be the source for many of those anecdotes from the ape language projects (e.g., Patterson, 1978c) that can help reveal proclivities of the apes’ mentality and perhaps suggest new avenues of experimentation. Experimentation, of course, needs to be done, but pioneers such as Piaget made great advances into revelations of the child’s cognitions by careful reporting of anecdotes and by devising experiments derived from the anecdotes. We might be able to do as well in our studies of animal mentality.

LANGUAGE IN THE GREAT APES

24 1

While encouraging the careful use of anecdotes, we must again strongly emphasize the need for filmed or videotaped records, probably taken simultaneously from several vantage points. 2.

Further Investigations into Meaning

If we hope to understand the continuity between human and animal mind, we should strive to investigate the apes’ meaning in their use of signing or artificial lexicons. Several approaches seem possible. a . Investigate Errors. We should investigate the errors made by apes, especially when apes first learn new signs and generalize the signs to other objects and events. Recall how Washoe generalized the sign forflower first to odors, not to other flowers, until she was specifically trained to do so by the experimenters. Apes’ errors in other experiments, such as those by Premack investigating Sarah’s concepts of objects, could also be revealing. h. Gather Sraristicul Dutu. As we have often suggested, researchers should gather statistical information on the apes ’ production of novel word combinations that appear to be used either to label an item for which they have no sign, describe and event, or possibly indicate some kind of grammatical structure. Such data will also yield information relevant to meaning. c. Compure Concept-Learning Difficulties. We should compare the apes’ relative difficulties in learning different kinds of concepts so as possibly to reveal their natural proclivities. We could continue developing some of the approaches used by Premack ( 1976). Some important changes in the paradigms include, e.g., giving the apes more alternatives to choose among, indicating only first trial data in published reports when appropriate to do so, and varying the concepts tested in any one session so as to have a clearer sense of the apes’ grasp of the concepts. d . Demonstrate Understanding of Temporal Concepts. Among the concepts of particular interest which have not yet been adequately demonstrated is the ape’s understanding of time or temporal concepts. Following is one possible paradigm. An animal’s episodic memory (memory for a specific event) could be revealed by the use of two indices for an event. One of these indices is used by the experimenter to question the ape about the event and the other index is used by the ape to corroborate with the experimenter the event referred to. A label, such as a plastic chip or sign, could indicate whether the event in question occurred in the past, is currently underway, or will occur in the future. One index could be one of a sequence of colored lights, while the other index might be the kind of reward received by the animal. Zaidel (1978) has used somewhat similar nonverbal paradigms in studying cognitive capacities of the right hemisphere of split-brain patients. e . “Functional” vs “Labeling” Procedures. The “functional” vs “labeling” procedures for learning “words” investigated by Savage-Rumbaugh ef al.

242

CAROLYN A. RISTAU A N D DONALD ROBBINS

(1978b) should be investigated more thoroughly, so as better to understand the variables affecting the great differences in rates of learning between those two procedures. f. Analyze f o r Meaning. We should muse about how to analyze for meaning. We can look to the ethologists’ methods in animal communication, the anthropologists’ and tourists’ ways of understanding a language other than their own, and the developmental psychologists’ and parents’ manners of understanding the meaning in communications from their developing children. We should consider the contributions of philosophers and encourage their assistance in these difficult areas. Will similar methods work in each field? Can we learn more of how to study meaning in one of these areas by comparing the methods and results from another area? 3 . Does Language or at Least the Use of Labels Aid Thought or Learning:‘

It will of course be easier to determine if labels aid the rate of learning; if so, one may next attempt to devise experiments to reveal whether they aid the animal’s thought. The effect of labeling on acquisition of learning has been studied in children (e.g., Ristau, 1965). One might do a classic discrimination task with and without labels, in which each set of objects or figures to be discriminated could be preceded or accompanied by one of two plastic chips or some other labeling device. This task bears some similarity to the categorical sorting task of Savage-Rumbaugh et al. (1980b), but theirs presumably dealt with categories already familiar to the chimpanzees and did not compare rates of learning categories with and without labels. 4.

W e Could Look to Natural Animal CommunicationlBehavior for Suggestions

The animals’ natural way of life may well suggest which concepts we should attempt to begin to teach the apes. As we have previously suggested (Section VIII,C), words describing social relationships, such as boss, male, femule, or juvenile might be of interest as would words to describe deference, neutrality, and dominance. We might also wish to search in the apes’ natural communication system for signals that are not highly arousing and thereby disruptive to the research, and try then to interpret the meaning of those signals. Accompanied by our good guesses, we can next attempt to teach signs which “map” those meanings. In this way we may possibly bridge the transition from ape calls and gestures (traditionally assumed to be largely reflexive) to more clearly voluntarily used signs or artificial representations. Such may be an important way that children first learn to use language, and, we speculate, may also have been involved in the acquisition of language by early man. As an example from a child

LANGUAGE IN THE GREAT APES

243

developing language, No is first learned accompanied by negative affect and nonverbal signals of refusal or denial. The acoustic characteristics of No likewise mirror the vocalizations that occur naturally in the preverbal child in those contexts in which No is first used. Gradually the word No becomes dissociated from much of the affect and nonverbal signaling (Ristau, 1974). While observing the natural behavior of animals in their environment, we might also search for the “schema” that may be utilized by young apes in their play, perhaps similar to those of 1- to 3-year-old children. In play, at different ages, children perform certain acts repetitiously. For example, they can spend long times sorting items into categories that are obvious and not-so-obvious. At other times, children repeatedly put items into containers, remove them, and put them back in again and again. One could speculate that children are developing the cognitive foundations for categorical analyses, for the preposition “in” and related grammatical structures. Do young apes exhibit comparable behaviors?

5 . Mental States Another direction for future research is to inquire into the existence of mental states, such as belief, and the attribution of mental states by the ape. Bennett (1978) and others, developing their ideas from those of the philosopher Lewis (1 969), have suggested devising coordination problems (see discussion in Section VII). Specific workable methodologies have not been suggested for such problems nor are they particularly easy to devise. There are, however, basic parallels between the coordination problem and cooperative hunting that has been observed in chimpanzees. In such hunting one animal must base his behavior on the actual or expected behavior of the other. Roles differ in hunting such as those of the enticer of prey and the grabber. If chimpanzees do switch roles, we might speculate that the chimpanzees must have some understanding of the role of the other, and with that possibly some notion of “other.” If a notion of “other” is possible, so, we might guess, is a notion of “self.” 6 . Prevarication Experiments

Prevarication experiments are particularly important, because they help to reveal the distinction between “knowing how” and “knowing that. ” That distinction is often used to denigrate the accomplishments of animals within their natural communication systems (e.g., honeybee dance) (Premack and Woodruff, 1978b) and of the apes within their artificial language systems. Similarly, prevarication experiments, viewed from an intentional analysis, can reveal levels of mental complexity (Dennett, 1978b). a . Sorting Tasks. Sorting tasks might be done with the ape given choices of videotape sequences to be labeled as “truth” or “falsehood.” Premack and Woodruff (1978b) have indicated that experiments somewhat similar to this were underway, but we are not aware of the outcome of this work.

244

CAROLYN A. RISTAU AND DONALD ROBBINS

b. Complex Lying. We might ask whether an ape can do more complex lying which entails an understanding of the role of the “other” (see discussion in Section VII). Again one could present a videotape sequence to the ape, and then let him choose among two or more alternative sequences to let him determine “who did it” or (revealing his understanding of general capabilities or roles) who could have “done it.” For example, one could ask the ape who could have broken the vase: the person walking near the table with the vase on it or a bird locked up in a cage? The ape would not have been shown the sequence in which the vase was being broken, only sequences containing the intact vase and then later the broken vase. Premack and his associates appear to have been thinking along similar lines when they proposed an experiment in which the ape saw humans of different capabilities, namely an adult and a child, given tasks or puzzles of varying difficulties. It was hoped the ape would predict that the child was unable to do more complicated problems. Although the results did not bear out the experimenters’ hypotheses, the approach is an intriguing and important one.One must be wary, as we have noted, of problems in interpretations.

