Linking up contrastive and learner corpus research (Language & Computers)

Introduction Gaëtanelle Gilquin Université catholique de Louvain Szilvia Papp University of Portsmouth María Belén Díez-Bedmar Universidad de Jaén Most of the papers collected in this volume were presented at the workshop “Linking up contrastive and learner corpus research”, held in September 2005 in conjunction with the 4th International Contrastive Linguistics Conference that took place in Santiago de Compostela (Spain). What these papers have in common is that they all use learner corpora (i.e. large computerised collections of authentic data produced by learners of a foreign/second language) and that they do so from a contrastive perspective. “Contrastive”, here, means several things. It can refer to “contrastive interlanguage analysis” (see Granger 1996), which involves either the comparison of learner language with native language (as exemplified by Ädel’s study) or the comparison of several learner languages (as in Lozano & Mendikoetxea’s chapter). While the former type of comparison makes it possible to highlight features of non-nativeness in the learners’ production, such as misuse, overuse or underuse, the latter may point to problems which are shared by learners from different mother tongue backgrounds and hence are likely to be developmental (age-related, proficiency-related, etc). Another sense in which the approach in this volume may be described as “contrastive” is that several studies combine learner corpus analysis with contrastive analysis, comparing the target language and the learner’s mother tongue (see e.g. Demol & Hadermann’s paper). This dual approach, exemplified by the “integrated contrastive model” (Granger 1996, Gilquin 2000/2001), sheds light on the place of the mother tongue in the acquisition of the target language and foregrounds the notion of language transfer. The volume is divided into four parts, which correspond to four fields of investigation. The first part deals with methodological issues. After setting the scene for the combination of contrastive and learner corpus research, Gaëtanelle Gilquin proposes a model for the study of transfer-related phenomena, namely the Detection-Explanation-Evaluation (DEE) transfer model. This model, which relies on two existing frameworks for transfer research (Granger 1996 and Jarvis 2000), uses corpus data and combines contrastive analysis and interlanguage analysis. By drawing a number of comparisons between the mother language, the target language and the interlanguage, the model aims to detect cases of transfer, explain them and evaluate their pedagogical relevance. The DEE transfer model is illustrated by means of a case study of the use of even if by French-speaking

viii

Gaëtanelle Gilquin, Szilvia Papp and María Belén Díez-Bedmar

learners, which shows how and why this expression displays traces of transfer and what consequences, if any, it should have on the foreign language teaching curriculum. Some limitations of the model are also discussed, with a view to further improving it and thus gaining an even better understanding of transfer. In the second paper, Annelie Ädel tackles the issue of corpus comparability and demonstrates that differences in foreign language production attributed to the learner’s mother tongue may in fact result from discrepancies in task setting (i.e. how much time is available) and/or intertextuality (i.e. whether access is given to secondary sources). Focusing on the presence of involvement features in the written production of Swedish learners of English, Ädel compares corpus data differing along the two parameters of task setting and intertextuality, and shows that the degree of involvement in learner writing is reduced when learners are involved in untimed tasks and when they are given secondary sources as input. In particular, the first factor turns out to affect the presence of first person pronouns, whilst the second factor is shown to have an influence on features such as disjuncts, questions and exclamations. The paper ends with some recommendations for testing practices in language teaching. In the second part of the volume, which focuses on learner lexis, John Cross and Szilvia Papp study verb + noun combinations as used by Chinese learners of English. By comparing their corpus data with similar data for Greek learners (Giovi 2006) and German learners (Nesselhauf 2003), they discover that Chinese learners use more non-native combinations than German and Greek learners. They also argue that Chinese learners are less creative in their production of verb + noun combinations than the other two groups of learners, who use such combinations in more experimental and innovative ways. Cross and Papp offer several explanations for this difference. Besides the influence of the mother tongue and the lack of exposure to the target language, they suggest that environment-related factors might be at work, including general beliefs widely held in the Peoples’ Republic of China and, more importantly, teaching and learning practices such as the memorisation and (inappropriate) recall of chunks of language. The next four papers investigate learner syntax. Cristóbal Lozano and Amaya Mendikoetxea examine the properties that govern the occurrence of postverbal subjects in English texts produced by Spanish and Italian advanced learners, and argue that postverbal subjects can only be accounted for by considering three interfaces, namely the lexicon-syntax interface, the syntaxphonology interface and the syntax-discourse interface. For each of these interfaces, they propose a hypothesis, which they then test on the basis of learner corpus data. The results of the analysis confirm the three hypotheses, namely that Spanish and Italian learners of English produce postverbal subjects with unaccusative verbs, not unergative verbs; that they tend to produce heavy subjects in postverbal position and light ones in preverbal position; and that they consistently produce focus subjects in postverbal position and topic subjects in preverbal position. No significant differences were found to exist between the two groups of learners regarding the conditions under which postverbal subjects

Introduction

ix

occur. John Osborne’s paper is devoted to adverb placement, and more particularly Verb-Adverb-Object order, in the written production of learners of English from a variety of L1 backgrounds. The author shows that non-native-like adverb placement is still likely to occur at a post-intermediate stage, although more so among learners whose mother tongue has obligatory verb-raising (Romance languages), which may suggest a transfer of syntactic patterns. However, the fact that speakers of V2 Germanic languages, which also permit VAdv-O order, produce very few such sequences and that speakers of non-raising Slavic languages produce some of them proves that transfer is not the only explanation and that other factors must be at work too. These, according to Osborne, include semantic and phraseological factors, as evidenced by learners’ tendency to produce V-Adv-O sequences with “lighter” NPs and “weaker” VAdv collocations than native speakers would do. The paper closes with some desiderata for enriching the potential for contrastive studies based on learner corpora. María Belén Díez-Bedmar and Szilvia Papp are interested in the use of the English article system by two groups of learners, one having no article system in their mother tongue, viz. Chinese-speaking learners, and the other having an article system slightly different from the English one, viz. Spanish-speaking learners. Using Bickerton’s (1981) semantic wheel and Huebner’s (1983) subsequent taxonomy, as well as the theory of definiteness within pragmatics, the authors combine a contrastive analysis, which highlights the differences between the article use in English, Spanish and Chinese, with a contrastive interlanguage analysis, in which they test the hypotheses put forward on the basis of the contrastive analysis. The results of the contrastive interlanguage analysis show, among others, that, as predicted by the contrastive analysis, Chinese learners exhibit more non-native features in their overall use of the English article system than Spanish learners. In her paper, Christelle Cosme focuses on the use of adverbial and adnominal present/past participle clauses. Using comparable and translation corpus data, she examines whether the underuse of participle clauses by French- and Dutch-speaking learners of English observed by Granger (1997) may be explained by cross-linguistic influence, considering the common claim that English makes greater use of participle constructions than French and Dutch. Her data suggest that transfer may be put forward as a plausible explanation for Dutch-speaking learners. In the case of French-speaking learners, on the other hand, transfer seems to be at work only with present participle clauses. Cosme’s paper ends with a caveat that other factors may contribute to the underuse of participle clauses by French- and Dutch-speaking learners and should therefore also be investigated. The last part of the volume includes three papers, which all deal with learner discourse. Marcus Callies considers the phenomenon of raising constructions, adopting a functional-typological approach to second language acquisition. After reviewing the literature on the topic and highlighting the high degree of typological markedness of English raising constructions, he examines the frequency of occurrence and contextual use of raising constructions in the written production of advanced German and Polish learners of English. It

x

Gaëtanelle Gilquin, Szilvia Papp and María Belén Díez-Bedmar

emerges from this analysis that raising constructions, and in particular toughmovement (or object-to-subject raising) constructions, are problematic for both groups of learners, being as a rule underrepresented and misused in terms of information structuring and thematic progression. Callies concludes that the use of raising constructions by learners is the result of a complex interplay of factors including avoidance, transfer of training and unawareness of raised structures’ discourse functions. Mike Hannay and Elena Martínez Caro’s paper investigates the theme zone in the written production of Dutch and Spanish learners of English. Dutch and Spanish differ from English with regard to the structure of the theme area, and they also differ from each other. Yet, the authors show that the two groups of learners construct the beginning of their sentences in essentially the same way, using the various thematic patterns in comparable proportions (including an overuse of more complex structures). The differences that do exist point to a higher degree of grammatical competence among the Dutch learners, as well as a more elaborate use of the thematic options available in English. The Spanish learners, by contrast, produce a number of ill-formed structures, some of them possibly resulting from transfer, and they tend to exhibit a more limited range of functionalities. In the final chapter, Annemie Demol and Pascale Hadermann study discourse organisation in written narratives (film retellings) produced by French and Dutch native speakers, French-speaking learners of Dutch and Dutch-speaking learners of French. Alongside a contrastive analysis (French vs. Dutch), they are thus able to perform a contrastive interlanguage analysis, comparing the learner varieties with the native varieties (to detect possible traces of transfer), as well as the two interlanguages (to identify any common strategies). Their hypotheses concern the degrees of packaging, dependency and integration of the narratives. One important difference emerging from the analysis is the higher frequency of present participles in secondary predication in French as compared to Dutch, which may explain the tendency of Dutch learners of French to underuse this type of structure. The comparison of French L2 and Dutch L2 reveals no clear organisational pattern typical of interlanguage, although they both display slightly less complex discourse organisation than the native varieties in terms of average number of words per sentence and proportion of simple sentences. Although they deal with different topics, the ten papers brought together in this volume all underline the relevance of linking up contrastive analysis with learner corpus research. This is indeed the case if the goals and methods of contrastive linguistics are applied to learner corpora, as numerous studies in the past have already made clear and as confirmed by this book. But this is also true if contrastive data (ideally in the form of parallel and/or comparable corpora) are actually integrated into the analysis of learner corpora, a trend which has been relatively discreet until now but whose potential has been illustrated by several of the papers in this volume. Judging by the present collection and the workshop from which it originated, the combination of contrastive and learner corpus research is likely to stimulate more research studies in the near future and provide new insights into second language acquisition. Such insights could benefit

Introduction

xi

professionals working in the field, such as language teachers, materials writers or language testers. In fact, it is our hope that some of the findings described in this volume may also lead to practical applications in the classroom, so that the ground-breaking research reported here can be linked up with teaching. References Bickerton, D. (1981), Roots of Language. Ann Arbor, MI: Karoma Press. Gilquin, G. (2000/2001), ‘The Integrated Contrastive Model. Spicing up your data’, Languages in Contrast, 3(1): 95-123. Giovi, C. (2006), The use of verb-noun collocation by Greek advanced learners of English using corpora analysis and some implications for teaching. Unpublished MA dissertation, University of Portsmouth. Granger, S. (1996), ‘From CA to CIA and back: an integrated approach to computerized bilingual and learner corpora’, in: K. Aijmer, B. Altenberg and M. Johansson (eds) Languages in Contrast. Papers from a Symposium on Text-based Cross-linguistic Studies. Lund 4-5 March 1994. Lund: Lund University Press. 37-51. Granger, S. (1997), ‘On identifying the syntactic and discourse features of participle clauses in academic English: native and non-native writers compared’, in: J. Aarts, I. de Mönnink and H. Wekker (eds) Studies in English Language and Teaching. In Honour of Flor Aarts. Amsterdam & Atlanta: Rodopi. 185-198. Huebner, T. (1983), A Longitudinal Analysis of the Acquisition of English. Ann Arbor, MI: Karoma. Jarvis, S. (2000), ‘Methodological rigor in the study of transfer: identifying L1 influence in the interlanguage lexicon’, Language Learning, 50(2): 245309. Nesselhauf, N. (2003), ‘The use of collocations by advanced learners of English and some implications for teaching’, Applied Linguistics, 24(2): 223-242.

Combining contrastive and interlanguage analysis to apprehend transfer: detection, explanation, evaluation Gaëtanelle Gilquin Centre for English Corpus Linguistics F.N.R.S. – Université catholique de Louvain Abstract This paper shows how contrastive analysis and interlanguage analysis can be combined into a model which aims to detect phenomena of transfer, explain them and evaluate them in terms of pedagogical relevance. The model, referred to as the Detection-ExplanationEvaluation (DEE) transfer model, relies on Granger’s (1996) Integrated Contrastive Model and Jarvis’s (2000) unified framework for transfer research, and uses corpus data as a basis for a range of comparisons between L1 (mother tongue), L2 (target language) and IL (interlanguage). As an illustration, the use of even if by French-speaking learners is investigated, and the possible influence of L1 is examined in the light of the different comparisons involved in the model. Finally, some limitations of the DEE transfer model are discussed, all of which point to possible developments that could be implemented in the future.

1.

Introduction

The importance of transfer in second language acquisition can now be regarded as an established fact. As observed by Kohn (1986: 21) two decades ago already, “there is no doubt that, despite its sometimes irritatingly elusive character, transfer is one of the major factors shaping the learner’s interlanguage competence and performance”. Two main sources of data have been used over the years to investigate the influence of the mother tongue (L1) on the interlanguage (IL), namely contrastive analysis (CA) data and learner data. Selinker (1989: 287) notes that “CA is the best place to begin language transfer studies since structural congruence (or at the least, partial structural similarity) is most probably necessary, though not sufficient, for most types of language transfer to occur”. Indeed, Schachter’s (1974) analysis, which investigated the influence of L1 on the use of English restrictive relative clauses by learners from several mother tongue backgrounds, was demonstrated to be flawed, for lack of reliable CA data (Zhao 1989, Kamimoto et al. 1992). Without samples of learner production to rely on, however, studies of transfer, as faultless as they may be from a contrastive point of view, could offer nothing but useless predictions. Both contrastive analysis and interlanguage analysis are therefore necessary for a sound and systematic assessment of the role of transfer in second language acquisition.1

4

Gaëtanelle Gilquin

In this paper, two models of transfer which combine a CA and an IL approach, Granger (1996) and Jarvis (2000), are compared in an attempt to develop a refined method to apprehend transfer. Three varieties of language will be considered, viz. L1, L2 (target language) and IL, and will lead to a number of comparisons which aim to detect transfer (when is it at work?), explain it (what are its origins?) and evaluate it from a pedagogical perspective (is it pedagogically relevant?). Reflecting this threefold purpose, the model will be referred to as the Detection-Explanation-Evaluation (DEE) transfer model. Because the interest here is in phenomena of transfer as they manifest themselves in naturally-occurring language, use will be made of corpora, whose value for linguistic descriptions has been underlined by numerous studies over the last few years. The paper is organised as follows. First, the phenomenon of transfer is briefly described and recontextualised, with special emphasis on its links with contrastive analysis and interlanguage analysis. Granger’s (1996) Integrated Contrastive Model and Jarvis’s (2000) unified framework for the study of L1 influence are then presented and compared. On the basis of this comparison, a new model of transfer is proposed, the DEE transfer model, which serves the threefold purpose of detecting, explaining and evaluating cases of transfer. The model is illustrated by means of a case study of the use of even if by Frenchspeaking learners, after which some of its limitations are outlined. The article ends with some concluding remarks. 2.