7 . Consciousness or Self-Description The elusive subject of consciousness or at least of self-description, might begin to be investigated by giving an ape a personality and preference type of questionnaire in which he could reveal information about himself. A plastic chip or some other artificial representation could be used to represent the ape himself, while questions could be straightforward queries of preferences or ‘*if-then” propositions. An instance of the latter might be, lf Jane hits Sarah, then ? Such questions, of course, would require “mapping” states such Sarah as “like, * ’ “prefer,” andvarious moods-no easy task,perhaps an almost impossible one. More sophisticated “if-then’’ questions might be used to reveal prominence of events or objects in the environment. 8 . General Approaches

General approaches that may be fruitful have been sugges!ed by Griffin (1978) and particularly noted by Gould (1978). These are (a) Situations for which an animal is not prepared by evolution can reveal the plasticity and versatility of the animal’s behavior and the alternatives that lie within the realm of the animal’s cognitive abilities; and (b) by entering into a dialogue with the animal in his own natural communication system, we can hope better to understand the meaning of his utterances, his intentions, and the contextural determinations of his communication and behaviors. Some of this is fairly easily done in ”simple” systems such as the flickerings of a firefly (Lloyd, 1977), but would be far more difficult in animals such as primates. Should anthropologists, developmental psychologists, ethologists, philosophers, and perhaps even linguists join in efforts to explore the larger issue of

LANGUAGE IN T H E GREAT APES

245

mental continuity, of which the ape language experiments are a part. we can hope for innovative approaches in a fascinating field.

XI.

CONCLUDIN STATEMENTS G

The ape language projects are perhaps best considered in their relation to the issue of mental continuity between man and animal. As such, they may reveal a potential bridge between ape abilities that may be precursors to man's linguistic and cognitive abilities. Communicative media utilized i n the projects are hand signing, similar to ASL, and artificial visual lexicons. The signing projects, pioneered by the Gardners, have the advantage that the full repertoire of vocabulary items is available to the apes at all times. Therefore, the apes have more opportunity for creative productions to describe and name objects and events in the environment and, potentially, to create grammatically novel sentences. This is an especial advantage when the ape is free to communicate spontaneously in a variety of contexts lrldoors and outdoors. The signing projects suffer the disadvantage of great difficulty i n recording apes' productions, particularly when the apes are in a nonstructured environment and outside of a regular training session. Terrace and his associates, investigators i n the Nim project, have used videotape most extensively in gathering data. They are, to date, the only ones to have done such laborious and detailed analysis of videotapes; yet even their work encompasses only 3.5 hr of such tape. The Nim project has apparently laid greatest emphasis on drills. It has also been beset by the problem of having a large number of trainers. Upon close inspection, however, it is clear that other investigators, in particular, the Gardners, are not at all free from this difficulty. This matter is of consequence, because a social bond is presumed to be very important in the early language learning of children and presumably in general learning of chimpanzees. The main conclusions reached by the Nim project are that his productions and presumably, those of other apes, are not linguistic, primarily because of a lack of grammar. The apes' use of signing is unlike humans', insofar as the utterance length is very small, utterances become highly repetitive when the length increases, are imitative of prior signing by humans, and, very importantly, apparently lack spontaneity. Critics have claimed that at least some of these characteristics, such as the imitation and extreme lack of spontaneity Terrace and colleagues find in Nim, may be due to the emphasis on drills. Patterson has conducted studies with a gorilla, and has done videotaping which has not yet been subjected to extensive data analysis. It is still difficult to evaluate many of her findings, for there is much anecdotal evidence often presented only in popular journals and some interpretations of the gorilla Koko's abilities seem overgenerous.

246

CAROLYN A . RISTAU A N D DONALD ROBBINS

Other signing projects include those by Fouts, who has attempted to have chimpanzees sign to each other rather than to humans, as is typically the case in the other projects; it is still difficult to determine in these interchimpanzee signing bouts how much cuing is being done by the people around them. Miles is working with signing in an orangutan but there is no published work yet available. The projects utilizing artificial lexicons have the advantage over the signing projects of an ease in recording data, and of greater control over the “linguistic” experiences provided to the apes, but they suffer greatly from a loss in the richness and versatility potentially available to the signing apes. (At least a larger vocabulary could be made available to the apes who use artificial lexicons merely by increasing the number of active keys or artificial signs available to the ape at any one time.) Such a potentiality has not yet been realized in the signing apes, quite possibly due to the shortness of utterances. (Mean length of utterance of most of the signing apes was between 2 and 3.) Lana, the chimpanzee in the original Rumbaugh work, used computer lexigrams and could, according to the rules of the artificial language Yerkish, make rather long strings of “words” to request various rewards. Yet, as the experimenters themselves later noted, Lana, especially when she grew proficient at producing long strings of lexigrams, did not seem to know the meaning of the individual items. No detailed analysis of Lana’s productions has been published, but the computer could record all communications to and from the chimpanzees, and was thereby potentially able to record and analyze an ape’s entire “linguistic” history. Savage-Rumbaugh et al. are beginning to explore the meaning of a word to an ape and the development of that meaning in such tasks as tool use and categorical sorting. Investigations into meaning are an important future direction for research. Premack has sought to investigate cognitive abilities of apes, some of which he notes may be precursors to language abilities. Some of his earlier experiments are related to concept-identification tasks done in the laboratories of experimental psychologists. These experiments have been criticized as needing additional alternatives for the ape, and more explicit presentations of methods and results including, in particular, presentation of first trial data; interpretations of the apes’ abilities in such tasks have been cited as problem solving rather than linguistic (Terrace, 1979a). More recent work by Premack and Woodruff has sought to investigate intriguing issues of intentionality, attribution of mental states, and prevarication using paradigms involving videotaped presentation to the ape of problems faced by another individual. The approaches are important ones for future research, although, at present, the particular tasks and videotaped sequences are often too tortuous, and explanations simpler than those offered by Premack and his associates often seem to suffice.

LANGUAGE IN THE GREAT APES

247

The conclusions that can be drawn are that signs and items in the artificial lexicons do, on some occasions and for some of the apes, appear to represent objects or events in the real world. The apes do sometimes use signs, chips, or lexigrams as symbols; although precisely how similar their use of symbols is to man’s is yet to be determined. In the least, apes appear able to use lexical items to refer to objects not immediately present; i.e., apes sometimes exhibit displacement. There is, however, no convincing evidence that apes’ utterances are grammatical. They tend to repeat items, as if for emphasis; their utterance length is typically very short. Compared to young human children, apes acquire signs much more slowly and sequence them within an utterance very differently. It is difficult to conclude that the productions of “linguistic-like’’ apes bear more than a rudimentary similarity to human language. Indeed, we would probably find it difficult to describe young children’s productions as linguistic, if we did not know these children will almost invariably grow up into language-using adults. However, the ape language projects do indicate cognitive abilities and possible mental states of the apes. Drawing from the results of field studies about natural ape abilities, communication and social relations could improve the direction of the ape language research. Furthermore, as shown by some recent work, more careful exploration into issues of meaning and representation may help to reveal more precisely the differences in achievements of man and ape and may suggest possible avenues to investigate in language learning by human children.

Acknowledgments We would like to thank our colleagues for helpful discussions and, in particular, Colin G. Beer, C. Granier-Deferre. Donald R. Griffin, and Robert Seyfarth for constructive comments on earlier versions of this manuscript and Daniel C. Dennett, Sue Savage-Rumbaugh, and Herbert Terrace for helpful comments on selected portions of this paper presented at the 1981 Dahlem Conference on “Animal Mind-Human Mind. The responsibility for any shortcomings of the paper is, however, the authors’, and not theirs. We are grateful to R. Fouts, the Gardners, F. Patterson, H. Terrace, E. S. Savage-Rumbaugh, and D. M. Rumbaugh for furnishing us with reprints of their papers, to H. Terrace for allowing us to borrow a videotape of training sessions with Nim, and to the editors for their patience. We acknowledge the financial support of the Harry Frank Guggenheim Foundation. We thank Rosanne Kelly for her painstaking preparation of the manuscript. Finally, we would like to express our appreciation of being part of an academic scientific community that espouses acceptance of criticism in hopes of furthering scientific understanding. ”

References Baker, C. 1975. Regulators and turn-taking in American Sign Language discourse. In “On the Other Hand: New Perspectives on American Sign Language” (L.Friedman,ed.), pp. 215-236. Academic Press, New York.

248

CAROLYN A. RISTAU AND DONALD ROBBINS

Barkow, J. H . 1976. Attention structure and the evolution of human psychological characteristics. I n “The Social Structure of Attention’’ (M. R. A. Chance and R. R. Larsen, eds.), pp. 203-219. Wiley, New York. Bates, E., Camaioni, L., and Volterra, V . 1975. The acquisition of performatives prior to speech. Merrill-Palmer Q . 21, 205-226. Beck, B. B. 1980. “Animal Tool Behavior.” Garland, New York. Bellugi, U., and Klima, E. S. 1976. Two faces of sign: Iconic and abstract. Ann. N . Y . Acad. Sci. 240, 514-538. Bennett, J . 1976. “Linguistic Behavior.” Cambridge Univ. Press, London and New York. Bennett, J . 1978. Some remarks about concepts. Cognition and consciousness in nonhuman species: Open peer commentary. B e h v . Brain Sci. 4, 557-560. Bemstein, I. S. 1978 Awareness, intention, expectancy, and plausibility. Cognition and consciousness in nonhuman species: Open peer commentary. Behav. Brain Sci. 4, 560. Bindra, D. 1981. Ape language-Letter to Science. Science 211, 86. Bloom, L. M. 1973. “One Word at a Time: The Use of Single-Word Utterances Before Syntax.” Mouton, The Hague. Bloom, L. M., Rocissano. L., and Hood, L. 1976. Adult child discourse. Cognit. Psychol. 8, 52 1-552. Bronowski, J., and Bellugi, U. 1970. Language, name and concept. Science 168, 669-673. Brown, R. 1970. The first sentences of child and chimpanzee. In “Selected Psychoinguistic Papers’’ (R.Brown, ed.), pp. 808-822. Macmillan, New York. Brown, R. 1973. “A First Language.” Harvard Univ. Press, Cambridge, Massachusetts. (As quoted in B. T. Gardner and Gardner, 1979.) Bruner. J. S. 1966. On cognitive growth, 1. and 11. In “Studies in Cognitive Growth” (J. S. Bruner, R. R. Olver, and P. M. Greenfield, eds.0, pp. 1-67. Wiley, New York. Campbell, C. B. G., and Hodos, W. 1970. The concept of homology and the evolution of the nervous system. Brain. Behav. Evol. 3, 353-367. (As quoted in Malmi, 1976.) Chomsky, N . 1959. Review of Skinner’s Verbal Behavior. Language 35, 26-58. Reprinted (1964) with comments in “The Structure of Language in the Philosophy of Language” (J. A. Fodor and J. J . Katz, eds.), pp. 547-578. Prentice-Hall, Englewood Cliffs, New Jersey. Chomsky, N. 1979. Human language and other semiotic systems. Semiorics 25, 31-44. Reprinted (1980) in “Speaking of Apes: A Critical Anthology of Two-way Communication With Man” (T. A. Sebeok and J . Umiker-Sebeok, eds.), pp. 429-440. Plenum, New York. Crook, J . H. 1980. “The Evolution of Human Consciousness.” Oxford Univ. Press, London and New York. Dawkins, R., and Krebs, J . R. 1978. Animal signals: Information or manipulation? In “Behavioural Ecology” (J. R. Krebs and N . B. Davies, eds.), pp. 282-309. Sinauer. Sunderland, Massachusetts. Dennett, D. C. 1978a. “Brainstorms.” Bradford, Montgomery. Vermont. Dennett. D. C. 1978b. Beliefs about beliefs. Open peercommentary. Behav. Brain Sci. 4, 568-570. Desmond, A. J . 1979. “The Ape’s Reflexion.” Dial, New York. Epstein, R., Lanza, R. P., and Skinner, B. F. 1980. Symbolic communication between two pigeons (Columba livis domestira). Science 1,07, 543-545. Farrer, D. N. 1969. Picture memory in the chimpanzee. Percept. Motor Skills 25, 305-315. Fischer, S. D. 1974. The ontogenetic development of language. In “Language and Language Disturbances: The Fifth Lexington Conference on Phenomenology” (E. W. Straus. ed.). Duquesne Univ. Press, Pittsburgh, Pennsylvania. (As quoted in Gardner and Gardner, 1979.) Fouts. R. S . 1972. Use of guidancc in teaching sign language to a chimpanzee (Pun troglodvtes). J . Comp. Phvsiol. Psychol. 80, 515-522.

LANGUAGE IN THE GREAT APES

249

Fouts. R. S. 1973. Acqu on and testing of gestural signs in four young chimpanzees. Science 180, 978-980. Fouts. R. S. 1974. Capactities for languages in great apes. I n “Proceedings of the 18th International Congress of Anthropological and Ethnological Sciences, ” pp. 1-20, Mouton, The Hague. Fouts, R. S. 1977. Ameslan in Pan. In “Progress in Ape Research” (G. H. Bourne, ed.), pp. 117-123. Academic Press, New York. Fouts, R. S., Chown. W., and Goodin, L. T. 1976. Transfer of signed responses in American Sign Language from vocal English stimuli to physical object stimuli by a chimpanzee ( P a n ) . Learn. M ~ t i v 7. , 458-475. Fouts, R. S . , Kimball, G . H., Davis, D. L., and Mellgren, R. L. 1979. Washoe and her children. Paper presented at Twentieth Annual Meeting of the Psychonomic Society, Phoenix, Arizona, November 8, 1979. Fouts, R. S., Hannum, M. E., O’Sullivan, C., and Schneider, K. Chimpanzee conversations: Language, cognition and theory. I n “Problems, Theories and Controversies in Language Development: Language, Cognition and Culture” (S. Kuczaj 11, ed.). Erlbaum, Hillsdale, New Jersey. (In press.) Furness, W . 1916. Observations on the mentality of chimpanzees and orangutans. Proc. Am. Philos. SUC. 6 5 , 281-290. Gallup, G . G . . Jr. 1977. Self-recognition in primates: A comparative approach to the bidirectional properties of consciousness. Am. Psychol. 32, 329-338. Gardner, B. T., and Gardner. R. A. 1971. Two-way communication with an infant chimpanzee. I n ”Behavior of Nonhuman Primates,” Vol. 4 (A. M. Schrier and F. Stollnitz, eds.), pp. 117183. Academic Press, New York. Gardner. B. T.. and Gardner, R. A. 1974. Comparing the early utterances of child and chimpanzee. In “Minnesota Symposium in Child Psychology,” Vol. 8 (A. Pick, ed.), pp. 3-23. Univ. of Minnesota Press. Minneapolis. Gardner. B. T.. and Gardner, R. A. 1975. Evidence for sentence constituents in the early utterances of child and chimpanzee. J . Exp. Psychol.: Gen. 104, 244-267. Gardner, B. T.. and Gardner. R. A . 1979. Two comparative psychologists look at language acquisition. ln “Children’s Language,” Vol. 2 (K.E. Nelson, ed.), pp. 309-369. Halsted. New York. Gardner, R. A.. and Gardner. B. T. 1969. Teaching sign language to a chimpanzee. Science 165, 664-672. Gardner. R. A.. and Gardner, B. T. 1974. Teaching sign language to the chimpanzee: Washoe (text of the soundtrack of the film). Bulletin d’Audio Phoriologie 4, 1-38. Gardner, R. A,, and Gardner, B. T. 1975. Early signs of language in child and chimpanzee. Science 187, 752-753. Gardner, R. A . , and Gardner, 8. T. 1978. Comparative psychology and language acquisition. Ann. N . Y. A l ’ U t l . S C ~309, . 37-76. Gardner, R. A . . and Gardner, B. T. 1980. Comment on Terrace. NYU Educational Quarterly. Gill. T. 1977. Conversations with Lana. I n “Language Learning by a Chimpanzee” (D. M. Rumbaugh, ed.). pp. 225-246. Academic Press, New York. Gillan. D. J . 1981. Reasoning in the chimpanzee: 11. Transitive inference. J. Exp. Psychol. 7, 150- 164. Gillan. D. J . , Premack. D., and Woodruff, G. 1981. Reasoning in the chimpanzee: I . Analogical reasoning. J . Exp. Psychol. 7 , 1-17. Ginsburg, N . 1963. Conditioned talking in the mynah bird. J . Comp. Phwiof. Psychol. 5 6 , 10611063. Goffman. E. 1963. “Behavior in Public Places: Notes on the Social Organization of Gatherings.” Free Press, New York.