The transfer mystery

After a period in the 1950s and 60s when transfer was seen as all-pervasive and a period in the 1970s when its existence was denied, we have now reached a time when it is “generally accepted that transfer does occur, but is a far more complex phenomenon than hitherto believed” (Benson 2002: 68). Its exact nature, the circumstances in which it occurs or the processes it relies on in the learner’s mind are still, to a large extent, a mystery to linguists. As Dechert & Raupach (1989: ix) put it in their introduction, “[i]n spite of three decades of intensive research […] there is still no generally accepted agreement of what transfer in language actually is”. Lado (1957: 2) refers to the transfer of “the forms and meanings and the distribution of forms and meanings of [individuals’] native language and culture to the foreign language and culture”. Odlin’s (1989: 27) definition is broader, as it covers the influence of “any other language that has been previously (and perhaps imperfectly) acquired”. In other words, transfer is not limited to the influence of L1 on L2, but could involve the influence of, say, L2 on L3 or L3 on L5. Neither of these definitions, however, explicitly raises the possibility of a simultaneous influence of several languages. Nor do they allow for the fact that a language, most typically the native language, may be influenced by subsequently acquired languages. In order to take such phenomena into account, transfer will be defined here in very general terms, as the influence, within an individual’s

Combining contrastive and interlanguage analysis

5

linguistic system, of one or more languages over another.2 Moreover, we will consider both cases of negative transfer, when the influence leads to erroneous usage,3 and positive transfer, when it leads to correct usage. The notion of transfer is closely related to contrastive analysis. In fact, as pointed out by Ringbom (1995: 581), the original idea behind contrastive analysis was to bring to light the problems that learners from a specific mother tongue background might experience in learning a given foreign language. However, the study of learner language (so-called “interlanguage”, see Selinker 1972), initially investigated within the frame of error analysis, revealed that problems could occur in cases not predicted by contrastive analysis and, conversely, that not all areas predicted as difficult by contrastive analysis were actually problematic for learners (Ringbom 1995). Thus, while both contrastive analysis and interlanguage analysis appeared necessary to make discoveries about transfer, it also turned out that the link between the two was far from obvious and required further research. The model of transfer presented here, the DEE transfer model, relies on two models combining contrastive and interlanguage analysis, namely Granger (1996) and Jarvis (2000), and aims to provide a systematic method for detecting transfer, explaining it and evaluating its pedagogical relevance. It involves three varieties of language (L1, L2 and IL) and six types of comparison,4 as summarised in Table 1, which also shows the abbreviations that will be used throughout the article. Of these six comparisons, the first three are instances of contrastive analysis, whereas the last three involve so-called contrastive interlanguage analysis.5 Each of them serves a particular (primary) purpose, namely detection for NL1/IL and ILa/ILb…ILz (see Section 6.1), explanation for OL1/OL2 and SL1/2/TL2/1 (see Section 6.2) and evaluation for NL2/IL and ILa/ILa (see Section 6.3). Table 1. Types of comparison and their abbreviations Abbreviation

Explanation

OL1/OL2

Comparison of original L1 and original L2 Comparison of source L1 and translated L2 or vice versa

SL1/2/TL2/1 NL1/IL NL2/IL ILa/ILa ILa/ILb…ILz

Comparison of native L1 and IL Comparison of native L2 and IL Comparison of interlanguage by learners from the same L1 Comparison of interlanguage by learners from different L1s

Example for French learner of English Original French vs. original English Source French vs. translated English or source English vs. translated French Native French vs. learner English Native English vs. learner English French learner English vs. French learner English French learner English vs. Spanish learner English vs. German learner English, etc

6

Gaëtanelle Gilquin

Although the model is mainly illustrated through cases of transfer from L1 French to L2 English, and most of the examples concern lexical (and some syntactic) phenomena, it is possible to apply the DEE transfer model to other languages, other types of transfer – e.g. from L2 to L3 (Hammarberg 2001: 23) or from L2 to L1 (Pavlenko & Jarvis 2002) – and other phenomena (phonological, pragmatic, etc).6 And while all the illustrations come from written corpus data, the model could also serve to study transfer in speech, provided the necessary material is available. 3.

Granger’s (1996) Integrated Contrastive Model

In Granger’s (1996) Integrated Contrastive Model (ICM), two components are combined in an attempt to shed light on the phenomenon of transfer. The contrastive analysis (CA) component compares (i) original data from one language with original data from another language and (ii) source language and its translation in another language. The contrastive interlanguage analysis (CIA) component, similarly, consists of two types of comparison: (i) native language and a non-native variety of this language and (ii) different non-native varieties of one and the same language. The peculiarity of the model is that it exclusively relies on authentic data coming from computerised corpora. This has not always been the case in contrastive and interlanguage studies. The first comparisons of two or more languages were often intuition-based and the traditional error analyses of learner language usually relied on very small collections of texts.7 Recently, however, with the computer revolution, large amounts of machine-readable authentic data have become available. Parallel corpora (that is, corpora of original texts and their translations into one or several languages) and comparable corpora (that is, corpora of original texts in two or more languages which are similar in terms of domain, genre, etc) have made it possible to reassess a number of contrastive claims. Tisthoud (1992), for instance, shows on the basis of bilingual corpus data that the morphological flexibility of English as compared to French has been largely overestimated, with a number of French derivations being translated by means of a paraphrase in English (e.g. clochettes = little bells, improductivité = lack of productivity). Large and balanced corpora of interlanguage have also been compiled over the last few years. One of the first corpora of this type was the International Corpus of Learner English (ICLE, see Granger et al. 2002), the latest version of which includes essays written by advanced learners of English from 16 different mother tongue backgrounds (Granger et al. forthcoming). Not only have such corpora allowed for the contextualisation of errors, but they have also enabled researchers to investigate what learners get right, what they overuse, i.e. use significantly more than native speakers, and what they underuse, i.e. use significantly less than native speakers. Corpora, therefore, have contributed to the

Combining contrastive and interlanguage analysis

7

development of both CA and CIA, thus making the model resulting from their integration a powerful tool, as will be shown presently. As appears from Figure 1, which represents the ICM, the notion of transfer occupies a central position, linking CA and CIA predictively and diagnostically. The predictive hypothesis (also called “CA a priori” or “strong CA hypothesis”) starts from the comparison of two languages (CA), either as original languages (OL vs. OL) or as source language and translated language (SL vs. TL), and on this basis makes predictions about the learner’s interlanguage. The assumption is that in the case of discrepancies between the learner’s mother tongue and the target language, the learner is likely to transfer the L1 pattern to his/her interlanguage, hence producing an erroneous L2 pattern (negative transfer). In the case of similarities between L1 and L2, on the other hand, the learner is expected to produce a correct pattern in L2 (positive transfer). By examining learner production data (CIA) and looking for traces of L1, one can test the accuracy of the predictions and thus establish the (potential) presence, or otherwise, of transfer. The diagram in Figure 1, by using a broken line to connect the CA and the CIA boxes, emphasises the hypothetical nature of the CA-based predictions: L1/L2 mismatches do not always lead to errors8 in interlanguage and, conversely, L1/L2 identity does not necessarily imply error-free use by learners (Gilquin 2000/2001: 101). The diagnostic hypothesis (also called “CA a posteriori” or “weak CA hypothesis”) follows the opposite path, starting from CIA and moving on to CA. By comparing the learner data with a reference native corpus (NL vs. IL) and with data produced by learners from other mother tongue backgrounds (IL vs. IL), one may notice L1-specific errors and look to contrastive analysis for an explanation. The diverging arrows in the diagram indicate that the explanation for an error will not always be found in the relation between the learner’s mother tongue and the target language. The errors observed may be due to other factors than L1 influence, e.g. intralingual factors, developmental factors, teachinginduced factors or communication strategies in interlanguage (see James 1980: 146). As pointed out by Granger (1996: 46), the ICM “involves constant to-ing and fro-ing between CA and CIA”. This can be illustrated by means of the analysis of the use of the passive by French-speaking learners of English. A quantitative study of the passive in native English and French-speaking learners’ English [CIA: NL vs. IL] brings out a significant underuse of the passive by the learners (Granger 1997). In order to test the role of L1 in this underuse, one may compare the frequency of the passive in native French and native English. A comparable corpus of newspaper articles [CA: OL vs. OL] reveals that the passive is about twice as frequent in English as in French (Fivet 1995, Granger 1998), which seems to suggest that the French learners’ underuse is transferrelated. Before drawing any such conclusion, however, it is safe to consider other varieties of learner English. From this examination [CIA: IL vs. IL], it turns out that the underuse of the passive is not limited to French-speaking learners, but also concerns learners from other L1 backgrounds (Granger 1997 highlights this tendency for Swedish- and Finnish-speaking learners as well), which appears to

8

Gaëtanelle Gilquin

challenge the transfer-related explanation. One last verification is necessary, however, namely checking whether the other native languages do not exhibit a lower frequency of the passive than in English, as French does. Using an English/French and an English/Swedish parallel corpus [CA: SL vs. TL], Granger (1998) shows that the proportion of English passives that are translated by means of a passive is higher in Swedish than in French. The frequency of the passive in Swedish is close to that of the passive in English, so that Swedish learners’ underuse of the passive cannot be attributed to L1 influence. The underuse of the passive exhibited by French-speaking learners of English therefore seems to be a universal feature of interlanguage, which may be related to learners’ preference for the unmarked rather than marked option (Ellis 1994), or for a personal rather than impersonal style (see e.g. Petch-Tyson 1998). The influence of L1 French, however, cannot be excluded, and it is probable that transfer reinforces the effect of other factors.

Figure 1. Integrated Contrastive Model (Gilquin 2000/2001: 100, based on Granger 1996: 47)

Combining contrastive and interlanguage analysis 4.

9

Jarvis’s (2000) unified framework for transfer research

In his (2000) article, Jarvis, noting the amount of confusion that reigns in the field of transfer research, offers a possible solution in the form of a unified framework consisting of three components, viz. a theory-neutral definition of transfer, a list of the types of evidence to be considered and a set of outside variables to be controlled. Jarvis’s proposed definition will not be discussed here (although it is important to stress that the types of evidence he advocates are derived from this definition), but the other two components will be examined, with particular emphasis on the evidence that should be produced to establish the existence of transfer-related phenomena with some confidence. Jarvis recognises that the essence of transfer “lies beyond the reach of the researcher” (ibid. 253-254), but he proposes a list of three effects which are subject to direct investigation and which should be considered before one is justified in positing the existence (or at least likelihood) of L1 influence. These three effects are: (i) (ii) (iii)

intra-L1-group homogeneity in learners’ IL performance; inter-L1-group heterogeneity in learners’ IL performance; intra-L1-group congruity between learners’ L1 and IL performance.

The first effect of L1 influence, the intra-L1-group homogeneity in learners’ IL performance, refers to the fact that learners who share the same mother tongue should exhibit a similar behaviour when using the target language. In other words, a particular feature should not be the idiosyncrasy of one or two individuals, but should characterise the group as a whole. The second effect, the inter-L1-group heterogeneity in learners’ IL performance, means that learners coming from different mother tongue backgrounds should perform differently in the target language. The behaviour of Italian learners of English, for example, should be specific to this learner population, and distinct from the behaviour displayed by, say, Finnish or Chinese learners of English. The third effect, the intra-L1-group congruity between learners’ L1 and IL performance, points to a parallelism between the learner’s mother tongue and his/her interlanguage. If the interlanguage presents a particular feature, there should be a corresponding feature in L1. While each of these effects has the potential to reveal the existence of L1 influence, Jarvis demonstrates that none of them is necessary nor sufficient in itself, as transfer may take place despite the absence of one of the effects and, conversely, an effect may be present but not transfer-related. To give but one example, several L1s may share a particular feature (especially if they belong to the same language family) and therefore produce the same effect on IL. In such a case, inter-L1-group homogeneity would not exclude the possibility of transfer. This leads Jarvis to argue that at least two effects out of the three he mentions should be identified before concluding that transfer is at play.

10

Gaëtanelle Gilquin

In addition, Jarvis claims that any rigorous investigation of transfer should control for a number of outside variables, which may either promote or inhibit L1 influence. On the basis of the literature, he draws the following list of variables (Jarvis 2000: 260-261): a) b) c) d) e) f) g) h) i)

age; personality, motivation and language aptitude; social, educational and cultural background; language background (all previous L1s and L2s); type and amount of target language exposure; target language proficiency; language distance between the L1 and target language; task type and area of language use; prototypicality and markedness of the linguistic feature.

Jarvis illustrates his methodological framework by means of a study of the referential lexical choices made in English by Finnish-speaking and Swedishspeaking Finns. Using three different elicitation tasks (film retell, lexical listing and selection of lexical items), he investigates how a number of objects and actions are referred to by the participants, a carefully selected subject pool consisting of several learner groups at various levels of age, L2 exposure and L2 proficiency, and their L1-control group counterparts. All the variables listed by Jarvis (see above) were controlled for, except the second one (personality, motivation and language aptitude). The analysis of the results shows that, generally speaking, the learners’ lexical choices exhibit the three effects identified by Jarvis as typical of L1 influence – although their presence may at times be quite subtle. First, it turns out that, despite differences in age and L2 exposure, learners with the same mother tongue background tend to make similar lexical choices. Second, while the comparison of the lexical choices made by learners with distinct mother tongues does not reveal any significant differences, learners from different L1 backgrounds and matched according to age and L2 exposure still exhibit higher levels of heterogeneity than learners from the same L1 background and with differences in age and L2 exposure. Finally, although the comparison of learners’ interlanguage and mother tongue provides only partial support for the third effect, the other two types of comparison (first and second effects) show evidence of L1-IL similarities. Jarvis also demonstrates that, while some outside variables appear to have an influence on the results (e.g. age, L2 exposure, task type), none of them produces as consistent effects as the L1 variable. His study therefore underlines the importance of the mother tongue background in the interlanguage lexicon of Finnish- and Swedish-speaking learners of English. More generally, it also demonstrates the usefulness of a rigorous methodological framework to investigate the role of transfer in interlanguage studies. While Jarvis uses elicitation data to illustrate his model, one could envisage using authentic corpus data instead, in view of the advantages of corpora

Combining contrastive and interlanguage analysis

11

outlined in Section 3. Paquot (2006) adopts a corpus-based approach to Jarvis’s framework, testing the role of L1 influence on French-speaking learners’ overuse of the expression on the contrary. Her study relies on learner data from several ICLE components representing different mother tongue backgrounds, as well as comparable native data produced by French- and English-speaking students (CODIF and LOCNESS, see later). She tests the first and second effects (intra-L1group homogeneity and inter-L1-group heterogeneity in learners’ IL performance) by means of a simple analysis of variance (ANOVA), which indicates that the difference in frequency between the ICLE components is more significant than the difference between the distinct essays within each component. Computing pairwise t-tests with Bonferroni adjustment, she is then able to show that it is the French data which are mainly responsible for the statistically significant difference between the ICLE components. In other words, it turns out from the statistical analysis that the variability between the essays written by learners with different mother tongues is significant (inter-L1-group heterogeneity) and is greater than the variability within the essays written by the French-speaking learners (intra-L1-group homogeneity). In order to investigate the third effect, the intra-L1-group congruity between learners’ L1 and IL performance, Paquot compares the relative frequency of on the contrary in French-speaking learners’ interlanguage with its relative frequency in a corpus of native English on the one hand, and with the relative frequency of the French equivalent, au contraire, in a corpus of native French on the other hand. It appears that the results of the learner corpus are more similar to the results of the native French corpus than to those of the native English corpus, which suggests that French-speaking learners tend to transfer the frequency of au contraire when producing on the contrary. The presence of Jarvis’s three effects leads Paquot to conclude that the use of on the contrary by French-speaking learners is (at least partly) L1-induced.9 5.