250

CAROLYN A. RISTAU AND DONALD ROBBlNS

Goldin-Meadow, S . , and Feldman, H. 1977. The development of language-like communication without a language model. Science 197, 401-403. Gould, J . L. 1978. Behavioral programming in honeybees: Open peer commentary. Behuv. Bruin Sci. 4, 572-573. Green, S. 1975. Variation of vocal pattem with social situation in the Japanese monkey (Mucuca fuscuta); A field study. In “Primate Behavior,” Vol. 4 (L. A. Rosenblum, ed.), pp. 1-102. Academic Press, New York. Grice, H. P. 1967. Logic and conversation. William James Lectures, Harvard University. In “Studies in Syntax” (P. Cole and J . L. Morgan, eds.), Vol. 3. Academic Press, New York, 1975. Griffin, D. R . 1976. “The Question of Animal Awareness: Evolutionary Continuity of Mental Experience,” 2nd ed.. 1981. The Rockefeller Univ. Press, New York. Griffin, D. R. 1978. Prospects for a cognitive ethology. Behav. Bruin Sci. 4, 527-538. Hallowell, A. I. 1955. “Culture and Experience.” Univ. of Pennsylvania Press, Philadelphia. Hayes, C. 1951. “The Ape in Our House.” Harper,New York. Hayes, K., and Hayes. C. 1951. The intellectual development of a home-raised chimpanzee. Proc. Am. Philos. Soc. 95, 105-109. Healy, A. F. 1973. Can chimpanzees learn a phonemic language’? J. Comp. Physiol. Psycho/. 56, 1061-1063. Reprinted (1980) in “Speaking of Apes: A Critical Anthology of Two-way Communication with Man”(T. A. Sebeok and J . Umiker-Sebeok, eds.), pp. 141-143. Plenum, New York. Hediger, H. 1968. “The Psychology and Behaviour of Animals in Zoos and Circuses.” Dover, New York. Held, R., and Hein, A. 1963. Movement-produced stimulation in the development of visually guided behavior. J. Comp. Physiol. Psycho/. 56, 872-876. Herman, L. M. 1980. Cognitive characteristics of dolphins. In “Cetacean Behavior: Mechanisms and Functions” (L. M. Herman, ed.), pp. 363-429. Wiley, New York. Hewes, G. W. 1977. Language origin theories. In “Language Learning by a Chimpanzee: The Lana Project” (D. M. Rumbaugh, ed.). pp. 3-53. Academic Press, New York. Hewett, F. M. 1965. Teaching speech to an autistic child through operant conditioning. A m . J . Orthopsychiut. 35, 927-936. Hockett, C. F. 1958. “A Course in Modem Linguistics.” Macmillan, New York. Hockett, C. F., and Altmann. S. A. 1968. A note on design features. In “Animal Communication” (T. A. Sebeok, ed.), pp. 574-575. Indiana Univ. Press, Bloomington. Humphrey, N. K . 1980. Nature’s psychologists. In “Consciousness and the Physical World” (B. Josephson and B. S. Ramachandra, eds.), pp. 57-80. Pergamon, New York. Jakobson, R., Fant, C. G. M., and Halle, M. 1969. “Preliminaries to Speech Analysis.” MIT Press, Cambridge. Massachusetts. Kavanagh. M. 1980. Invasion of the forest by an African savannah monkey: Behavioral adaptations. Behavior 73, 238-260. Kellogg. W. N . . and Kellogg. L. A. 1933. “The Ape and the Child: A Study of Environmental Influence Upon Early Behavior.” Whittlesey House, New York. [Reprinted (1967) by Haffner, New York] King, M. C., and Wilson, A. C. 1975. Evolution at two levels in humans and chimpanzees. Science 188, 107-1 16. Klima, E. S . , and Bellugi. U. 1972. The signs of language in child and chimpanzee. In “Communication and Affect” (T. Alloway, ed.), pp. 67-96. Academic Press, New York. Klima, E. S., Bellugi, U., Battison, R., Boyes-Braem, P., Fischer, S., Frishberg, N . , Lane, H., Lentz, E. M., Newkirk. D., Newport, E., Pedersen, C. C., and Siple, P. 1979. “The Signs of Language. Harvard Univ. Press, Cambridge, Massachusetts. Kohler, W . 1925. “The Mentality of Apes.” Harcourt, New York. ”

LANGUAGE IN THE GREAT APES

25 1

Kroodsma, D. E. 1980. Winter wren singing behavior: A pinnacle of song complexity. Condor 82, 357-365. Kummer, H. 1982. Social knowledge in free-ranging primates. In “Animal Mind-Human Mind” (D. R. Griffin, ed.), pp. 113-130. Springer-Verlag, Berlin and New York. Laidler, K . 1978. Language in the orang-utan. In “Action, Gesture and Symbol: The Emergence of Language” (A. Lock, ed.), pp. 133-155. Academic Press, New York. Langbauer, W. 1982. Prerequisites for language in a porpoise? Ph.D. Thesis, Boston University. Lenneberg, E. H. 1967. “Biological Foundations of Language.” Wiley, New York. Lenneberg, E. H. 1969. A word between us. In “Communication” (J. D. Roslansky, ed.), pp, I I I - I3 I . North-Holland, Amsterdam. Reprinted ( I 980) in “Speaking of Apes: A Critical Anthology of Two-way Communication with Man” (T. A. Sebeok and J. Umiker-Sebeok, eds.), pp. 71-83. Plenum, New York. Lenneberg, E. H. 1971. Of language knowledge, apes, and brains. J. Psycholing. Res. 1, 1-29. Reprinted (1980) in “Speaking of Apes: A Critical Anthology of Two-way Communication With Man” (T. A. Sebeok and J. Umiker-Sebeok, eds.), pp. 115-140. Plenum, New York. Lenneberg. E. H. 1975. A neuropsychological comparison between man, chimpanzee and monkey. Neuropsvchology 13, 125. Lewis, D. 1969. “Convention: A Philosophical Study.” Harvard Univ. Press, Cambridge. Massachusetts. Lieberman, P. 1967. “Intonation, Perception and Language.” MIT Press, Cambridge, Massachusetts Lieberman, P. 1968. Primate vocalizations and human linguistic ability. J. Acoust. SOC. Am. 44, 1574- 1584. Lieberman. P. 1975. The evolution of speech and language. In “The Role of Speech in Language” (J. F. Kavanagh and J . E. Cutting, eds.). pp. 83-106. MIT Press, Cambridge, Mass. Lieberman, P., Crelin, E. S., and Klatt, D. H. 1972. Phonetic ability and related anatomy of the newborn and adult human, neanderthal man, and the chimpanzee. Am. Anthropol. 74,287-307. Linden, E. 1974. “Apes, Men, and Language.” Saturday Review PresdDutton, New York. Lloyd, J . E. 1977. Bioluminescence and communication. In “How Animals Communicate” (T. A. Sebeok, ed.), pp. 164-183. Indiana Univ. Press, Bloomington, Indiana. McDermott, D. 1981. Artificial intelligence meets natural stupidity. In “Mind Design” (J. Haugeand, ed.), pp. 143-160. Bradford Books, Montgomery, Vermont. McNeill. D. 1974. Sentence structure in chimpanzee communication. In “The Growth of Competence” ( K . Connolly, and J. Bruner, eds.), pp. 75-94. Academic Press, New York. Malmi, W. A. 1976. Chimpanzees and language evolution. Ann. N . Y. Acad. Sci. 280, 598-603. Reprinted (1980) in “Speaking of Apes: A Critical Anthology of Two-way Communication with Man” (T. A. Sebeok, and J. Umiker-Sebeok, eds.), pp. 191-196. Plenum, New York. Marler, P. 1977a. The evolution of communication. In “How Animals Communicate” (T. A. Sebeok, ed.), pp. 45-70. Indiana Univ. Press, Bloomington, Indiana. Marler, P. 1977b. The structure of animal communication sounds. In “Recognition of Complex Acoustic Signals” (T. H. Bullock, ed.). Abakon Verlagsgesellschaft, Berlin. Mayr, E. 1969. “Principles of Systematic Zoology.” McGraw-Hill, New York. (As quoted in Malmi, 1976.) Medin, D. L., Roberts, W. A , , and Davis, R. T. (Eds.) 1976. “Processes of Animal Memory.” Erlbaum, Hillsdale, New Jersey. Menzel, E. W. 1974. A groupof young chimpanzees in aone-acre field. In “Behavior of Nonhuman Primates” (A. M. Schrier and F. Stollnitz, eds.), Vol. 4 , pp. 83-153. Academic Press, New York. Miles, H. L. W. 1978. Conversations with apes: The use of sign language by two chimpanzees. Ph.D. Thesis, University of Connecticut. Moynihan, M. 1976. “The New World Primates.” Princeton Univ. Press, Princeton, New Jersey.