Comparison of Granger’s (1996) ICM and Jarvis’s (2000) framework

Granger’s (1996) and Jarvis’s (2000) models share a number of important features. The most obvious one is that they are both concerned with the phenomenon of transfer. In addition, although Jarvis does not use these specific terms, the two models approach transfer by combining contrastive analysis (comparison of two or more languages) and contrastive interlanguage analysis (comparison of two or more varieties of a language, including non-native language). While the specific types of comparison within CA and CIA are essentially different, as we will see below, there is one type which is shared by the two models, namely the comparison of several non-native varieties of the same language (ILa/ILb…ILz, which corresponds to Jarvis’s second effect, interL1-group heterogeneity in learners’ IL performance). Finally, Granger and Jarvis both acknowledge the existence of a number of outside variables that should be rigorously controlled. While these variables are not part of the ICM as such, Granger (1996: 44) underlines their importance, with particular emphasis on the

12

Gaëtanelle Gilquin

type of target language exposure (foreign or second language), the target language proficiency and the task type. In fact, the learner corpus she uses to test the ICM, the International Corpus of Learner English, has been carefully designed so as to keep these three variables constant (foreign language rather than second language, advanced proficiency level10 and essay writing) and has been encoded with additional variables which can easily be manipulated when selecting the data, e.g. amount of L2 exposure, knowledge of other foreign languages, gender, task setting, etc (see Granger 2003). Next to these similarities, the two models also present differences. Despite their common research object (transfer), the models have slightly different goals. The ICM, as we saw, has both a predictive and a diagnostic goal, whereas Jarvis’s model simply aims to identify instances of transfer. Second, Granger and Jarvis use different data – corpus data and experimental data, respectively – although, as illustrated by Paquot’s (2006) study, Jarvis’s model can be operationalised by means of corpus data as well. The models also differ with respect to the types of CA and CIA analyses they advocate. For Granger, the contrastive analysis should consist in a comparison of the target language and the learner’s mother tongue (either as original languages in a comparable corpus or as source and translated languages in a parallel corpus). Jarvis, on the other hand, proposes the comparison of the learner’s mother tongue with his/her interlanguage (see endnote 5 on why this is considered a case of CA). In terms of contrastive interlanguage analysis, besides the common ILa/ILb…ILz comparison, Granger includes as an individual (and essential) component of her model a comparison of the interlanguage and its native counterpart. Jarvis uses this type of comparison too, but only as part of the larger IL-IL comparative analysis, treating NL2 as if it were another IL. Moreover, he regards this type of evidence as unreliable (Jarvis 2000: 300). Finally, Jarvis, unlike Granger, includes in his model a comparison of the interlanguage of several learners from the same mother tongue background. The similarities and differences between the two models are summarised in Table 2. It will appear from the next section that all these elements have a role to play in a model of transfer, but that they serve different purposes. Table 2. Comparison of the features of Granger’s ICM and Jarvis’s framework Granger’s (1996) ICM Jarvis’s (2000) framework Object of study = transfer Combination of CA and CIA ILa/ILb…ILz Control of outside variables Predictive and diagnostic Identification Corpus data Experimental data (but applicable to corpus data) NL1 vs. IL L1 vs. L2 (OL1/OL2 or SL1/2/TL2/1) NL2 vs. IL as an individual component ILa vs. ILa

Combining contrastive and interlanguage analysis 6.

13

Towards a comprehensive model of transfer: the DEE transfer model

The advantage of using authentic corpus data has been briefly demonstrated in Section 3, both for CA and CIA (see also Ellis & Barkhuizen 2005 on the value of naturally-occurring samples of learner language for investigating second language acquisition). The importance of considering outside variables has also been underlined by Granger and Jarvis, as well as other authors (e.g. Ellis 1994). This section will therefore focus on the six types of comparison borrowed from Granger’s and Jarvis’s models (see Table 1), and deal with them according to the primary purpose they serve, namely detection, explanation or evaluation.11 The resulting model, the DEE transfer model, does not pretend to be a theory of transfer, but it proposes a methodology which seeks to apprehend transfer in a way that takes several of its facets into account, thus striving for a comprehensive view of the phenomenon. 6.1

Detection of transfer

Two types of comparison serve the primary purpose of detecting the (potential) presence of transfer, viz. NL1 vs. IL and ILa vs. ILb (or more precisely ILa vs. ILb vs. … vs. ILz). Of these, the first one is an instance of CA and the second one, an instance of CIA, so that the detection of transfer relies on the combination of CA and CIA, as is also the case in Granger (1996) and Jarvis (2000). While ILa/ILb…ILz is common to both models, NL1/IL is only found in Jarvis’s model. The comparison of the learner’s interlanguage with his/her mother tongue (NL1 vs. IL) is essential to detect transfer. Only if one can establish a similarity between the learner’s behaviour in IL and NL1 does the presence of transfer seem plausible enough.12 The operationalisation of the NL1/IL comparison may involve frequency, as proposed by Paquot (2008), with the requirement that the frequency of the NL1 item be as close as possible to that of the IL item (cf. frequency of au contraire in native French and of on the contrary in learner English in Paquot’s [2006] study). The similarity may also be formal, in which case the method of reversed translation, which translates interlanguage back into the learner’s mother tongue, may be helpful.13 Thus, disregarding frequency, it is clear that in fact may be considered as the formal equivalent of French en fait (en corresponds to the preposition in and fait is the French noun for fact; both expressions have the same function of adding information to emphasise what has just been said). Similarly, from a formal point of view, on the contrary would be the nearest equivalent of French au contraire, with contraire being related to contrary (more, for example, than to contrast – cf. by contrast – which has another equivalent in French, contraste). Using the method of reversed translation, Borgatti (2006) discovered that 57% of the occurrences of the verb make in the interlanguage of Frenchspeaking advanced learners could be translated into French by means of the equivalent verb faire (e.g. make sacrifices = faire des sacrifices).14 Interestingly, he also found out that 83% of all the erroneous uses of make corresponded to French faire (e.g. *make a description, translatable into French by faire une

14

Gaëtanelle Gilquin

description), which strongly suggests that these errors are due to negative transfer. It should be noted that the method of reversed translation involves authentic learner data and that contrastive corpus data may be helpful in identifying translational equivalence (see below). However, the comparison itself does not rely on corpus data, but on the researcher’s capacity to translate learners’ interlanguage back into their mother tongue. In that, it stands as an exception to the generally corpus-based approach of the DEE transfer model. NL1/IL congruity is not enough to establish transfer with some degree of confidence. This is because a learner’s behaviour may be similar to his/her mother tongue, but also similar to the behaviour of learners from other mother tongue backgrounds, in which case the most likely explanation is not interlingual (i.e. due to L1 influence), but intralingual (i.e. due to the inherent difficulties of learning L2).15 The examination of the relationship between the learner’s interlanguage and the interlanguage of learners from other mother tongue backgrounds (ILa/ILb…ILz) should therefore come as a complement to the NL1/IL comparison. On the basis of a simple NL1/IL comparison, one may for instance come to the conclusion that the overuse of more and more by French-speaking learners is due to transfer, given that the frequency of this expression in their interlanguage is very similar to the frequency of its French counterpart, de plus en plus, in a comparable corpus of native French (CODIF, see Section 6.4). Examining data produced by learners from other mother tongue backgrounds, however, reveals that the overuse of more and more is a general characteristic of learner English, since out of the 16 L1 backgrounds represented in ICLE, Version 2 (Granger et al. forthcoming), 14 present such an overuse (statistically significant at the 0.001 level). While transfer cannot be totally excluded in the case of French-speaking learners, it seems as if intralingual factors prevail here. Since both Granger and Jarvis agree on the necessity of this type of comparison, we will not discuss it any longer. One may wonder why the comparison of the two languages involved has been operationalised through an NL1/IL analysis, rather than an OL1/OL2 or SL1/2/TL2/1 analysis as is the case in the Integrated Contrastive Model. In contrast with the latter two types of comparison, the NL1/IL analysis makes it possible to approach the learner’s interlanguage as an independent linguistic system. Comparing the two native languages in their original (or source and translated) forms and directly applying these results to the question of transfer (i.e. the relationship between the learner’s interlanguage and his/her mother tongue) would imply equating the interlanguage and the target language. However, several second language acquisition specialists have guarded against such an equation, underlining the dangers of what is called the “comparative fallacy”, i.e. “the failure […] to take the structure of the interlanguage on its own terms and a reliance on a comparison of the target language in order to study the structure of the interlanguage” (Bley-Vroman 1983: 3-4). It will be argued later that the criticism levelled by some against the native norm in interlanguage study is not necessarily justified, but it is still true that the learner’s interlanguage, not the target language, should be taken as a baseline for the comparison with NL1, since

Combining contrastive and interlanguage analysis

15

equivalence (or lack of equivalence) between NL1 and NL2 does not necessarily imply a similar relationship between NL1 and IL. Thus, there may be no correspondence in a parallel corpus between English sit on the bus and French être assis sur le bus, the latter describing someone sitting on the roof of a bus, but reversed translation would highlight the equivalence between the two expressions, since sur le bus is a literal translation of on the bus. It would also suggest that an English-speaking learner of French using the preposition sur in such cases is probably acting under the influence of English. The inappropriateness of bilingual corpora to deal with the relationship between the learner’s interlanguage and his/her mother tongue is also particularly evident in the case of non-target forms in the learner’s IL. The French component of ICLE (ICLE-FR) contains the expression make abstraction of, standing for disregard. An OL1/OL2 or SL1/2/TL2/1 comparison would fail to establish a correspondence between these two items, since make abstraction of does not exist in the NL2 system. Using the method of reversed translation, however, one may draw a parallel between make abstraction of and French faire abstraction de, thus showing that transfer is a likely explanation for the IL behaviour. What precedes should not be taken to mean that OL1/OL2 or SL1/2/TL2/1 comparisons have no role to play in the detection of transfer, or that NL2 is completely out of the picture because of the comparative fallacy. Comparable or parallel corpora may be necessary before the NL1/IL comparison proper, to identify the best candidate for translational equivalence. In the case of cognates (e.g. fait/fact, description/description), the equivalence is easily established. In some other cases, however, it may be more difficult, especially for non-native speakers of the learner’s L1, to determine the most probable equivalent. It is here that bilingual corpora may help, by showing in context what words fulfil the same function and express the same meaning in the learner’s mother tongue and in the target language.16 As Danchev (1991: 89) puts it, translation obviates “the familiar difficulty of establishing comparatibility”. (See, however, Section 7 for an important caveat.) As for NL2, it is often necessary as a baseline for comparing the frequencies of an item in IL and NL1. The difference in frequency between nevertheless in French learners’ IL and the equivalent néanmoins in a comparable corpus of native French may at first sight appear too great to be the result of transfer, with a relative frequency of 327 per million words for the former and 206 per million words for the latter. However, compared with the frequency of nevertheless in native English, which amounts to a mere 9 occurrences per million words, this difference between IL and NL1 becomes insignificant, thus making the transfer hypothesis more likely. Despite their possible participation in the detection of transfer, OL1/OL2, SL1/2/TL2/1 and NL2/IL comparisons have another, primary role to play in the DEE transfer model. The first two types of comparison make it possible to explain the presence, or otherwise, of L1 influence, whereas the last type of comparison is especially helpful in a more applied context, when evaluating whether an observed phenomenon of transfer should have consequences in

16

Gaëtanelle Gilquin

foreign language teaching. These two roles, explanation and evaluation, are described in the next two sections. 6.2

Explanation of transfer

The existence of a formal equivalent in the other language system is not necessarily sufficient to lead to transfer. Thus, although both Swedish and French have structures corresponding to causative make + infinitive (with få in Swedish and faire in French), transfer seems to be at work only in the case of Swedish learners (compare Altenberg 2002 and Gilquin 2000/2001). This difference in IL behaviour may be explained by resorting to the concept of “language distance”, that is “the degree of similarity between two languages” (Odlin 1989: 32). As noted by Ellis (1994: 327), “[t]here is substantial evidence to indicate that the actual distance between the native and the target languages acts as a constraint on transfer”. Equivalents which are identical in all respects between the two language systems are therefore more likely to be transferred than equivalents which present differences.17 The degree of cross-linguistic equivalence can be established on the basis of parallel corpus data (SL1/2 vs. TL2/1). A useful measure here is that of “mutual translatability” (or “mutual correspondence”), which refers to the frequency with which two items are translated as one another (on this concept, see Altenberg 1999). A value of 0% would indicate that the two items are never translated as one another, and a value of 100%, that the two items are always translated as one another. If we apply this measure to causative constructions in parallel corpus data of English and Swedish on the one hand (Altenberg 2002), and English and French on the other (Gilquin 2000/2001), a noticeable difference emerges. While causative make and its Swedish equivalent få exhibit a mutual translatability of 52%, in the case of causative make and its French equivalent faire, the value is under 14%.18 Considering all the English equivalents of causative faire in a sample of the parallel corpus PLECI19 (from source texts to translations and vice versa), for example, it turns out that a synthetic causative verb is the preferred option, with a total percentage of 56.7% (Gilquin 2000/2001: 105). This is illustrated by (1), where faire disparaître (‘make disappear’) corresponds to dispel. (1)

Aussitôt qu’elle avait établi son bon droit, elle s’efforçait de faire disparaître toute trace de ressentiment chez son adversaire. = Her own way once insisted upon and secured she was swift to dispel any smallest hint of unpleasantness. (PLECI fiction TF-OE)

The degree of cross-linguistic equivalence between two items can also be established by means of a comparable corpus (OL1 vs. OL2), preferably as a supplement to the parallel corpus analysis. While this approach does not make it possible to identify the word or construction corresponding to the target item in a specific context, since the texts are not translations of each other, it has the

Combining contrastive and interlanguage analysis

17

advantage of not being affected by any possible translation effects (“translationese”, see Gellerstam 1986). By carefully comparing the factors conditioning the use of two supposedly equivalent items in authentic original data, one may gain insight into the extent to which the two items overlap in terms of, say, lexical preferences, pragmatic functions or syntactic patterns. This, in turn, may help determine the “distance” between the two items and hence the likelihood of transfer. In the case of English and French causative constructions, the analysis of the comparable corpus data confirms the low degree of equivalence brought to light by the parallel corpus analysis (see Gilquin forthcoming). To give but one example, a sample of original texts from the PLECI corpus shows that causative constructions with faire are quite frequently causeeless, i.e. with no mention of the entity changed or influenced by the subject (this represents over one quarter of all the occurrences of causative faire), cf. (2). In English, by contrast, causeeless constructions with make are limited to the idiomatic expressions make believe and make do, which, incidentally, are not found in the PLECI corpus. (2)

Fais vérifier ses papiers. ‘Make check his papers’ Make someone check his papers/Have his papers checked. (PLECI fiction OF-TE)

In sum, OL1/OL2 and SL1/2/TL2/1 comparisons, by highlighting the degree of similarity between two items cross-linguistically, may provide an explanation for the presence or absence of transfer among learners. Two remarks are in order, here. The first one is that, although in what precedes the bilingual corpus data have been presented as a way of explaining transfer, thus corresponding to the diagnostic hypothesis of Granger’s model, they may also be used with a predictive goal, as a way of predicting where transfer is likely to occur or not. In other words, the OL1/OL2 and SL1/2/TL2/1 comparisons may precede or follow the detection of transfer, or indeed, as advocated by Granger, occur at different stages in the analysis. The second remark is that, in the same way as distance between two languages is a matter of degree (cf. degree of mutual translatability or degree of overlap between the conditioning factors of two items), the likelihood of transfer may be relatively high or low. One cannot determine a threshold below which transfer becomes impossible, so that both the explanatory (or diagnostic) power and the predictive power of the DEE transfer model are hypothetical, as in the Integrated Contrastive Model. 6.3

Evaluation of transfer

The two remaining types of comparison found in Granger (1996) and Jarvis (2000), NL2/IL and ILa/ILa, also have a role to play in a comprehensive model of transfer. However, while Granger and Jarvis use these comparisons as (partial) evidence for transfer, they will be assumed here to be mainly (though by no

18

Gaëtanelle Gilquin

means only) useful for pedagogical purposes, as a way of evaluating whether the presence of transfer should affect foreign language teaching (FLT) or not.20 Once a phenomenon of transfer has been detected with a certain degree of likelihood, a comparison of the interlanguage and the target native language (NL2) is necessary to distinguish between positive and negative transfer. Consider the expressions according to me and according to X (X being another entity than the subject) as used by French-speaking learners of English. Applying the criteria established in Section 6.1, one can detect possible traces of transfer in both cases, from French selon moi and selon X respectively. However, it takes an examination (or at least knowledge) of NL2 to reveal that according to me, unlike according to X, is not acceptable in native English. The presence of according to me in French-speaking learners’ interlanguage may therefore be described as a case of negative transfer, that is, a case where learners produce a non-target form in the interlanguage as a result of negative influence of their mother tongue, whereas the presence of according to X could be a case of positive transfer, where learners produce a target form in the interlanguage as a result of positive influence of their mother tongue. Note that the comparison with native data is not only necessary to distinguish between cases of positive and negative transfer of form, but also of frequency.21 By transferring the frequency of an item from NL1 to IL, a learner may “hit the right amount” (positive transfer), or s/he may overuse or underuse the item (negative transfer). The apparent positive transfer of form of according to X among French-speaking learners of English is accompanied by what looks like a positive transfer of frequency, with no statistically significant difference in frequency between non-native and native English. Negative transfer, on the other hand, seems to be at work in French-speaking learners’ overuse of on the contrary (see above). Given the time constraints that exist in the FLT classroom, it is best to focus on instances of negative transfer (e.g. the use of according to me or the overuse of on the contrary), showing learners how to avoid it (negative feedback), and let positive transfer (e.g. the use of according to X) make its way into the interlanguage. This, of course, implies that one takes native language as a target for foreign language learners, which may sound like a debatable issue in view of the “comparative fallacy” hypothesis (Bley-Vroman 1983) and the movement of English as a Lingua Franca (or English as an International Language, see e.g. Jenkins 2000, Seidlhofer 2001, Mauranen 2003), which both stigmatise the NL norm. However, I would like to argue that the native language still has a proper place (and an important one) in FLT contexts.22 In a survey conducted among some 200 students, Mukherjee (2005) found that 80% of the subjects chose a national variety of native English as the norm they wanted to approximate to (47% for British English, 30.5% for American English and 2.5% for other national varieties). By contrast, less than 17% of the subjects accepted as their target norm the English used by competent non-native speakers in international contexts. Although native English may not be the norm aimed at by all learners (many of the students who took part in Mukherjee’s survey were future English teachers; the results might have been different with, say, business

Combining contrastive and interlanguage analysis

19

students), it is still a target norm for some of them. In addition, especially in the early stages of FLT, students may not know what type of norm they will need in their future interactions (e.g. only comprehensible to non-native speakers, or also comprehensible to native speakers) and teachers may have to address mixed groups of learners, some of them aiming at a native norm and others happy to speak English as competent non-native speakers. Native language, therefore, appears as a reasonable common target in FLT education, and one which is relatively well-defined.23 Like NL2/IL comparisons, ILa/ILa comparisons will be taken here to be mainly useful for evaluative purposes. The fact that an item is used by only a few learners within an L1-group does not perforce exclude the possibility of transfer. In fact, it has been regularly pointed out in the literature on transfer that “the extent to which the L1 influences L2 development can vary strongly from individual to individual” (Kellerman 1995: 133).24 Going back to the expression make abstraction of (see Section 6.1), there is no doubt that it is the result of transfer from French, although it occurs only once in ICLE-FR. Given its low frequency, however, it is less interesting in an FLT context than an expression such as according to me, which is found in ICLE-FR with a relative frequency of almost 10 occurrences per 100,000 words. By establishing the probability of a particular transfer-related phenomenon, ILa/ILa comparisons, therefore, provide relevant information to go through the inevitable process of material selection in the FLT curriculum. The model of transfer proposed here, with its six types of comparison and three main stages, is illustrated in Figure 2. In what follows, the model will be applied to the use of even if by French-speaking learners of English. 6.4

Case study: the use of even if by French-speaking learners

This section presents a case study using the DEE transfer model. More precisely, we will go through the stages of detection, explanation and evaluation in order to study any phenomenon of transfer that might be attached to French-speaking learners’ use of the expression even if. The learner corpus data on which the study is based come from ICLE Version 2 (Granger et al. forthcoming). Like all the data from this corpus (see Section 5), the selected data keep three variables constant, namely type of learner (foreign language learner rather than second language learner), stage of learning (advanced level) and text type (essay writing). A number of additional variables were controlled for when selecting the samples: mother tongue background (French), text type (argumentative essays) and conditions of writing (untimed and with reference tools). Finally, only those samples that contained at least 100,000 words after applying the above criteria were kept in the final analysis.25 Table 3 presents the word counts of the remaining nine components. In addition, comparable corpora of native French and native English were used in the study – 208,419 words from the Corpus de Dissertations Françaises (CODIF) and 328,406 words from the Louvain Corpus of Native English Essays (LOCNESS) – as well as French-English corpus data from PLECI.26

20

Gaëtanelle Gilquin

Figure 2. The DEE transfer model

Table 3. Word counts of ICLE samples ICLE component Czech (ICLE-CZ) Dutch (ICLE-DU) Finnish (ICLE-FI) French (ICLE-FR) German (ICLE-GE) Polish (ICLE-PO) Russian (ICLE-RU) Spanish (ICLE-SP) Turkish (ICLE-TU) TOTAL

Number of words 131,435 163,908 126,599 137,431 110,890 141,535 168,046 99,905 105,006 1,184,755

Detection of transfer according to the DEE transfer model involves two types of comparison, viz. NL1/IL and ILa/ILb…ILz. A reversed translation of even if points to même si as a formal equivalent, both as a multi-word unit and as individual words (even = même; if = si). This equivalence is confirmed by the corpus data. The first type of comparison also reveals a striking similarity in

Combining contrastive and interlanguage analysis

21

frequency between even if in ICLE-FR and même si in the corpus of native French essays, with a relative frequency of 458.41 and 489.40 occurrences per million words respectively.27 The inter-L1-group comparison (ILa/ILb…ILz), the results of which are displayed in Table 4, shows that, despite the variability among the L1 groups, French-speaking learners’ behaviour differs markedly from that of the other learners. ICLE-FR emerges as the ICLE component which has the highest relative frequency of even if, and a Ryan procedure (REGWQ) indicates that French-speaking learners distinguish themselves significantly from all the other learner populations in their use of even if.28 These two comparisons suggest that there might be a phenomenon of transfer from French même si to French-speaking learners’ IL even if.29 Table 4. Frequency of even if in ICLE samples ICLE component ICLE-FR ICLE-FI ICLE-GE ICLE-PO ICLE-CZ ICLE-RU ICLE-TU ICLE-DU ICLE-SP

Absolute frequency 63 35 26 30 26 26 12 17 7

Relative frequency per million words 458.41 276.46 234.47 211.96 197.82 154.72 114.28 103.72 70.07

Going on to the next stage, that of explanation, we can use bilingual corpus data to examine the degree of equivalence between English even if and French même si. Parallel corpus data from PLECI reveal a relatively high degree of correspondence between the two expressions. Out of 61 occurrences of even if, 40 correspond to même si in French, e.g. (3). (3)

Dès lors, on comprend pourquoi la solution négociée arrangerait les affaires de tout le monde, la France et les Nations unies devant se targuer d’avoir provisoirement marqué un point, même si Washington, inquiet de la triple entente Kinshasa-Tripoli-Khartoum, n’a pas encore dit son dernier mot. = Clearly, a negotiated solution would be in everybody’s interest. It would enable both France and the UN to score a point, at least temporarily, even if Washington, worried by the triple alliance between Kinshasa, Tripoli and Khartoum, may not have said its last word. (PLECI non-fiction OF-TE)

However, it also appears from the SL1/2/TL2/1 and OL1/OL2 comparisons that even if and même si do not always fulfil the same function. In particular, while even if

22

Gaëtanelle Gilquin

is exclusively used to introduce a condition, même si can also introduce a concession, as in (4), where même si corresponds to English even though. (4)

Or, dans les sociétés africaines traditionnelles, on demande toujours l’avis des femmes avant de prendre une décision, même si celles-ci ne parlent jamais en public. = But in traditional African societies the women are always asked for their opinion before a decision is taken, even though they never speak in public. (PLECI non-fiction OF-TE)

Although the equivalence between even if and même si is only partial, the high degree of correspondence between the two expressions and the inclusion of even if in the functional field of même si suggest that there is little distance between the expressions and thus serve to explain French-speaking learners’ readiness to transfer the form, function and frequency of même si into their IL (see next paragraph on the transfer of function). In terms of evaluation, finally, the comparison between NL2 and IL underlines the negative nature of the transfer, and hence the relevance such material may have in an FLT context. This negative transfer manifests itself in two ways. First, it takes the form of an overuse of even if by French-speaking learners, as appears from Table 5 (Ȥ² = 60.24, statistically significant at the 0.001 level). Second, it results in a number of incorrect uses of the expression, in contexts where a concession, rather than a condition, is introduced. This type of error is illustrated by (5), where a native speaker would have used even though instead of even if.30 As noted earlier, même si can fulfil a concessive function, which makes the influence of L1 quite clear in such cases.31 These two types of negative transfer stress the importance of drawing French-speaking learners’ attention to the native-like use of even if. Table 5. Frequency of even if in learner English (ICLE-FR) and native English Language variety Learner English (ICLE-FR) Native English (5)

Absolute frequency 63 33

Relative frequency per million words 458.41 100.49

The problem is that even if such an international principle for equality has been established, some countries still keep on violating it (mostly in countries where religion prevails over international rights). (ICLE-FR)

This point is reinforced by the intra-L1-group comparison (ILa/ILa), which shows that the use of even if is not limited to a few learners only. Querying each essay from ICLE-FR separately, one notices that the expression occurs in 22.37% of them (51 essays out of 228). While this proportion may seem relatively low, it

Combining contrastive and interlanguage analysis

23

roughly corresponds to the proportion with which frequent words such as difficult, look, feel or idea occur, as emphasised by Paquot & Fairon (2006). This brief case study has illustrated the operationalisation of the DEE transfer model, as described in Sections 6.1 to 6.3. While this model makes it possible to shed light on some important aspects of the phenomenon of transfer, it does not pretend to answer all questions related to the issue. In the next section, we examine some possible limitations of the model. 7.

Some limitations of the DEE transfer model

In what follows, three possible limitations of the DEE transfer model are briefly discussed, one having to do with the CA component, the other with the CIA component and the last one with the model in general. The discussion, at the same time, opens up possibilities for further improving the model, in an attempt to achieve an even better understanding of transfer. A first and, in my view, major limitation of the DEE transfer model as it stands is that, in its investigation of the relation between L1 and L2, it only takes account of the objective distance, as evidenced in parallel or comparable corpora. Since transfer is a phenomenon that takes place first and foremost in the learner’s cognitive system, however, what really counts is not the objective distance between two items, but the distance as it is perceived by the learner, that is, the learner’s psychotypology (cf. Kellerman 1983).32 If a learner perceives two expressions as similar, this may lead him/her to transfer the characteristics of the L1 expression to the L2 expression, no matter whether the two expressions actually correspond to each other in authentic language data. On the other hand, if two expressions are not connected with each other in the learner’s mental lexicon, their objective similarity is very unlikely to trigger transfer-related phenomena. While objective distance may be an indication of subjective distance (Gilquin 2000/2001), the ideal would be to probe directly into learners’ perceptions. This could be done by means of elicitation tests, asking learners to provide the translational equivalent of a word or expression presented to them. The results of such tests might provide an explanation for some cases of transfer not accounted for by bilingual corpus data. Thus, the use of indeed by French-speaking learners presents all the signs of transfer. Its frequency in ICLE-FR is very different from its frequency in all the other ICLE components, but very similar to the frequency of its intuitive equivalent en effet in a comparable corpus of native French. It is also significantly overused in ICLE-FR (unlike the other ICLE components, which display an underuse or no significant difference) and appears in a large proportion of essays, which makes it particularly relevant to FLT. However, what seems to be lacking is an explanation for this transfer, since parallel corpus data from PLECI reveal a very low degree of correspondence, with only 4 out of 48 occurrences of indeed translated as en effet,33 and hence, a great (objective) distance between the two expressions. A test carried out among 23 Frenchspeaking advanced learners of English, however, shows that en effet is perceived

24

Gaëtanelle Gilquin

as the direct equivalent of indeed by most of them (20 learners).34 The subjective similarity between indeed and en effet, therefore, explains why transfer is possible, although the expressions are, in reality, quite different. Inaccurate perception may also be at work in the case of so-called false friends. A learner may establish a link between, say, actually and actuellement (French for ‘at the moment’) because of the formal similarity between the words, but no bilingual corpus, or, indeed, competent “reversed” translator would suggest such a link. Elicitation data, because they offer more direct access to the learner’s mental lexicon, may explain phenomena unaccounted for by the corpus data and would, for this reason, deserve a place in the DEE transfer model. Turning to the CIA component, it has been argued earlier that, despite recent criticism, a native norm is still useful in the study of transfer. However, what this norm should be is still very much an open question. In the case study of Section 6.4, the English native data were extracted from LOCNESS, a corpus of essays written by American and British students. American English and British English, however, display some differences, as is well known, and it might not be such a good idea to mix the two varieties. If only one variety is chosen, on the other hand, which one should it be? British English, American English or yet another national variety of English? Han (2004: 232) argues that the norm against which the interlanguage is examined should be “what the subjects have been exposed to”, but for foreign language learners, exposure might be limited to English as spoken by a non-native teacher. Some may also argue that a corpus of expert writing is what is needed, not a corpus of novice writing, since learners (or at least, some learners) ultimately aim to become expert writers. The problem is that the choice of the norm may affect the results of the analysis. An example of this is the word chance, which has a direct (though partly deceptive) equivalent in French, chance. Compared with data from LOCNESS, ICLE-FR reveals no significant difference in frequency (Ȥ² = 1.51). Compared with a corpus of expert academic writing, however, the difference becomes significant (Ȥ² = 22.64, p<0.001), with the French-speaking learners overusing the word. Consequently, if transfer could be demonstrated to occur with chance, a comparison with a corpus of novice writing would suggest that it is unnecessary to draw learners’ attention to the word chance (or, at least, to its frequency), whereas a comparison with a corpus of expert writing would speak in favour of raising this issue in the FLT classroom. As long as the question of the norm has not been resolved, therefore, the DEE transfer model may make inappropriate recommendations in terms of FLT (although it remains valid for detecting and explaining transfer). Another possible problem with the model proposed here is that it may give the impression that, once the influence of L1 has been highlighted, no other explanation is to be sought to account for the observed IL behaviour. Transfer, however, does not exclude the possibility of other factors acting in conjunction with L1 influence. Consider the overuse of let’s/let us by French-speaking learners, which Paquot & Fairon (2006) have shown to be transfer-related. An examination of the “Essay Writing” section of the French-English bilingual dictionary Collins Robert (Atkins et al. 1998) reveals the presence of a large

Combining contrastive and interlanguage analysis

25

number of expressions with let’s/let us, presented as translational equivalents of plural imperatives in French, e.g. let us now consider… (for French considérons maintenant…), let us take … as a starting point (for French prenons comme point de départ…) or let us not forget that… (for French n’oublions pas que…). This indicates that in the case of the imperative, the influence of the mother tongue may be reinforced by teaching-induced effects, and more precisely by the unsuitability of some pedagogical materials and/or reference tools (Gilquin & Paquot 2008). Ideally, the DEE transfer model should therefore be combined with the analysis of other possible factors affecting the IL behaviour, so as to determine the extent to which transfer is responsible for this behaviour. 8.