252

CAROLYN A . RISTAU AND DONALD ROBBINS

Nagel. T. 1974. What is it like to be a bat'? Philos. Rev. 83, 435-450. Natsoulas. T. 1978a. Consciousness. Am. Psychol. 33, 906-914. Natsoulas. T. 1978b. Toward a model for consciousness, in the light of B. F. Skinner's contribution. Brhuviorism 6, 139-175. Nelson, K . 1973. Structure and strategy in learning to talk. Monogr. Soc. Res. Child Drv. 38, 1-137. Nottebohm, F. 1969. The "critical period" for song learning. Ibis 111, 386-387. Parkel, D. A., and Smith, S. T., Jr. 1979. Application of computer-assisted language designs. In "Language Intervention from Ape to Child" (R. L. Schiefelbusch and J . H. Hollis, ed;.), pp. 44-464. Univ. Park Press, Baltimore, Maryland. Passingham. R. E. 1979. Specialization and the language areas. It1 "Neurobiology of Social Communication in Primates: An Evolutionary Perspective" (H.D. Steklis and M. J . Raleigh, eds.), pp. 221 -256. Academic Press. New York. Patterson, F. G. 1978a. The gestures of a gorilla: Language acquisition in another pongid. Bruiu Lung. 5, 72-97. Patterson, F. G. 1978b. Linguistic capabilities of a young lowland gorilla. In "Sign Language and Language Acquisition in Man and Ape: New Dimensions in Comparative Pedolinguistics" (F. C. Peng, ed.), pp. 161-201. Westview, Boulder, Colorado. Patterson, F. 1978~.Conversations with a gorilla. Nail. Geogr. 154, 438-465. Patterson. F. G . 1979a. Linguistic capabilities of a lowland gorilla. Ph.D. Thesis, Stanford University. (University Microfilms International Edition, Ann Arbor, Michigan.) Patterson, F. G. 1979b. Innovative uses of language by a gorilla: A case study. In "Children's Language," Vol. 2 (K. E. Nelson, ed.). pp. 497-561. Gardner, New York. Patterson, F. G. 1981. Ape language - Letter to Science. Science 211, 86-87. Patterson, F., and Linden, E. 1981. "The Education of Koko." Holt. New York. Pepperberg. I. M. 1981. Functional vocalizations by an African grey parrot (Psirrucus eriihacus). Z. Tierpswhol. 5, 139-160. Petitto, L. A., and Seidenberg, M. S. 1979. On the evidence for linguistic abilities in signing apes. Bruin Lung. 8, 162-183. Piaget. J . , and Inhelder. B. 1969. "The Psychology of the Child." Basic Books, New York. Plooij, F. X. 1978. Some basic traits of language in wild chimpanzees? In "Action. Gesture, and Symbol: The emergence of Language" (A. Lock, ed.), pp. I 11-131. Academic Press. New York. Plooij, F. 1980. The behavioural development of free-living chimpanzee babies and infants. Ph.D. Thesis, Rijksuniversiteit Te Groningen. Poirier, F. E. 1973. "Fossil Man: An Evolutionary Journey." Mosby, St. Louis, Missouri. Premack, D. 1971. On the assessment of language competence in the chimpanzee. In "Behavior of Nonhuman Primates,'' Vol. 4 (A. M. Schrier an0 F. Stollnitz, eds.), pp. 186-228. Academic Press, New York. Premack, D. 1972. Teaching language to an ape. Sri. Am. 227, 92-99. Premack. D. 1976. "Intelligence in Ape and Man." Erlbaum, Hillsdale, New Jersey. Premack, D.. and Schwartz. A. 1966. Preparations for discussing behaviorism with a chimpanzee. In "Genesis of Language" (F. Smith and G. A. Miller, eds.), pp. 295-335. MIT Press, Cambridge, Massachusetts. Premack, D., and Woodruff, G. 1978a. Does the chimpanzee have a theory of mind? Behuv. Brain Sci. 4, 515-526. Premack, D., and Woodruff, G. 1978b. Cognition and consciousness in nonhuman species: Authors' responses. Behav. Bruin Sri. 4, 616-629. Ristau, C. A. 1965. Concept formation with and without verbal labels, under conditions of social and non-social reinforcement. Master's Thesis, University of Pennsylvania. Ristau. C. A. 1974. Infant vocal communication: A comparison of normal and mongoloid humans,