Conclusion

While the existence and the pervasiveness of transfer in second language acquisition is now widely acknowledged, one must recognise that transfer is still something of a mystery to linguists. In this article, the DEE transfer model has been proposed which, by integrating methods from Granger (1996) and Jarvis (2000), aims to detect cases of transfer, explain them and evaluate their pedagogical relevance, hence making it possible to come to a better understanding of the phenomenon. This model combines contrastive analysis and interlanguage analysis, both of which have been shown to be necessary for a sound investigation of transfer, and relies on corpus data, the guarantee for the authenticity of the observations. It would be naive, however, to think that the model of transfer presented here provides definite and definitive answers to all transfer-related questions. Transfer is a slippery phenomenon that does not lend itself easily to apprehension. It sometimes comes up when one does not anticipate it, and can fail to appear where one would expect it, thus escaping the attention of even the best of observers. But with enough patience and by putting in enough effort and rigour in the task, it is to be hoped that one day transfer will reveal all of its secrets. This model offers a key to unlock some of them. Notes 1

I gratefully acknowledge the financial support of the Belgian National Fund for Scientific Research (F.N.R.S.). For helpful comments on an earlier version of this paper, I wish to thank (in alphabetical order) María Belén Díez-Bedmar, Sylviane Granger, Claire Hugon, Scott Jarvis, Szilvia Papp and Magali Paquot.

2

Sharwood Smith & Kellerman (1986: 1-2) prefer the term “crosslinguistic influence” to “transfer” (or “interference”). One of their arguments is that the term “transfer” has traditionally been reserved for the influence of L1 on L2 or L2 on L3, not the other way round. However, following Pavlenko

26

Gaëtanelle Gilquin & Jarvis (2002), “transfer” will be understood here in a bidirectional sense. The term “interference” will be avoided because of its negative connotations (see Sharwood Smith & Kellerman 1986: 1).

3

“Erroneous usage” is to be taken here both in a qualitative sense, referring to misuse (i.e. errors in the strict sense), and in a quantitative sense, covering over- and underuse (see later on these notions). The same is true of “correct usage”, which can be correct both from a qualitative and a quantitative perspective.

4

The three same varieties of language are examined by Borin & Prütz (2004) and Tono (2004), but none of them offers the range of comparisons proposed here.

5

See Granger (1996) and Section 3 of this paper on contrastive interlanguage analysis. Note that, while the third type of comparison, NL1/IL, includes a non-native variety of language, it is considered a case of contrastive analysis because it involves distinct languages (e.g. native French and non-native English), and not two varieties of one and the same language as is the case in contrastive interlanguage analysis.

6

It should be borne in mind, however, that the claims of the model might be slightly different for, say, phonological or pragmatic phenomena, or for languages which are radically different from European languages in these aspects.

7

See Gilquin (2000/2001) for a brief history of CA and CIA, as well as some bibliographical references on the subject.

8

“Errors” in a broad sense, see endnote 3.

9

Other influences are not excluded. It seems plausible to assume, following Paquot (2006), that the overuse of on the contrary is a general characteristic of interlanguage, given that all ICLE components exhibit this tendency (although it is more marked in the French component). This overuse may also be teaching-induced since, as Lake (2004) notes, many textbooks make no distinction between on the contrary and on the other hand, listing both discourse markers under the functional heading of “contrast”, which leads to potential confusion in the learner’s mind.

10

It should be emphasised, however, that the stage of learning was determined on the basis of external criteria (students in their third or fourth year of university study) and that a rater’s assessment of some 200 texts from ICLE based on internal criteria revealed proficiency levels ranging from B2 to C2 in the Common European Framework (but with a majority of C1), see Granger & Thewissen (2005).

Combining contrastive and interlanguage analysis

27

11

Detection roughly corresponds to Jarvis’s (2000) idea of “identification”. Explanation may be related to the “diagnostic” hypothesis of Granger’s (1996) ICM. Evaluation is not explicitly present in either model.

12

Although Jarvis (2000: 258) cites a case where transfer could be present despite a lack of correspondence between NL1 and IL, he acknowledges that NL1-IL performance congruity “may, in fact, be the strongest type of evidence for L1 influence”.

13

While reversed translation as such is not used by Jarvis (2000), his method of comparing NL1 and IL is similar in that it examines congruence between the words used in NL1 and in IL to refer to the same denotatum.

14

Borgatti (2006) uses the term “back-translation” to describe this technique. Here, however, this term will be avoided for its potential confusion with the concept of back-translation as defined by Ivir (1983), which refers to the observation, in a bilingual corpus, of translated words and their equivalents in the source language.

15

That is, unless all L1s share the feature under study, in which case similarity in the learners’ behaviour is to be expected in case of transfer.

16

This does not exclude the possibility that a word has several equivalents, all of which may result in transfer. Thus, English perhaps and maybe correspond to just one word in French, peut-être. While maybe is a closer equivalent from an etymological point of view (peut = may and être = be), perhaps is more similar phonologically. There is therefore no reason to favour one word over the other as an equivalent of peut-être, and hence one could envisage that an English-speaking learner of French is influenced both by perhaps and maybe when producing peut-être in his/her interlanguage.

17

The opposite view, namely that identity between L1 and L2 leads to avoidance, has also been expressed (Kellerman [2000] refers to this phenomenon as “homoiophobia”, the fear of resemblance). However, this seems to be true of certain items only (typically marked forms such as idioms, cf. Kellerman 1978) and to be limited to intermediate learners (Sjöholm 1998). Since the illustrations provided here are based on advanced learner language and non-idiomatic items, it is reasonably safe to assume that language identity is more likely to lead to transfer than language distance.

18

This difference in mutual translatability may be argued to originate from the structural difference between English and Swedish on the one hand, and French on the other. In the former case, the normal position of the non-finite complement is after the causee (cf. He made her laugh, Han fick henne att skratta), whereas in French the causative verb and its nonfinite complement are in principle inseparable (cf. Je l’ai fait rire).

28

Gaëtanelle Gilquin However, the Swedish causative construction and its English counterpart also differ in two important respects, namely få, which is the verb used in the Swedish construction, corresponds to English ‘get’, not ‘make’ (the normal equivalent of ‘make’ in Swedish, göra, cannot be used as a causative verb followed by a non-finite verb; see Viberg 2006: 251-252), and the infinitival complement in Swedish has to be preceded by the marker att, ‘to’.

19

The PLECI (Poitiers-Louvain Échange de Corpus Informatisés) corpus is an English-French bilingual corpus of journalistic and fictional texts (see Gilquin 2000/2001).

20

The context here is that of FLT for learners from the same mother tongue background. In the case of teachers having to address students with different mother tongues or teaching materials designed for several learner populations, it is difficult to take transfer into account, except, perhaps, if the learners’ mother tongues belong to the same language family.

21

See Paquot (2008) on the notions of transfer of frequency and transfer of form.

22

See also Tenfjord et al. (2006), who argue that the tagging of errors in a learner corpus, which entails a native norm, does not necessarily lead to comparative fallacy.

23

By contrast, the concept of English as a Lingua Franca is still relatively vague and to date no one has yet been able to compile a complete repertoire of “non-core” features of English which would represent “no threat to intelligibility for the (non-native) receiver” (Jenkins 1998: 123). Mollin (2006), in fact, shows that what is generally referred to as “EuroEnglish” is too heterogeneous to be considered a “variety of English”.

24

It may also vary in one and the same learner according to external circumstances. Nickel (1989: 298) notes that stress, for example, is a factor that tends to increase transfer.

25

The only exception to this is the Spanish component, which was included since its size is very close to the limit (99,905 words).

26

Use was made of the Poitiers web interface (http://cabal.rezo.net, last accessed on January 13, 2007), which contains about 350,000 words of original newspaper articles in English or French and their translations in the other language. While the use of a journalistic corpus may be questionable, given the nature of the other corpora (argumentative essays), it should be emphasised that bilingual corpora are not available for all genres, which severely limits the choice. Bilingual corpora of scientific articles do exist (cf. Kraif & Tutin 2006), but they are still relatively rare (and small), and not readily available. Newspaper articles, which are more widely available, have been chosen here as the next best solution.

Combining contrastive and interlanguage analysis

29

27

To be compared with 100.49 in native English (see later).

28

I thank Magali Paquot and Yves Bestgen for their help with the statistical analysis.

29

While the Italian component of ICLE was excluded because of its small size, Prat Zagrebelsky (2001) has shown that Italian learners also have a tendency to overuse even if, a result which she explains by L1 influence.

30

Interestingly, the overuse of even if by French-speaking learners goes hand in hand with an underuse of even though (significant at the 0.001 level), which is probably to be linked to the learners’ common confusion between the two expressions.

31

The same phenomenon seems to be at work in Italian learner English, since the Italian conjunction anche se can be used to express the meanings of both even if and even though (Prat Zagrebelsky 2001).

32

For an example of a study showing the importance of psychotypology, see Singleton (1987).

33

An equivalent such as même, in fact, is more common than en effet in the corpus data (6 occurrences).

34

Two learners provided en fait (‘in fact’) as a translational equivalent, and one, the rather formal adverb certes (‘admittedly’). Interestingly, both equivalents occur twice each in the parallel corpus data.

References Altenberg, B. (1999), ‘Adverbial connectors in English and Swedish: semantic and lexical correspondences’, in: H. Hasselgård and S. Oksefjell (eds) Out of Corpora. Studies in Honour of Stig Johansson. Amsterdam & Atlanta: Rodopi. 249-268. Altenberg, B. (2002), ‘Causative constructions in English and Swedish: A corpusbased contrastive study’, in: B. Altenberg and S. Granger (eds) Lexis in Contrast. Corpus-based Approaches. Amsterdam & Philadelphia: John Benjamins Publishing Company. 97-116. Atkins, B.T., A. Duval, R.C. Milne, P.-H. Cousin, H.M.A. Lewis, L.A. Sinclair, R.O. Birks and M.-N. Lamy (eds) (1998), Collins Robert Unabridged. French-English, English-French Dictionary. Fifth edition. New York: HarperCollins. Benson, C. (2002), ‘Transfer/cross-linguistic influence’, ELT Journal, 56(1): 6870. Bley-Vroman, R. (1983), ‘The comparative fallacy in interlanguage studies: the case of systematicity’, Language Learning. A Journal of Applied Linguistics, 33(1): 1-17.

30

Gaëtanelle Gilquin

Borgatti, E. (2006), The use of the verbs ‘make’ and ‘do’ by French- and Dutchspeaking EFL learners: A corpus-based study. Unpublished MA dissertation, Université catholique de Louvain. Borin, L. and K. Prütz (2004), ‘New wine in old skins? A corpus investigation of L1 syntactic transfer in learner language’, in: G. Aston, S. Bernardini and D. Stewart (eds) Corpora and Language Learners. Amsterdam & Philadelphia: John Benjamins Publishing Company. 67-87. Danchev, A. (1991), ‘Some notes on a Bulgarian model of expanded contrastive analysis’, in: V. Ivir and D. Kalogjera (eds) Languages in Contact and Contrast: Essays in Contact Linguistics. Berlin: Mouton de Gruyter. 87101. Dechert, H. and M. Raupach (eds) (1989), Transfer in Language Production. Norwood, NJ: Ablex. Ellis, R. (1994), The Study of Second Language Acquisition. Oxford: Oxford University Press. Ellis, R. and G. Barkhuizen (2005), Analysing Learner Language. Oxford: Oxford University Press. Fivet, O. (1995), A contrastive study of the passive in English and French journalese. Unpublished MA dissertation, Université catholique de Louvain. Gellerstam, M. (1986), ‘Translationese in Swedish novels translated from English’, in: L. Wollin and H. Lindquist (eds) Translation Studies in Scandinavia. Lund: CWK Gleerup. 88-95. Gilquin, G. (2000/2001), ‘The Integrated Contrastive Model. Spicing up your data’, Languages in Contrast, 3(1): 95-123. Gilquin, G. (forthcoming), ‘Causative make and faire: A case of mismatch’, to appear in: M.L.A. Gómez González, J.L. Mackenzie and E.M. González Álvarez (eds) Current Trends in Contrastive Linguistics. Amsterdam & Philadelphia: John Benjamins Publishing Company. Gilquin, G. and M. Paquot (2008), ‘Too chatty: Learner academic writing and register variation’, English Text Construction, 1(1): 41-61. Granger, S. (1996), ‘From CA to CIA and back: an integrated approach to computerized bilingual and learner corpora’, in: K. Aijmer, B. Altenberg and M. Johansson (eds) Languages in Contrast. Papers from a Symposium on Text-based Cross-linguistic Studies. Lund 4-5 March 1994. Lund: Lund University Press. 37-51. Granger, S. (1997), ‘Automated retrieval of passives from native and learner corpora: precision and recall’, Journal of English Linguistics, 25(4): 365374. Granger, S. (1998), ‘Interlanguages in contrast: translationese vs learnerese’. Paper presented at the conference Languages in Contrast. Information Structure in Parallel Texts, University of Oslo, 30-31 October 1998. Granger, S. (2003), ‘The International Corpus of Learner English. A new resource for foreign language learning and teaching and second language acquisition research’, TESOL Quarterly, 37(3): 538-546.

Combining contrastive and interlanguage analysis

31

Granger, S., E. Dagneaux and F. Meunier (eds) (2002), The International Corpus of Learner English. Handbook and CD-ROM. Louvain-la-Neuve: Presses Universitaires de Louvain. Granger, S., E. Dagneaux, F. Meunier and M. Paquot (eds) (forthcoming), The International Corpus of Learner English. Handbook and CD-ROM. Version 2. Louvain-la-Neuve: Presses Universitaires de Louvain. Granger, S. and J. Thewissen (2005), ‘Towards a reconciliation of a “Can Do” and “Can’t Do” approach to language assessment’. Paper presented at the Second Annual Conference of EALTA (European Association of Language Testing and Assessment), Voss, Norway, 2-5 June 2005. Hammarberg, B. (2001), ‘Roles of L1 and L2 in L3 production and acquisition’, in: J. Cenoz, B. Hufeisen and U. Jessner (eds) Cross-linguistic Influence in Third Language Acquisition: Psycholinguistic Perspectives. Clevedon: Multilingual Matters. 21-41. Han, Z. (2004), ‘Fossilization: five central issues’, International Journal of Applied Linguistics, 14(2): 212-242. Ivir, V. (1983), ‘A translation-based model of contrastive analysis’, Jyväskylä Cross-Language Studies, 9: 171-178. James, C. (1980), Contrastive Analysis. London: Longman. Jarvis, S. (2000), ‘Methodological rigor in the study of transfer: identifying L1 influence in the interlanguage lexicon’, Language Learning, 50(2): 245309. Jenkins, J. (1998), ‘Which pronunciation norms and models for English as an International Language?’, ELT Journal, 52(2): 119-126. Jenkins, J. (2000), The Phonology of English as an International Language. Oxford: Oxford University Press. Kamimoto, T., A. Shimura and E. Kellerman (1992), ‘A second language classic reconsidered – the case of Schachter’s avoidance’, Second Language Research, 8(3): 251-277. Kellerman, E. (1978), ‘Giving learners a break: Native language intuitions as a source of predictions about transferability’, Working Papers in Bilingualism, 15: 59-92. Kellerman, E. (1983), ‘Now you see it, now you don’t’, in: S. Gass and L. Selinker (eds) Language Transfer in Language Learning. Rowley, MA: Newbury House. 112-134. Kellerman, E. (1995), ‘Crosslinguistic influence: transfer to nowhere?’, Annual Review of Applied Linguistics, 15: 125-150. Kellerman, E. (2000), ‘Lo que la fruta puede decirnos acerca de la transferencia léxico-sémantica: una dimensión no estructural de las percepciones que tiene el apprendiz sobre las relaciones lingüisticas’ [What fruit can tell us about lexicosemantic transfer: A non-structural dimension to learners’ perceptions of linguistic relations], in: C. Muñoz (ed.) Segundas lenguas. Adquisición en la aula. Barcelona: Ariel. 21-37.

32

Gaëtanelle Gilquin

Kohn, K. (1986), ‘The analysis of transfer’, in: E. Kellerman and M. Sharwood Smith (eds) Crosslinguistic Influence in Second Language Acquisition. New York: Pergamon Press. 21-34. Kraif, O. and A. Tutin (2006), ‘Looking for semi-frozen expressions using an aligned corpus: an application for academic writing for EFL learners’. Paper presented at the 7th Conference of Teaching and Language Corpora (TaLC), Université Paris 7 – Denis Diderot, 1-4 July 2006. Lado, R. (1957), Linguistics across Cultures. Ann Arbor: University of Michigan Press. Lake, J. (2004), ‘Using “on the contrary”: the conceptual problems for EAP students’, ELT Journal, 58(2): 137-144. Mauranen, A. (2003), ‘The corpus of English as lingua franca in academic settings’, TESOL Quarterly, 37(3): 513-527. Mollin, S. (2006), Euro-English. Assessing Variety Status. Tübingen: Gunter Narr Verlag. Mukherjee, J. (2005), ‘The native speaker is alive and kicking: linguistic and language-pedagogical perspectives’, Anglistik, 16(2): 7-23. Nickel, G. (1989), ‘Some controversies in present-day error analysis: “contrastive” vs. “non-contrastive” errors’, International Review of Applied Linguistics in Language Teaching, 27(4): 293-305. Odlin, T. (1989), Language Transfer. Cambridge: Cambridge University Press. Paquot, M. (2006), ‘EAP phraseology in the International Corpus of Learner English: the role of transfer’. Paper presented at the Congreso Internacional de Fraseología Y Paremiología, Universidad Santiago de Compostela, 19-22 September 2006. Paquot, M. (2008), ‘Exemplification in learner writing: a cross-linguistic perspective’, in: F. Meunier and S. Granger (eds) Phraseology in Foreign Language Learning and Teaching. Amsterdam & Philadelphia: John Benjamins Publishing Company. 101-119. Paquot, M. and C. Fairon (2006), ‘Investigating L1-induced learner variability: Using the Web as a source of L1 comparable data’. Paper presented at the International Computer Archive of Modern and Medieval English (ICAME) Conference (Variation, Contacts and Change), University of Helsinki, 24-28 May 2006. Pavlenko, A. and S. Jarvis (2002), ‘Bidirectional transfer’, Applied Linguistics, 23(2): 190-214. Petch-Tyson, S. (1998), ‘Writer/reader visibility in EFL written discourse’, in: S. Granger (ed.) Learner English on Computer. New York: Addison-Wesley Longman. 107-118. Prat Zagrebelsky, M.T. (2001), ‘Even if or even though? A corpus-based investigation of the Italian-ICLE subcorpus’. Poster presented at the International Computer Archive of Modern and Medieval English (ICAME) Conference (Future Challenges in Corpus Linguistics), Université catholique de Louvain, 16-20 May 2001.

Combining contrastive and interlanguage analysis

33

Ringbom, H. (1995), ‘Error analysis’, in: J. Verschueren, J.-O. Östman and J. Blommaert (eds) Handbook of Pragmatics. Amsterdam & Philadelphia: John Benjamins Publishing Company. 581-583. Schachter, J. (1974), ‘An error in error analysis’, Language Learning, 24(2): 205214. Seidlhofer, B. (2001), ‘Closing a conceptual gap: the case for a description of English as a lingua franca’, International Journal of Applied Linguistics, 11: 133-158. Selinker, L. (1972), ‘Interlanguage’, International Review of Applied Linguistics in Language Teaching, 10(3): 209-231. Selinker, L. (1989), ‘CA/EA/IL: The earliest experimental record’, International Review of Applied Linguistics in Language Teaching, 27(4): 267-291. Sharwood Smith, M. and E. Kellerman (1986), ‘Crosslinguistic influence in Second Language Acquisition: An introduction’, in: E. Kellerman and M. Sharwood Smith (eds) Crosslinguistic Influence in Second Language Acquisition. New York: Pergamon Press. 1-9. Singleton, D. (1987), ‘Mother and other tongue influence on learner French. A case study’, Studies in Second Language Acquisition, 9: 327-345. Sjöholm, K. (1998), ‘A reappraisal of the role of cross-linguistic and environmental factors in lexical L2 acquisition’, in: K. Haastrup and Å. Viberg (eds) Perspectives on Lexical Acquisition in a Second Language. Lund: Lund University Press. 209-236. Tenfjord, K., J.E. Hagen and H. Johansen (2006), ‘The “hows” and “whys” of coding categories in a learner corpus (or “How and why an error-tagged learner corpus is not ‘ipso facto’ one big comparative fallacy”)’, Rivista di Psicolinguistica Applicata, 6(3): 93-108. Tisthoud, N. (1992), Conceptual and lexical gaps: A problem of translation. Unpublished MA dissertation, Université catholique de Louvain. Tono, Y. (2004), ‘Multiple comparisons of IL, L1 and TL corpora: The case of L2 acquisition of verb subcategorization patterns by Japanese learners of English’, in: G. Aston, S. Bernardini and D. Stewart (eds) Corpora and Language Learners. Amsterdam & Philadelphia: John Benjamins Publishing Company. 45-66. Viberg, Å. (2006), ‘What one verb can do: The Swedish verb göra in a crosslinguistic perspective’, in: M. Suominen, A. Arppe, A. Airola, O. Heinämäki, M. Miestamo, U. Määttä, J. Niemi, K.K. Pitkänen and K. Sinnemäki (eds) A Man of Measure: Festschrift in Honour of Fred Karlsson on his 60th Birthday. Special Supplement to SKY Journal of Linguistics, 19: 243-257. Zhao, R. (1989), ‘A discourse analysis of relative clauses in Chinese and English: An error in “An error in error analysis”’, IDEAL, 4: 105-117.

An exploratory study of discourse organisation in French L1, Dutch L1, French L2 and Dutch L2 written narratives Annemie Demol and Pascale Hadermann Universiteit Gent Abstract Three major questions are dealt with in this contribution: (i) what principles of discourse organisation do French and Dutch native speakers use when writing narratives?; (ii) what strategies do (pre-)advanced French-speaking learners of Dutch and Dutch-speaking learners of French display in their texts?; and (iii) to what extent do the interlanguages differ from the source and target languages, and resemble each other? Our analyses reveal that it is not so much the degree of packaging, nor the use of subordinate clauses that distinguishes French from Dutch narratives, but the frequency of non-finite clauses and the use of present participles in secondary predication. Furthermore, the impact of target and source language on the learner’s interlanguage differs according to the construction that is being examined. The most striking differences were the absence of secondary predication in French L2 and the frequent use of complement clauses in Dutch L1. Finally, no clear organisational pattern typical of (pre-)advanced interlanguage was evident.

1.

Introduction

This article is concerned with cross-linguistic differences in discourse organisation and, more specifically, with the way in which foreign language learners deal with them. Previous studies have suggested that French and Dutch show different preferences in terms of their principles of discourse organisation. Our aim is to explore to what extent these claims hold for unprepared written film retellings. Secondly, we examine whether similar reports by (pre-)advanced learners of French and Dutch display a distinctive interlanguage organisation, whether they show signs of the transfer of discourse organisation principles from the writers’ native language, or whether the system of the target language has already been acquired. Discourse organisation is studied through an analysis of the number of clauses per sentence, the use of parataxis and hypotaxis, and the use of integrated structures (finite and non-finite clauses) with special emphasis on the kinds of subordinate clauses used. In Section 2 we briefly summarise the integrated contrastive approach on which this study is based. We also report on the findings of previous research on discourse organisation in French, Dutch and English, as well as in different kinds of interlanguages. In Section 3 we formulate our research hypotheses and define the features that are examined, and in Section 4 we present the data and the subjects. Section 5 gives a detailed account of the results of our study, while

256

Annemie Demol and Pascale Hadermann

Section 6 summarises the most important findings and compares them with those discussed in Section 2. 2.

Previous studies

2.1

The integrated contrastive model

In line with the objectives of this volume, the data we have collected are contrasted in various ways. Granger (1996) and Gilquin (2000/2001) convincingly argue that the integrated approach, which combines both traditional CA (contrastive analysis) and the younger discipline of CIA (contrastive interlanguage analysis), provides us with a solid framework, allowing not only for a thorough analysis of learner language and the phenomenon of transfer, but also for a more comprehensive study of both the learners’ mother tongue and the language they are in the process of acquiring. The contrastive analysis of source language and translated language will not be dealt with here, although in Section 2.2.1 we briefly summarise some of the results obtained by Cosme (2005, 2006, 2008) on discourse organisation and clause-linking in English, French and Dutch as source and translated languages. With respect to the comparison of interlanguages, and contrary to the current trend,1 we do not compare the interlanguage of learners who are acquiring the same foreign language and who have different mother tongues. Rather, we contrast the interlanguage of Dutch-speaking learners of French and Frenchspeaking learners of Dutch living in Brussels and the Walloon region of Belgium. This particular kind of CIA should cast light on both the transfer of strategies from the native language, and the properties of the general interlanguage system, irrespective of the native and foreign languages involved. It should however be borne in mind that any conclusions will be tentative, given that the present study only deals with two learner groups. 2.2

Discourse organisational preferences

2.2.1 From a contrastive perspective (CA) As Cosme (2006) points out, recent contrastive literature2 has claimed that different languages show different preferences in the way discourse is organised. The linguistic field of discourse organisation is, of course, a very extensive one, which covers phenomena such as information density, the structuring of information (thematic structure, focus assignment) and clause-combining. The present study mainly deals with this last aspect. A distinction can be made between two major clause-linking patterns. On the one hand, a language may prefer to combine clauses in a paratactic manner, mainly employing juxtaposition (commas and colons) and coordinating devices (conjunctions). On the other hand, it may opt for a more hypotactic organisation

An exploratory study of discourse organisation

257

in which subordination plays a major part. The first pattern is called incremental, horizontal or linear, while the second is termed hierarchical or vertical (FabriciusHansen 1996, 1999, Asher & Vieu 2005, Cosme 2008).3 In reality, these patterns can be seen as the extremes of a continuum along which languages can be ordered in terms of their preferences. On this continuum, French and Dutch are often claimed (Cosme 2008) to tend towards the extremes (see Figure 1): French is considered to display a hierarchical discourse structure, while Dutch arguably presents a more linear discourse organisation. English seems to prefer a combinatory style and Dutch a choppier one. INCREMENTALITY Parataxis (juxtaposition and coordination)

DUTCH

ENGLISH

HIERARCHY Hypotaxis (subordination)

FRENCH

Figure 1. The dependency hypothesis (adapted from Cosme 2008) Translational data corroborate these findings. While French translations of English texts are characterised by a high number of shifts from coordinate structures to subordinate ones, the opposite is true for English translations of French. In the translation of English into Dutch, it is also the shift from subordination to coordination that prevails. In the present study, we therefore need to bear in mind that French often seems to resort to subordination, whereas Dutch speakers frequently seem to make use of coordinated and juxtaposed clauses, both in newspaper texts and in fiction. These patterns will be compared to the narrative data we collected. 2.2.2 From an acquisitional perspective (CIA) Bartning & Kirchmeyer (2003) have looked at discourse organisation from an acquisitional point of view. They have contrasted the film retellings of 24 preadvanced and advanced Swedish learners of French with similar data obtained from Swedish and French native speakers in order to study the level of textual competence achieved by the learners. They examined, among other things, syntactic complexity, which was evaluated by three closely related features: (i) the degree of packaging, (ii) the degree of integration and (iii) the degree of ellipsis. With respect to the degree of packaging, the data show that intermediate and lower-advanced level learners use more simple sentences (consisting of only one clause) than native speakers do and that they seldom combine more than two clauses into one sentence. This results in a more linear account of the events

258

Annemie Demol and Pascale Hadermann

reported. As for the degree of integration, it seems that learners use fewer nonfinite subordinate clauses (i.e. infinitival and participial clauses, which are highly integrated structures) than native speakers. Finally, the use of elliptical relations, which presuppose referential continuity and imply that one or more elements are not explicitly mentioned, is rather rare in the learner data. When compared to data on (post-)basic and (very) advanced levels, these results reveal an acquisitional path which corresponds to gradually increasing complexity in the construction of sentences, as shown in Figure 2. decomposition synthetisation linearity hierarchisation x degree of packaging: juxtaposition of simple sentences > biclausal sentences > multiclausal sentences x degree of integration: juxtaposition > subordination > infinitival/participial > nominalisation clauses x degree of ellipsis: full constructions > subject ellipsis (elliptical coord.) > elliptical clauses (inf./part.)

Figure 2. Continuum representing the increasing syntactic complexity of sentences4 (from Bartning & Kirchmeyer 2003: 19) According to Bartning & Kirchmeyer (2003: 19-20), the development towards the use of more complex, more integrated and more elliptical structures is accompanied by an increasing ability of the learners to plan their discourse, to organise information hierarchically and to think in a more abstract way in a foreign language. Thus, the use of syntactically more complex structures means that the tasks of planning, organising and abstracting become increasingly important, and represent a considerable cognitive burden for the learner. This probably explains why the use of very integrated and complex structures only occurs at more advanced stages, since less proficient learners are already dealing with morpho-syntactic and lexical difficulties of other kinds. These take up much of their cognitive effort on the one hand, and on the other hand, may inhibit the use of complex constructions because the learner has not mastered the necessary tools. Other studies in second language acquisition (SLA) (Chini 2003, Lambert et al. 2003) have mainly concentrated on the presence and the role of subordinate clauses in learner language. None of them, however, include Dutch. Lambert et al. (2003: 49) deal with very advanced French- and German-speaking learners of English, whose film retellings are compared to those of native speakers of

An exploratory study of discourse organisation

259

English, French and German. Table 1 shows that the proportion of subordinate clauses differs in the native speakers’ texts: whereas subordinate clauses account for 19.1% of all clauses in the English language corpus, they only make up 11.2% of the German corpus, but represent 37% of the French corpus. The results for French and English confirm the descriptions of these languages in the previous section. German-speaking learners of English use even fewer subordinate clauses in English than in their native language (9.4%). French-speaking learners of English also reduce their use of hypotactic structures below that of their native language, although they still use them considerably more than the native speakers of English (30.9%). Table 1. Proportion of subordinate clauses (adapted from Lambert et al. 2003: 50) English Subordinate clauses/total clauses % subordination

742/3,886 19.1%

German

French

Ger. L1/ Eng. L2

Fr. L1/ Eng. L2

366/2,740 850/2,297

274/2,908

617/1,997

9.4%

30.9%

11.2%

37.0%

Lambert et al. (2003) also observed that the less advanced learners who participated in their research stuck closely to the chronology of events in the films they were describing. They created the necessary conditions for topic continuity (with the protagonist in subject position). This basic organisational mode has been observed in other studies dealing with other combinations of native languages and interlanguages, and is called the “prototypical treatment” (Watorek 1996). It appears to be a typical feature of the interlanguage system, which occurs at relatively early stages in the acquisition process. However it is still used by quite advanced learners who are confronted with a very complex verbal task, to simplify the task and manage all the cognitive and linguistic processes in on-line experiments (Lambert et al. 2003: 61). Chini (2003) examined the frequency and functions of subordinate clauses in the interlanguage of learners who have reached varying degrees of post-basic proficiency. She points out clearly that the presence of more complex syntactic structures depends not only on the increasing complexity of the interlanguage, or the characteristics of the mother tongue and/or target language, but also on the personal stylistic preferences of the subjects and the rhetorical effects they are pursuing. The data in Chini’s (2003) study are film retellings of the silent movie Modern Times5 starring Charlie Chaplin, produced by eighteen German learners of Italian, thirteen native speakers of Italian and ten native speakers of German. The learners were divided into four groups according to their proficiency level, ranging from post-basic to advanced. First, the narratives produced by eight learners from across the groups were compared to the native-speaker data in

260

Annemie Demol and Pascale Hadermann

terms of the proportion of subordinate clauses in general, and of complement clauses, relative clauses, pseudo-relative clauses (a category typical of Italian which we will not consider) and adverbial clauses in particular. As Table 2 shows, the percentage of subordinate clauses (as opposed to the total number of clauses) is somewhat smaller in the learner data: 27.4%, compared to 33.3% in Italian and 32.4% in German.6 In addition, the learners use fewer non-finite subordinate clauses than the native speakers of Italian, which coincides with the results reported by Bartning & Kirchmeyer (2003). Chini (2003) does not provide the data for German in this respect. Table 2. Proportion of subordinate and non-finite clauses in Chini (adapted from Chini 2003: 81, 83)

Subordinate clauses/total clauses % subordination Number of infinitival clauses Number of gerund clauses Number of participial clauses Total non-finite clauses % non-finite clauses (total subordinate clauses)

German

Italian

116/358 32.4% -

544/1,631 33.3% 109 25 4 138 25% (138/544)

-

Ger. L1/ Italian L2 234/864 27% 40 3 0 43 18.4% 43/234

Finally it is worth pointing out that Chini draws a parallel between the use of paratactic structures and that of over-explicit referential expressions, and suggests that these choices are linked to more local planning and elaboration strategies. These in turn lead to an element-by-element treatment, and the connections between the clauses and sentences are mainly implicit. 3.