LANGUAGE I N THE GREAT APES

253

other primates and carnivores. Ph.D. Thesis, University of Pennsylvania. (University Microfilms. Ann Arbor, Michigan.) Ristau. C. A. 1979. Are there talking chimpanzees’? Paper presented at the Symposium on Consciousness in Contemporary Psychology. Dalhousie University, Halifax. Nova Scotia, August 1979. Ristau, C. A. 1980. Ape language. cognition and culture, review of Desmond, Adrian, “The Ape’s Reflexion,” Dell, New York, 1979. Nature (London) 284, 684-685. Ristau. C. A,. and Robbins, D. 1979. A threat to man’s uniqueness? Language and communication in the chimpanzee. J . Psycholing. Res. 8, 267-300. Ristau. C. A,. and Robbins, D. 1982. Cognitive aspects of ape language experiments. In “Animal Mind-Human Mind” (D. R . Griffin, ed.), pp. 299-331. Springer-Verlag. Berlin and New York. Robbins, D. 1977. Cognitive processing in great apes (Pongidae). In “Progress in Ape Research’’ (G. H. Bourne, ed.), pp. 259-276. Academic Press, New York. Rosenthal, R.. and Rubin. D. B. 1978. Intrapersonal expectancy effects: The first 345 studies. Behav. Bruin Sci. 3, 377-415. Rumbaugh, D. M. (Ed.) 1977. “Language Learning by a Chimpanzee: The Lana Project.“ Academic Press, New York. Rumbaugh, D. M. 1981. Who feeds Clever Hans? Ann. N . Y. Acad. Sci. 364,26-34. Rumbaugh, D. M., and Gill, T. V. 1977. Lana’s acquisition of language skills. In “Language Learning by a Chimpanzee: The Lana Project” (D. M. Rumbaugh, ed.), pp. 165-192. Academic Press, New York. Ryle, G. 1949. “The Concept of Mind.” Barnes and Noble, New York. Savage-Rumbaugh, E. S. 1979. Symbolic communication-its origins and early development in the chimpanzee. New Dir. Child Dev. 3 , 1-15. Savage-Rumbaugh, E. S., and Rumbaugh, D. M . 1978. Symbolization, language and chimpanzees: A theoretical reevaluation based on initial language acquisition processes in four young Pan troglodytes. Brain Lang. 6 , 265-300. Savage-Rumbaugh, E. S., and Rumbaugh, D. M. 1979. Chimpanzee problem comprehension: Insufficient evidence. Science 206, 1201-1202. Savage-Rumbaugh, E. S., and Rumbaugh, D. M. 1980. Requisites of symbolic communication-a, Are words for birds? Psvchol. Rec. 30, 305-318. Savage-Rumbaugh, E. S., Wilkerson, B. J . , and Bakeman, R. 1977. Spontaneous gestural communication among conspecifics in the pygmy chimpanzee (Pan paniscus). I n “Progress in Ape Research“ (G. H. Bourne, ed.), pp. 97-1 16. Academic Press, New York. Savage-Rumbaugh, E. S., Rumbaugh, D. M., and Boysen, S. 1978a. Symbolic communication between two chimpanzees (Pan troglodytes). Science 201, 641-644. Savage-Rumbaugh, E. S., Rumbaugh, D. M., and Boysen, S. 1978b. Linguistically mediated tool use and exchange by chimpanzees (Pan troglodytes). Behav. Brain Sci. 4, 539-554. Savage-Rumbaugh, E. S.. Rumbaugh, D. M., and Boysen, S. 1980a. Do apes use language? Am. Sci. 68, 49-61, Savage-Rumbaugh, E. S., Rumbaugh, D. M., Smith, S. T . , and Lawson, J . 1980b. Reference- the linguistic essential. Science 210, 922-925. Sebeok, T. A . , and Rosenthal, R. 1981. “The Clever Hans Phenomenon: Communication with Horses, Whales, Apes, and People. ” Ann. N. Y . Acad. Sci. 364, 1-3 1 1. Seidenberg, M. S., and Petitto, L. A. 1979. Signing behavior in apes: Acritical review. Cognilion 7 , 177-215. Seyfarth, R . M.. and Cheney, D. L. 1980. The ontogeny of vervet alarm calling behavior: A preliminary report. Z. Tierpsychol. 54, 37-56. Seyfanh, R. M., Cheney. D. L., and Marler, P. (1980a). Vervet monkey alarm calls: Semantic communication in a free-ranging primate. Anim. Behav. 28, 1070-1094.

254

CAROLYN A . RlSTAU AND DONALD ROBBINS

Seyfarth, R. M.. Cheney, D. L., and Marler, P. 1980b. Monkey responses to three different alarm calls: Evidence of predator classification and semantic communication. Science 210, 801-803. Seyfarth, R. M., Beer, C. G.. Dennett, D. C., Gould, J . L., Lindauer, M., Marler, P.R . , Ristau, C. A,, Savage-Rumbaugh, E. S.. Solomon, R. C., and Terrace, H. S . 1982. Communication as evidence of thinking. In “Animal Mind-Human Mind” (D. R. Griffin, ed.), pp. 391 -406. Springer-Verlag, Berlin and New York. Sherman, P. W . 1977. Nepotism and the evolution of alarm calls. Science 147, 1246-1253. Simmons. K. E. L. 1955. The nature of the predator. Reactions of waders towards humans; with special reference to the role of the aggressive- escape- and brooding-drives. Behaviour 8, 130- 173. Simpson, G . G. 1961. “Principles of Animal Taxonomy.” Columbia Univ. Press, New York. Skinner, B. F. 1957. “Verbal Behavior.” Appleton, New York. Smith, W . J . 1965. Message, meaning, and context in ethology. Am. Nar. 99, 405-409. Smith, W. J . 1968. Message-meaning analyses. In “Animal Communication” (T. A. Sebeok, ed.), pp. 44-60,Univ. of Indiana Press, Bloomington, Indiana. Smith, W. J . 1977. “The Behavior of Communicating: An Ethological Approach.” Harvard Univ. Press, Cambridge, Massachusetts. Steklis, H. D., and Raleigh, M. J. 1979. Requisites for language: Interspecific and evolutionary aspects. In “Neurobiology of Social Communication” (H. D. Steklis and M. J. Raleigh, eds.), pp. 283-314. Academic Press, New York. Stokoe, W . J . , Jr. 1960. Sign language structure: An outline of the visual communication systems of the American deaf. (Studies in Linguistics, Occasional Papers 8). Univ. of Buffalo Press, Buffalo, New York. Straub. R. 0.. Seidenberg, M. S., Terrace, H. S . , and Bever, T. G. 1979. Serial learning in the pigeon. J . Exp. Anal. Behav. 32, 137-148. Struhsaker, T. 1967. Auditory communication among vervet monkeys (Cercopirhecus aerhiops). In “Social Communication Among Primates’’ (S. A. Altmann, ed.), pp. 281-324. Univ. of Chicago Press, Chicago, Illinois. Studdert-Kennedy, M. 1980. Language by hand and by eye. A review of E. S. Klima and U. Bellugi “The Signs of Language.” Cognition 8, 93-108. Sugiyama. Y., and Koman, J. 1979. Tool-using and making behavior in wild chimpanzees at Bossou, Guinea. Primares 20, 513-524. Terrace, H. S. 1979a. Is problem-solving language? J . Exp. Anal. Behav. 31, 161-175. Terrace, H. S. 1979b. “Nim.” Knopf, New York. Terrace, H. S., and Bever. T. G. 1976. What might be learned from studying language in the chimpanzee? The importance of symbolizing oneself. Ann. N . Y . Acad. Sci. 280, 579-588. Terrace, H. S . , Petitto, L.A,, Sanders, R. J.. and Bever, T. G. 1979a. Can an ape create a sentence? Science 200, 891-902. Terrace, H., Petitto. L. A., Sanders, R. J . , and Bever, T. G . 1979b. On the grammatical capacity of apes. In “Children’s Language,” Vol. 2 (K. Nelson, ed.), pp. 371-495. Gardner. New York. Terrace, H.. Petitto. L. A., Sanders, R. J.. and Bever. T. G. 1981. Ape language - Letter to Science. Science 211, 87-88. Thompson, C. R., and Church, R. M. 1980. An explanation of the language of a chimpanzee. Science 208, 313-314. Thorpe, W . H. 1974. “Animal Nature and Human Nature.” Anchor, Garden City, New York. Turney, T. H. 1980. Discriminative vocal responding in the Indian Hill mynah (Gracula religiosa). Paper presented at the Animal Behavior Society Annual Meeting, Fort Collins, 1980. Umiker-Sebeok. J., and Sebeok, T. A. 1980. Introduction: Questioning apes. In “Speaking of Apes: A Critical Anthology of Two-way Communication with Man” (T. A. Sebeok and J. UmikerSebeok, eds.), pp. 1-59. Plenum. New York.

LANGUAGE IN THE GREAT APES

255

van Lawick-Goodall, J . 1968. The behavior of free-living chimpanzees in the Gombe Stream Reserve. Ani. Behav. Monogr. 1, 161-31 I . van Lawick-Goodall, J . 1971. “In the Shadow of Man.” Houghton-Mifflin, Boston, Massachusetts. van Lawick-Goodall, J. 1973. Cultural elements in a chimpanzee community. In “Precultural Primate Behavior” (E. W. Menzel, ed.). Karger, Basel. Verner, J . 1976. Complex song repertoire of male long-billed marsh wrens in eastern Washington. The Living Bird 14, 263-300. von Glasersfeld, E. 1977. Linguistic communication: Theory and definition. In “Language Learning by a Chimpanzee: The Lana Project’’ (D. Rumbaugh, ed.). pp. 55-71. Academic Press, New York. Woodruff, G., and Premack, D. 1979. Intentional communication in the chimpanzee: The development of deception. Cognifion 7 , 333-362. Woodruff, G . , and Premack, D. 1981. Primitive mathematical concepts in the chimpanzee: Proportionality and numerosity. Nature (London) 293, 568-570. Yerkes, R. M., and Yerkes, A. 1929. “The Great Apes: A Study of Anthropoid Life.” Yale Univ. Press, New Haven. Connecticut. Zaidel, E. 1978. Of apes and hemispheres. Cognition and consciousness in nonhuman species: Open peer commentary. Behav. Brain Sci. 4, 607-609.