Research hypotheses and methodology

The following global hypothesis was formulated in accord with the contrastive literature reviewed in Section 2: MAIN HYPOTHESIS: Dutch-speaking learners of French and French-speaking learners of Dutch encounter difficulties when writing narratives in their second language because they have to apply, respectively, a more hierarchic or vertical discourse organisation, and a more incremental or horizontal discourse organisation in their L2 than they are accustomed to in their L1. In order to test this hypothesis and pinpoint specific problems, several subhypotheses and research questions were formulated. These concern the degree of

An exploratory study of discourse organisation

261

packaging, the degree of dependency and the degree of integration. They are related to the global hypothesis as shown in Figure 3. Hierarchic or vertical discourse organisation + + +

degree of packaging degree of dependency degree of integration

Incremental or horizontal discourse organisation   

Figure 3. Relationships between the type of discourse organisation and the degree of packaging, dependency and integration. 3.1

The degree of packaging

Bartning & Kirchmeyer (2003) introduced the idea of degree of packaging (“degré d’empaquetage”). It will be examined by taking account of: o o

the average number of clauses per sentence; and the proportions of mono-, bi- and multi-clausal sentences.

With this objective in mind, a fairly traditional approach was adopted (see, among others, Riegel et al. 1994: 472). This consists of defining a sentence on the basis of graphic features (i.e. a capital letter at the beginning and a full stop or question/exclamation mark at the end) and a clause on the basis of the presence of (at least) a subject (which can, of course, remain implicit)7 and a predicate. Sentences were classified as simple (consisting of one clause), complex (biclausal) or multiple (consisting of more than two clauses). Various combinations are possible within any one sentence. The following example illustrates this approach: (1)

(i) Là il fait un festin de roi et (iii) se présentant à la caisse, (ii) appelle l’agent de quartier lui-même (iv) en l’accostant par un signe de la main. (FR L1) ‘(i) There he dines in royal style and, (iii) presenting himself at the cash desk, (ii) he himself calls the local policeman (iv) by signalling to him with his hand.’

This is a multiple sentence, which contains an independent clause (i), coordinated to (ii) a main clause, which is in turn preceded by (iii) a participial clause and followed by (iv) another participial clause (constructed around a gerund). The following sub-hypothesis was formulated with respect to the degree of packaging: SUB-HYPOTHESIS 1: French is expected to exhibit a higher degree of packaging than Dutch, as shown by:

262

Annemie Demol and Pascale Hadermann o o

a higher average number of clauses per sentence; and a higher proportion of multiple sentences.

Accordingly, the following research question was posed: INTERLANGUAGE RESEARCH QUESTION 1: Do the learner data present a similar degree of packaging, do they present a degree of packaging similar to that of the learners’ L1 or have the packaging characteristics of the L2 already been acquired at the (lower) advanced stage? 3.2

The degree of dependency

The degree of dependency was evaluated by examining the frequency of parataxis (juxtaposition and coordination) and hypotaxis (subordination and secondary predication). According to Lehmann (1988), hypotaxis, which traditionally covers all kinds of subordinate clauses, is the result of a process he calls “hierarchical downgrading”. The following sub-hypothesis on dependency will be tested: SUB-HYPOTHESIS 2: French native speakers use hypotaxis more frequently than Dutch ones, who prefer parataxis. French narratives written in French therefore present a higher degree of dependency than narratives written in Dutch. With respect to the interlanguage of the two learner groups, the present study sets out to answer the following question: INTERLANGUAGE RESEARCH QUESTION 2: Do the narratives of language learners show a preference for parataxis in general, do they manifest the same tendencies as narratives written in the learners’ L1, or have the characteristics of the target language already been acquired? 3.3

The degree of integration

While the degree of dependency is certainly a very important feature of the distinction between more hierarchic and more incremental styles of discourse organisation (see Section 2.2), it is not all-revealing. It is possible to distinguish even further between more and less integrated structures in subordinate clauses (Bartning & Kirchmeyer 2003), taking the description of discourse organisational preferences to a further level. According to Lehmann (1988: 204), a subordinate clause may be desententialised, “turning it into a simple constituent of the main clause”. At the end of this process we find non-finite verb forms and verbal nouns. Thus, example (2a) below presents a more integrated structure than (2b):

An exploratory study of discourse organisation (2a) (2b)

263

Il prétend qu’il a volé le pain. ‘He claims that he stole the bread.’ Il prétend avoir volé le pain. ‘He claims to have stolen the bread.’

The frequency of finite and non-finite subordinate clauses was examined in the various data-sets. Furthermore, in order to get a more complete picture of the integration process, a distinction was made between non-finite subordinate clauses containing an infinitive, a past or present participle, and a gerund.8 In this, we departed from the rough distinction made by Bartning & Kirchmeyer (2003) between finite and infinitival/participial clauses in terms of different levels of integration. We felt that there was a need for a more fine-grained distinction, to distinguish for instance between (2b) and (2c-d) below: (2c)

(2d)

Buitengekomen uit het restaurant belt de agent naar het commissariaat zodat ze een wagen kunnen sturen om Charlie weg te voeren. (NL L1) ‘Having left the restaurant, the policeman calls the police station so that they can send a car to take Charlie away.’ Laissant vides chacun des plats sur la table, il se dirige vers la sortie où le récupère le gendarme. (FR L1) ‘Leaving every one of the dishes on the table empty, he goes to the exit where the policeman catches him again.’

The participial clauses in (2c) and (2d) do not behave like truly dependent clauses, in the way the finite and the infinitival clauses in (2a) and (2b) do. They resemble more closely detached secondary predications.9 The distinction between truly dependent clauses and detached secondary predication in the form of participial clauses is accounted for by Lehmann (1988). According to his clause linking model (ibid. 185), the participial construction in (2c) and (2d) is “part of the main clause and insofar embedded in it. However, its syntactic function […] is a blend, as it were, of an apposition and an adverbial, and thus not subject to government”. On the continuum of “hierarchical downgrading” (see Figure 4) put forward by Lehmann (1988: 189) the kind of participle used in (2c) and (2d), which he calls a “conjunct participle”, is considered to be less embedded than governed clauses (e.g. the finite clause in [2a] and the infinitival clause in [2b]), but more embedded than independent and adjoined clauses (i.e. paratactic structures). parataxis independent clause

adjoined clause

correlative diptych

medial clause

conjunct participle

embedding governed clause

Figure 4. Hierarchical downgrading (Lehmann 1988: 189) Overall, we distinguished between parataxis (which includes coordination and juxtaposition) and hypotaxis (which covers subordination – with finite or

264

Annemie Demol and Pascale Hadermann

non-finite verbs – and secondary predication of a participial kind) in order to give as complete a survey of the degree of dependency and integration as possible.10 Figure 5 summarises the various distinctions that we took into account. It also shows that, although dependency and integration are closely related, they are not synonymous. Dependency level PARATAXIS Juxtaposition Coordination

HYPOTAXIS Secondary predication

Secondary predication (i.e. participial clauses)

Subordination

Integration level Finite Non-finite subordinate subordinate clauses clauses (i.e. infinitives)

Figure 5. Distinction between dependency and integration Our sub-hypothesis with respect to the degree of integration, then, is the following: SUB-HYPOTHESIS 3: French is characterised by a higher degree of integration than Dutch. This means that the French data are expected to contain a higher proportion of more integrated kinds of structures than the Dutch: non-finite subordinate clauses (infinitives and gerunds) > finite subordinate clauses > secondary predication (participial clauses) > coordination > juxtaposition The following question will be addressed in the learner data: INTERLANGUAGE RESEARCH QUESTION 3: Do the narratives of the two learner groups resemble each other with respect to the degree of integration they show, do they have a similar degree of integration to that of the learners’ mother tongue, or do they resemble more closely the usual patterns in the language being learned? Overall, by comparing the degree of packaging, dependency and integration in the Dutch and French interlanguages, we hope to be able to shed some light on the question of whether linear and local treatment are indeed typical features of the interlanguage system, even at a rather advanced stage. Finally, we also took a closer look at the variety of subordinate clauses that native speakers and learners use.

An exploratory study of discourse organisation 4.

265

The data: subjects and task

We analysed the narratives written by 19 Dutch-speaking learners of French and 19 French-speaking learners of Dutch (all Belgians living in Brussels or in the Walloon region of Belgium). These were compared to the writings of 19 native speakers of Dutch and 19 native speakers of French (also all Belgians). The first group (FR L2) consisted of Dutch-speaking students enrolled in the first year of “Language and literature: French + a second language” at Ghent University (GU). These students thought the task to be a preliminary test designed to evaluate their linguistic abilities at the beginning of the academic year. All of them had taken French for eight years.11 The second group (NL L2, where NL refers to Nederlands, the Dutch word for “Dutch”) was composed of ten French-speaking students enrolled in the second year of “Language and literature: Dutch + a second language” at the Catholic University of Louvain (UCL) and of nine French-speaking students enrolled in the third year of the school for interpreters at Mons. All of these learners also undertook the task as part of their writing class. The native speakers of Dutch (NL L1) were all students enrolled in the Arts and Humanities faculty at Ghent University and fourteen of them had a “language profile”.12 They volunteered for the task, in return for a small fee. The last group (FR L1) consisted of French-speaking students studying Romance languages at the Free University of Belgium (ULB), who completed the task as part of a writing class.13 Since our aim was to compare groups of learners who had reached a fairly advanced stage (based on Bartning 1997 and Bartning & Krichmeyer 2003), we decided to work with French-speaking learners who were in the second and third year of university (rather than the first year, like their Dutch-speaking counterparts), because they had not received the same amount of instruction in Dutch as the Dutch-speaking learners had in French in primary and secondary school (see note 11). For the same reason, this group performed the task near the end of the academic year. Given that the learners had more or less the same level of proficiency, we decided not to perform any within-group analyses, although there was obviously a certain degree of individual variability.14 All the students undertook the same task. They were asked to describe in writing, as accurately as possible, the plot of the Alone and Hungry sequence (12 minutes) of the film Modern Times (cf. Lambert 1997, Chini 2003). They were allowed to take notes while viewing the film, but were not allowed to make a draft of their narrative first. They were given a maximum of one hour for this rather complex task.15

266

Annemie Demol and Pascale Hadermann

5.

Corpus findings

5.1

Some preliminary observations

Before we comment on the way information is organised in the four corpora, it must be pointed out that the learners’ narratives differ considerably from the native speakers’ in terms of length, as Table 3 brings out. These differences possibly point to the greater cognitive effort the learners had to make, in comparison with the native speakers, and, consequently, to the fact that they needed more time to activate their procedural knowledge. Table 3. Average length of the narratives Total no. of words per corpus Average no. of words per narrative Average no. of sentences per narrative Average no. of clauses per narrative

FR L1 5,669 298.37 18 44

FR L2 3,394 178.63 13 29

NL L1 5,965 313.95 20 45

NL L2 2,864 203.37 16 34

The complexity of the task and its greater cognitive demands are also related to the fact that five different characters appear in the extract, sometimes simultaneously. Charlie Chaplin is the main protagonist, and the girl he wants to help plays a major role at the beginning of the sequence. The policeman is present during more or less the entire story, whereas the baker and the witness are less important characters, only appearing at the beginning. Do the learners and the native speakers refer to all these characters in the same proportions? Charlie is clearly the protagonist in all four corpora, but he gets some competition from the girl in the FR L2 corpus. These learners focused on the girl at the beginning of their narratives and related in detail what happens to her.16 However, they seem to be unable to maintain the same level of detail afterwards and so give a more condensed account of the adventures of Charlie and the policeman. This suggests some planning difficulties and problems with time management. The learners of the NL L2 corpus seem to experience fewer cognitive (planning) problems.17 5.2

Discourse organisation

5.2.1 The degree of packaging As Cosme (2008) points out, sentence length can be seen as a first indication of syntactic complexity (see also Hannay & Mackenzie 1996: 44). It appears from Table 4 that the native-speaker Dutch sentences (NL L1) contain, on average, only slightly fewer words than the native-speaker French sentences (FR L1). Moreover the sentences in the learner corpora are only slightly shorter than those in the source and target languages. The shortest sentences were produced by the NL L2 group. The average number of clauses per sentence is nearly the same in

An exploratory study of discourse organisation

267

the different corpora, although the NL L2 is again the smallest. A very small overall difference can be observed between French and Dutch, which could perhaps be seen as a first tentative indication of the greater syntactic complexity of French. Both interlanguages closely resemble their target languages in the number of clauses. Table 4. Average number of words and clauses per sentence FR L1 16.62 2.6

Average no. of words Average no. of clauses

FR L2 14.26 2.5

NL L1 15.62 2.3

NL L2 12.46 2.2

Figure 6 shows the average number of simple (mono-clausal), complex (bi-clausal) and multiple (multi-clausal) sentences. This gives a better idea of the actual distribution of clauses over sentences.

9 8 7 6 5

simple

4

complex multiple

3 2 1 0 FR L1

FR L2

NL L1

NL L2

Figure 6. Average number of simple, complex and multiple sentences per narrative Native speakers The average number of simple and multiple sentences is equal in the two sets of native speaker data, although they differ in the number of complex sentences, which are used more often by the Dutch native speakers. Hence, the distribution of the clauses does not seem to provide any support for the hypothesis that Dutch has a choppier style (see Vanderauwera 1985) and French a more complex sentence structure.