This Page Intentionally Left Blank

Index

A

Activity cycles, integration of visceral and somatic activity and, 112-1 13 Adaptive complex, learning and, 92 interactions of selective pressures and adaptive traits and, 93-94 limits on adaptive precision and, 94-96 Amealan, see Signing Ape language, 142-143 artificial lexicons and, 178 Lana project, 178-185 Premack project, 186-188 history of, 143-145 implications for human language and cognitive development, 230 analogy or homology of linguistic-like behaviors to human language, 234-236 criteria for human language and, 230-234 implications for children's communication, 236-237 investigations of mental states and, 207-219 labeling and, research directions in. 242 meaning and categorical sorting and, 205-207 errors made in generalization and, 192-193 feature analysis of an object and its label, 196-198 functional definitions of words, 198-205 labeling and, 189-191 novel uses of a word and, 192 novel word combinations and, 193-196 research directions for, 241-242 mental states and. research directions in, 243 problems raised by and implications for research, 237 Clever Hans phenomenon, 237-239 257

overly critical analysis, 239-240 suggestions for future directions, 240-245 relation to animal cognition and natural animal communication, 219 artificial language projects in other species and, 227-228 aspects of linguistic abilities used in natural environment, 223-227 cognition and. 219-220 communication and, 221-223 evolutionary forces leading to linguisticlike abilities in apes and early man, 228 implications for research, 228-230 signing and, 172-178 comparison of projects in, 155-162 Gardners' projects in, 162-166 Gorilla Koko studies, 17 1 - 172 Nim project in, 167-171 theoretical issues in ability to be linguistic, 151-154 definitions of language and, 145-151 pragmatic applications of, 154 Approach behavior, Pavlovian conditioning and, 6-8 Aversion learning, poison rejection and, 19-22

B Backward conditioning, in interspecific defense, 37-38 Body imagery. metaphoric language and, 124128 Bradycardia, conditioned, 3 1-34 Brain rhythms, in visceral-somatic coordination, 115-1 16 Brooding solicitation, conditioning of, 47-48

258

INDEX

C

F

Cardiovascular activity ideational imagery associated with, 123- 124 visceral-somatic integration and, neural mechanisms of, 113-1 16 Central nervous system brain rhythms and, in visceral-somatic coordination, 115-1 16 greater complexity of, as cost of learning, 86-89 Children, communication in, implications of ape language for, 236-237 Clever Hans phenomenon, ape language and, 237-239 Cognition, in animals, ape language and, 219220 Communication, see Ape language; Language Conditioning. see Backward conditioning; Learning; Pavlovian conditioning Consciousness, ape language and, 244 Consummatory behavior, Pavlovian conditioning and, 8-14 Cost-benefit analysis, evolution of adaptations and, 70-73 Courtship, Pavlovian conditioning of, 43-44

Feature analysis, of object and its label, ape language and, 196-198 Flight response, 34-36 Food aversion learning, poison rejection and, 19-22 Food-procuring behavior, Pavlovian conditioning and, 8-14 Food recognition, Pavlovian conditioning and, 14-17 Foraging behavior, Pavlovian conditioning and, 6 consummatory and food-procuring behavior and, 8-14 development of food recognition and, 14-17 digestive responses and, 17-19 locomotory search and approach behavior and, 6-8 rejection and ingestion of toxins and, 19-28 Fright response, 31-34

D Defensive behavior, Pavlovian conditioning of, 28 backward conditional response and, 37-38 interspecific, 28-38 intraspecific, 38-43 Developmental fallibility, as cost of learning, 91-92 Digestive responses, Pavlovian conditioning and, 17-19 Disease, psychosomatic, concept of, 129- 134

E Environment natural. ape linguistic abilities used in, 223227 variability of, adaptation to, 74-76 Evolution learning and, 66-69 of linguistic-like abilities, in apes and early man, 228

G Generalization, errors in, ape language and, 192-193 Genome, greater complexity of, as cost of learning, 89-91

I Image(s), internal, of apes, 198-205 Imagery, ideational cardiac cycle and, 123-124 metaphoric language and, 124-128

J Juvenile vulnerability, increased, as cost of learning, 83-84

L Labeling, ape language and meaning and, 189-191 research directions in, 242 Language, see also Ape language definitions of, 145-151 metaphoric, body imagery and, 124-128

259

INDEX Learning, 65-66, see dso Backward conditioning; Pavlovian conditioning adaptive complex and, 92 interactions of selective pressures and adaptive traits and, 93-94 limits on adaptive precision and, 94-96 aversion, poison rejection and, 19-22 cost-benefit analysis and evolution of adaptations and, 70-73 developmental fallibility and, 91-92 ecological conception of, 69-70 evolution and, 66-69 selective benefits of, 73 adaptation to environmental variability, 74-76 lack of variation for other adaptive solutions, 77-79 sexual selection, 76-77 selective costs of, 79 delayed reproductive effort andor success, 79-82 developmental fallibility, 91-92 greater complexity of central nervous system, 86-89 greater complexity of genome, 89-91 increased juvenile vulnerability, 83-84 increased parental investment in each offspring, 84-86 study of, 96-98 Lexicons, artificial, 178 Lana project, 178- 185 Premack project, 186- 188 Locomotor activity, visceral-somatic integration and, neural mechanisms of, 113-1 16 Locornotory search, Pavlovian conditioning and, 6-8 Lying, complex, ape language and, 244

M Meaning, ape language and, 188-189 categorical sorting and, 205-207 errors made in generalization and, 192-193 feature analysis of an object and its label, 196- I98 functional definitions of words, 198-205 labeling and, 189-191 novel uses of a word and, 192 novel word combinations and, 193-196 research directions for, 241-242

Mental states, ape. language and, research directions in, 243 Milk-ejection reflex, conditioning of, 4 - 4 7

N Neuroendocrine reflexes, 111-1 12

0 Offspring, increased parental investment in, as cost of learning, 84-86

P Parental behavior increased investment in offspring, as cost of learning, 84-86 Pavlovian conditioning of, 46-48 Pavlovian conditioning, 1-3 of defensive behavior, 28 backward conditional response and, 37-38 interspecific, 28-38 intraspecific, 38-43 foraging behavior and, 6 consummatory and food-procuring behavior and, 8-14 development of food recognition and, 14-17 digestive responses and, 17-19 locomotory search and approach behavior and, 6-8 rejection and ingestion of toxins and, 19-28 prefiguring hypothesis and, 3-6, 48-49 naturally occurring conditional stimuli as learned releasers and, 50-51 non-Pavlovian learning and, 49-50 of reproductive behavior, 43-44 courtship, 44-46 parental behavior, 46-48 Prefiguring hypothesis, 3-6, 48-49 naturally occurring conditional stimuli as learned releasers and, 50-5 I non-Pavlovian learning and, 49-50 Prevarication experiments, 243-244 Psychosomatic disease, concept of, 129-134

R Recognition, of food, Pavlovian conditioning and. 14-17

260

INDEX

Reflexes milk-ejection, conditioning of, 46-47 neuroendocrine, I 1 I - I I2 somatovisceral, 109-1 10 underlying cardiovascular, respiratory, and locomotor activity, 113-115 viscerosomatic, 1 10-1 1 I Rejection, of toxins, Pavlovian processes in, 19-28 Reproduction, delayed effort or success of, as cost of learning, 79-82 Reproductive behavior, Pavlovian conditioning of, 43-44 courtship, 44-46 parental behavior, 46-48 Respiratory activity, visceral-somatic integration and, neural mechanisms of, 113I I6 S

Self-description, ape language and, 244 Sexual selection, learning and, 76-77 Signing, in apes, 172-178 comparison of projects in, 155-162 Gardners' projects in, 162-166 Gorilla Koko studies, 171-172 Nim project in, 167-171 Somatovisceral reflexes, 109- I 10 Sorting, ape language and, 243 categorical sorting, 205-207 Shuli exteroceptive, visceral-somatic integration and, 116-119 naturally occurring, as learned releasers, 50-5 1

T Tachycardia. conditioned, 34-36 Toxins, rejection and ingestion of, Pavlovian processes in, 19-28

v Visceral-somatic integration, 107-108 higher-order, I12 coupling of activity cycles and, 112- I 13 neural mechanisms underlying cardiovascular, respiratory, and locomotor activity, 113-1 16 perceptual-motor aspects of, 116- I19 subjective aspects of, 119-123 ideational imagery and, 123 cardiac cycle and, 123-124 metaphoric language and, 124- 128 linguistic, interpersonal, and aesthetic expressions of, 121-122 neuroendocrine reflexes in, 1 I I - I 12 physiological consequences of rhythmic activity, 122-123 psychogenic organic disease concept and, 129-134 somatovisceral reflexes in, 109-1 10 viscerosomatic reflexes in, 1 10- I 1 I Viscerosomatic reflexes, 110-1 1 1 W Words functional definitions of, ape language and, 198-205 novel combinations of, ape language and, 193- I96 novel uses of, ape language and, 192

Contents of Previous Volumes

Volume 1

Volume 3

Aspects of Stimulation and Organization in ApproachiWithdrawal Processes Underlying Vertebrate Behavioral Development T. C. SCHNEIRLA

Behavioral Aspects of Homeostasis D. J . McFARLAND Individual Recognition of Voice in the Social Behavior of Birds C. G. BEER

Problems of Behavioral Studies in the Newborn Infant H. F. R . PRECHTL

Ontogenetic and Phylogenetic Functions of the Parent-Offspring Relationship in Mammals LAWRENCE V. HARPER

The Study of Visual Depth and Distance Perception in Animals RICHARD D. WALK

The Relationships between Mammalian Young and Conspecifics Other Than Mothers and Peers: A Review Y . SPENCER-BOOTH

Physiological and Psychological Aspects of Selective Perception GABRIEL HORN

Tool-Using in Primates and Other Vertebrates JANE VAN LAWICK-GOODALL

Current Problems in Bird Orientation KLAUS SCHMIDT-KOENIG

Author

Habitat Selection in Birds P. H. KLOPFER and J . P. HAILMAN Author

IndexSubjecr index

Volume 4

IndexSubject index

Volume 2

Constraints on Learning SARA J. SHETTLEWORTH

Psychobiology of Sexual Behavior in the Guinea Pig WILLIAM C. YOUNG

Female Reproduction Cycles and Social Behavior in Primates T. E. ROWELL

Breeding Behavior of the Blowfly

The Onset of Maternal Behavior in Rats, Hamsters, and Mice: A Selective Review ELAINE NOIROT

V . G. DETHIER

Sequences of Behavior R. A. HINDE and J. G. STEVENSON

Sexual and Other Long-Tern Aspects of Imprinting in Birds and Other Species KLAUS IMMELMANN

The Neurobehavioral Analysis of Limbic Forebrain Mechanisms: Revision and Progress Report KARL H. PRlBRAM

Recognition Processes and Behavior, with Special Reference to Effects of Testosterone on Persistence R. J. ANDREW

Age-Mate or Peer Affectional System HARRY F. HARLOW Aurhor

IndexSubject index

Author

26 I

Index-Subject Index

262

CONTENTS OF PREVIOUS VOLUMES

Volume 5

Volume 7

Some Neuronal Mechanisms of Simple Behavior KENNETH D. ROEDER

Maturation of the Mammalian Nervous System and the Ontogeny of Behavior PATRICIA S. GOLDMAN

The Orientational and Navigational Basis of Homing in Birds WILLIAM T. KEETON The Ontogeny of Behavior in the Chick Embryo RONALD W. OPPENHEIM Processes Governing Behavioral States of Readiness WALTER HEILIGENBERG Time-sharing as a Behavioral Phenomenon D. J . McFARLAND Male-Female Interactions and the Organization of Mammalian Mating Patterns CAROL DIAKOW Author IndexSubject Index

Volume 6 Specificity and the Origins of Behavior P. P. G.BATESON The Selection of Foods by Rats, Humans, and Other Animals PAUL ROZIN

Functional Analysis of Masculine Copulatory Behavior in the Rat BENJAMIN D. SACHS and RONALD J. BARFIELD Sexual Receptivity and Attractiveness in the Female Rhesus Monkey ERIC B. KEVERNE Prenatal Parent-Young Interactions in Birds and Their Long-Term Effects MONICA IMPEKOVEN Life History of Male Japanese Monkeys YUKIMARU SUGIYAMA Feeding Behavior of the Pigeon H. PHILIP ZEIGLER Subjeci Index

Volume 8 Comparative Approaches to Social Behavior in Closely Related Species of Birds FRANK McKlNNEY

Social Transmission of Acquired Behavior: A Discussion of Tradition and Social Learning in Vertebrates BENNETT G . GALEF, JR.

The Influence of Daylength and Male Vocalizations on the Estrogen-Dependent Behavior of Female Canaries and Budgerigars, with Discussion of Data from Other Species ROBERT A. HINDE and ELIZABETH STEEL

Care and Exploitation of Nonhuman Primate Infants by Conspecifics Other Than the Mother SARAH BLAFFER HRDY

Ethological Aspects of Chemical Communication in Ants BERTHOLLDOBLER

Hypothalamic Mechanisms of Sexual Behavior, with Special Reference to Birds J. B. HUTCHISON

Filial Responsiveness to Olfactory Cues in the Laboratory Rat MICHAEL LEON

Sex Hormones, Regulatory Behaviors, and Body Weight GEORGE N. WADE

A Comparison of the Properties of Different Reinforcers JERRY A. HOGAN and T. J. ROPER

Subjecr Index

Subjecr Index

CONTENTS OF PREVIOUS VOLUMES

Volume 9 Attachment as Related to Mother-Infant Interaction MARY D. SALTER AINSWORTH Feeding: An Ecological Approach F. REED HAINSWORTH and LARRY L WOLF

263

Progress in the Study of Maternal Behavior in the Rat: Hormonal, Nonhormonal. Sensory, and Developmental Aspects JAYS. ROSENBLA'IT. HAROLD I. SIEGEL, and ANNE D. MAYER Subject Index

Volume 11 Progress and Prospects in Ring Dove Researct A Personal View MEI-FANG CHENG Sexual Selection and Its Component Parts, Somatic and Genital Selection. as Illustrated by Man and the Great Apes R. V . SHORT Socioecology of Five Sympatric Monkey Species in the Kibale Forest. Uganda THOMAS T. STRUHSAKER and LYSA LELAND Ontogenesis and Phylogenesis: Mutual Constraints GASTON RICHARD Subject Index

Volume 10 Learning, Change, and Evolution: An Enquiry into the Teleonomy of Learning H. C. PLOTKIN and F. J. ODLING-SMEE Social Behavior, Group Structure, and the Control of Sex Reversal in Hermaphroditic Fish DOUGLAS Y. SHAPIRO Mammalian Social Odors: A Critical Review RICHARD E. BROWN The Development of Friendly Approach Behavior in the Cat: A Study of Kitten-Mother Relations and the Cognitive Development of the Kitten from Birth to Eight Weeks MILDRED MOELK

Interrelationships among Ecological, Behavioral, and Neuroendocrine Processes in the Reproductive Cycle of A n d i s curdinensis and Other Reptiles DAVID CREWS Endocrine and Sensory Regulation of Maternal Behavior in the Ewe PASCAL POINDRON AND PIERRE LE NEINDRE The Sociobiology of Pinnipeds PIERRE JOUVENTIN AND ANDRE CORNET Repertoires and Geographical Variation in Bird Song JOHN R. KREBS AND DONALD E. KROODSMA Development of Sound Communication in Mammals GUNTER EHRET Ontogeny and Phylogeny of Paradoxical Reward Effects ABRAM AMSEL AND MARK STANTON lngestional Aversion Learning: Unique and General Processes MlCHAEL DOMJAN The Functional Organization of Phases of Memory Consolidation R. 1. ANDREW index

This Page Intentionally Left Blank