268

Annemie Demol and Pascale Hadermann

Learners The learners’ narratives are, on average, shorter and contain fewer sentences than the native speakers’, and this is reflected in the frequency of sentence types. Proportionally speaking, the gap between simple sentences on the one hand and complex and multiple sentences on the other appears to be smaller in the learner corpora. The FR L2 corpus contains as many complex as multiple sentences and the fewest simple sentences of all the corpora. The NL L2 corpus resembles the target language corpus (NL L1) with respect to the tendency to use complex sentences. Furthermore, the NL L2 corpus arguably presents the most horizontal or linear discourse organisation in terms of the construction-types examined, since it contains as many simple sentences as multiple sentences, and more simple sentences than the other data sets, and since the most frequently used type of sentence is a complex one. Summary We found only minor differences at the level of packaging: French texts display a slightly higher degree of packaging than Dutch ones, but the difference between the two data sets is certainly not as pronounced as that found by other researchers. Sub-hypothesis 1 is therefore not confirmed by the data. The learner corpora do not point to any clear features that are characteristic of a “general” interlanguage system. Although they contain, on average, approximately the same proportion of clauses as the native speaker corpora (see Table 4), they do seem to have a slightly less complex discourse organisation in terms of the average number of words per sentence and the proportion of simple sentences (and, for the NL L2 group, the proportion of complex sentences). 5.2.2 The degree of dependency In order to gain a better understanding of the internal sentence organisation, we need to examine the different types of clause-linking, i.e. parataxis (juxtaposition and coordination) vs. hypotaxis (subordination and secondary predication). Before we discuss the results for this feature, however, it is necessary to point out that the quantification of types of clause-linking is not a straightforward matter, at least in multiple sentences. In complex sentences, two clauses can be combined through juxtaposition, coordination or subordination. In multiple sentences, one clause can, for instance, be subordinated to another clause, while simultaneously being coordinated to a second subordinate clause. An example is given in (3), where the subordinating and the coordinating conjunctions have been italicised: (3)

Puis, une dame a raconté au policier qu’il avait fait erreur et un peu plus tard Charles et la fille se sont rencontré dans la camionnette ... parce que entretemps le bonhomme avait déjà fumé une cigare et mangé deux plats du jour sans rien payer. (FR L2) ‘Then, a lady has told the policeman that he had made a mistake and a bit later Charles and the girl met in the van… because in the meantime the

An exploratory study of discourse organisation

269

young chap had already smoked a cigar and eaten two meals without paying for anything. Such cases of “double linking” were taken into account in our analysis, and so two types of linking were quantified. The results are shown in Tables 5 and 6. Table 5. Types of linking in complex sentences Parataxis - juxtaposition - coordination total Hypotaxis - subordination - sec. predication total

FR L1

FR L2

NL L1

NL L2

13% 44% 57%

11% 38% 49%

2% 50% 52%

2% 44% 46%

38% 5% 43%

51% 0% 51%

48% 0% 48%

54% 0% 54%

Table 6. Types of linking in multiple sentences Parataxis - juxtaposition - coordination total Hypotaxis - subordination - sec. predication total

FR L1

FR L2

NL L1

NL L2

21% 35% 56%

21% 40% 61%

12% 48% 60%

13.7% 47% 60.7%

36% 8% 44%

39% 0% 39%

39% 1% 40%

39% 0.3% 39.3%

Native speakers In the French native speaker corpus (FR L1), coordination is the prevalent type of linking in complex sentences, while subordination is by one percent the most frequent linking type in multiple clauses. Juxtaposition is the least common type of paratactic linking in both kinds of sentences, but it plays a more important role in multiple sentences. The proportions of parataxis and hypotaxis are similar in complex and multiple sentences. In the Dutch native speaker corpus (NL L1), coordination is the most common type of linking in both complex and multiple sentences. However, in complex sentences it is very closely followed in popularity by the use of subordination. Juxtaposition is very seldom used in complex sentences, but it plays a larger role in multiple sentences. Overall we can conclude that, for complex sentences, the findings do not really correspond to the picture that has been given of Dutch in earlier studies (which do not distinguish complex and multiple sentences): hypotaxis is well represented, and indeed is even more

270

Annemie Demol and Pascale Hadermann

frequent than in the FR L1 corpus. In multiple sentences, however, the results seem to conform better to the traditional picture: there is a gap of 20% between the use of parataxis (juxtaposition 12% and coordination 48%) on the one hand, and hypotaxis on the other (subordination 39% and secondary predication 1%), whereas this gap is only 12% in the FR L1 corpus. Within the paratactic organisation, coordination plays the leading part in the NL L1 corpus.18 Overall, we cannot conclude that French is characterised by much more subordinate linking than Dutch, or Dutch by much more juxtaposition. Our data do not really support sub-hypothesis 2. The most important difference is in secondary predication, which reaches 5% in complex sentences and 8% in multiple sentences in the FR L1 corpus (see examples [4a] and [4b] below), whereas in the NL L1 corpus, there were no examples of secondary predication in complex sentences and very few in multiple sentences (see example [4c]). (4a)

(4b)

(4c)

Voyant le désarroi de la jeune fille, Charlie Chaplin se dénonce à sa place pour le vol et se fait donc arrêter par les agents de police. (FR L1) ‘Seeing the young girl’s desperation, Charlie Chaplin turns himself in for the theft in her place, and gets himself arrested by the policemen.’ C’est alors que monsieur Chaplin, touché sans doute par la jolie frimousse et la mauvaise position de la jeune fille, s’accuse du vol. (FR L1) ‘It was then that Mr Chaplin, touched, without a doubt, by the pretty face of the girl and the awkward position she was in, said that he was guilty of the theft.’ Charlie Chaplin, overtuigd van het feit dat hij die vrouw wil helpen, stapt een cafetaria binnen, vult twee schotels eten en eet dit alles smakelijk op. (NL L1) ‘Charlie Chaplin, convinced that he wants to help this woman, goes into a cafeteria, fills two dishes with food and eats it all up with great pleasure.’

The specific use of present participial clauses in the FR L1 corpus (see example [4a]) cannot be seen entirely separately from an apparent difference between French and Dutch, which will be further commented on in Sections 5.2.3 and 5.3 below. In principle, the present participle can function as a predicate in Dutch, as illustrated by the following examples taken from the E-ANS (Coppen et al. 2004: “Het tegenwoordig deelwoord”, 2.4.5): (5) (6)

Hij komt lopend. ‘He comes walking (= He comes on foot).’ Karel liep lachend weg. ‘Charles walked away, smiling.’

No indications are given in the E-ANS about the frequency of this construction in modern Dutch. However, our data suggest that these forms are not often used, since there were only six occurrences of participial clauses in our data (see example [4c]), none of which contained a present participle.19

An exploratory study of discourse organisation

271

Learners Let us now take a look at the interlanguages. We will first discuss the use of each type of clause-linking separately, by comparing their presence in complex and multiple sentences. Surprisingly, both learner corpora are characterised by high proportions of hypotaxis, consisting mainly of subordination. The learners’ complex sentences contain even more subordination than the Dutch nativespeakers’. In fact, subordination is the most common type of clause-linking used by learners in complex sentences. In multiple sentences, both learner groups use as many subordinate links as the Dutch native speakers. Subordination is no longer the preferred linking device, but it is still used almost as often as coordination in the FR L2 corpus. Coordination is the second most common type of clause-linking used by learners in complex sentences, and the most frequent one in multiple sentences. In complex sentences, the use of juxtaposition is very restricted in the NL L2 corpus (2%), but is more common in the FR L2 corpus (11%). Juxtaposition plays a more important role in multiple sentences in both learner corpora, but the learners of French use this linking device considerably more often than the learners of Dutch (21% in FL L2 and 14% in NL L2). Looking at these data from a different angle, and comparing the relative weight of parataxis and hypotaxis, we observe that the pattern in complex clauses is quite different from that of native speakers. Hypotaxis is used slightly more often than parataxis, although in multiple sentences, parataxis wins out over hypotaxis. But since subordination reaches 39% even in multiple sentences, it cannot be claimed that the interlanguage of our (pre-)advanced learners shows signs of “prototypical treatment”, i.e. of a more simplified, chronological and incremental organisation. Learners vs. natives Finally, we want to compare the data for interlanguages and native speakers. Overall, both interlanguages resemble their target languages more closely than their source languages. The most striking observation concerns the percentage of juxtaposed links in the FR L2 corpus: there are 9% more juxtaposed clauses in this corpus than in the NL L1 corpus, and only 2% less than in the FR L1 corpus. It seems as though the learners have succeeded in the process of acquiring the structuring of their target language with regard to juxtaposition. On the other hand, the Dutch-speaking learners of French (FR L2) do not make any use of secondary predication involving participial clauses, although this is quite characteristic of French L1 (cf. the examples in [4]). The French-speaking learners of Dutch (NL L2), however, seem to have understood the importance of coordination and subordination in their target language and often use it. These differences between the two learner groups might be explained by the nature of the properties they need to acquire: it is easier to use structures for which an analogous organisation pattern exists in the source language, than to use structures for which the equivalent in the source language seems to be rare and

272

Annemie Demol and Pascale Hadermann

perhaps slightly archaic (secondary predication involving present participial clauses in Dutch). Summary All things considered, subordination is relatively frequent in our Dutch corpus, but coordination is the most frequently-used linking device. The French native speakers often employ juxtaposition and secondary predication; more specifically, they combine a participial clause with a main clause. Overall, the learners seem to be well on their way to producing the characteristics of their respective target languages, but the Dutch-speaking learners still need to grasp the particular properties of the French hypotactic pattern. Surprisingly, both learner groups use at least as many subordinating links as the native speakers. This finding, however, would need support from a more qualitative analysis (see Section 5.3 below). Future research should also check whether there is a connection between this and the overuse of causal subordinate clauses documented by Carroll & Lambert (2003), Chini (2003) and Kirchmeyer (2003). 5.2.3 The degree of integration Even though our French and Dutch data do not differ greatly with respect to the degree of packaging (Section 5.2.1), they might well differ with respect to the degree of integration. As was stated in Section 3.3, not all hypotactic structures that are characterised by a dependency relation present the same degree of integration: secondary predication is less well-integrated with the main predication than finite or infinitival subordinate clauses. With respect to the latter two types of subordinate clauses, it has been shown (Lehmann 1988, Bartning & Kirchmeyer 2003) that infinitival clauses manifest a higher degree of integration than finite subordinate clauses. The degree of integration of hypotactic structures will therefore be evaluated by distinguishing between finite subordinate clauses, infinitival subordinate clauses, participials and gerunds. Following Wilmet (1997: 530-531), we do not consider gerunds as part of secondary predication. It should be noted that Dutch does not really possess gerunds, but the use of the present participle preceded by the word al can be compared to the French “gérondif”: (7)

(Al) zwemmend bereikte hij de overkant. En nageant, il est arrivé à l’autre côté. ‘Ø Swimming he reached the other side.’

No examples of this form were found in either the NL L1 or the NL L2 corpora. Table 7 shows how finite and non-finite subordinate clauses were represented in the corpora.

An exploratory study of discourse organisation

273

Table 7. Hypotaxis and integration Degree of integration sec. predication (participial clauses) finite sub. clauses non-finite sub. clauses infinitives gerunds sub-total

FR L1 19% 48%

FR L2 0% 71%

NL L1 2% 78%

NL L2 0.5% 73.5%

29% 4%

23% 6%

20% 0%

26% 0%

33%

29%

20%

26%

Native speakers The use of non-finite subordinate clauses was clearly more widespread among the French (33%) than the Dutch (20%) speakers, which seems to support subhypothesis 3. A substantial number of the FR L1 subordinate clauses are secondary predication participial clauses (see Table 7), which actually indicates a lower degree of integration. The non-finite clauses in the NL L1 corpus are mainly infinitival clauses. In the FR L1 corpus, the percentage of infinitival clauses is 9% higher than in the NL L1 corpus. This finding is again consistent with the idea that French has a more integrated discourse organisation than Dutch. Learners Both learner groups used more non-finite subordinate clauses than the Dutch native speakers. The French-speaking learners of Dutch (NL L2) seem to have acquired the use of integrated structures in their target language, since their nonfinite clauses were, with one exception, all infinitival clauses, as was the case for the target language corpus. Unlike the French native speakers (FR L1), the FR L2 group does not use participial clauses. The following examples illustrate that differences between the normal discourse organisation of Dutch and French have an impact on the interlanguage of the Dutch-speaking learners of French (FR L2). Example (8) shows how these learners prefer to use their native-language devices, while example (9) illustrates the French strategy: (8)

(9)

La dame qui a vu la jeune fille voler du pain informe le boulanger qui alerte la police. (FR L2) ‘The lady who saw the young girl steal a loaf of bread informs the baker, who alerts the police.’ Le boulanger, averti entre-temps par un témoin à l’œil vigilant, se précipite à la suite de la voleuse infortunée et la rejoint sans effort. (FR L1) ‘The baker, alerted in the meantime by a vigilant eye-witness, hurries to pursue the unfortunate thief and catches up with her easily.’

274

Annemie Demol and Pascale Hadermann

Summary In addition to the use of subordinate clauses – which can be seen as a first indication of a more integrated discourse organisation – the frequency of finite and non-finite clauses has been suggested as a criterion for measuring the level of integration. At first sight the French native speakers’ narratives do seem to be characterised by a higher level of integration, because the percentage of non-finite clauses is clearly higher in the FR L1 corpus than in the other data sets (cf. Table 7: 33%). On the other hand, 19% of the subordinate clauses are detached participial clauses which function as secondary predications. They are, as such, even less integrated than the finite subordinate clauses. Given the contradictory evidence of the high percentage of extremely integrated structures in the FR L1 corpus (33%), and the frequent use of the least integrated dependent clauses (namely secondary predication, which is rare in NL L1 corpus), there is no clear answer with regard to our third sub-hypothesis. 5.3

Zooming in on subordinate clauses

Table 8 shows the types of subordinate clause used in the four corpora. Table 8. Types of subordinate clause Complement Adverbial Relative Gerund Participial

FR L1 16% 31% 30% 4% 19%

FR L2 26% 32% 37% 6% 0%

NL L1 25% 35% 38% 0% 2%

NL L2 35% 35% 29% 0% 1%

Native speakers The Dutch native speaker corpus contains 8% more relative clauses than the French native speaker corpus (38% compared to 30%). A similar difference can be observed for the number of complement clauses in the two native speaker data sets: the Dutch-speaking subjects (NL L1) use 9% more complement clauses than the French-speaking subjects (FR L1). With adverbial clauses, there is only a difference of 4% between the two groups. The subordinate clauses used by the French native speakers also include 4% gerunds and 19% present participles, two structures which are essentially absent from the Dutch native speakers’ writings. Learners The types of subordinate clauses used in the interlanguages are similar to the learners’ mother tongue with respect to relative clauses, whereas the proportions of adverbial clauses resemble the target languages more closely. The FR L2 group uses almost the same proportion of complement clauses (26%) as the NL L1 group (25%). The NL L2 group uses an even higher proportion (35%) of complement clauses. Together with adverbial clauses, complement clauses are the

An exploratory study of discourse organisation

275

most frequent subordinate clauses in this learner corpus. There are almost 20% more complement clauses in the NL L2 group than in the French native speaker corpus (FR L1). Despite their reluctance to use present participles, the learners of French (FR L2) do use gerunds – in fact they make more use of this structure than French native speakers (FR L1). This indicates that they have acquired an important form-function relation that is used by native speakers. The structure is not unknown to the learners, since their mother tongue possesses a close variant of the French gerund (al + present participle), but this variant is seldom used in Dutch (see Section 5.2.3). We can conclude that the French interlanguage (FR L2) looks a lot like native Dutch in terms of the use of different types of subordinate clauses. The main difference between them is the presence of gerunds in the learner corpus. The Dutch interlanguage (NL L2) is striking for its high percentage of complement clauses, whereas the French native speaker (FR L1) corpus is marked by a small proportion of complement clauses. Lambert (1997) has pointed out that the use of complement clauses following verbs like want, believe, hope or desire can serve to make the characters’ intentions, hopes and beliefs explicit, and explain the reasons for some of their actions. It could thus be that a considerable proportion of the complement clauses in the learner corpus fulfil this function, and relate to the strong presence of logical relations in the French native speaker corpus. However, an analysis of these complement clauses showed that they mostly occur with declarative verbs (say, ask, convince, etc) and perception verbs (see, observe). The same types of verbs are used in the NL L1 (Dutch native speakers) corpus. 6.

Conclusion

6.1

Just how “vertical” is French and how “horizontal” is Dutch?

Our study gives a highly nuanced picture of the principles of discourse organisation in French and Dutch narratives and shows that there are no straightforward answers to the question of how vertical French is and how horizontal Dutch is. While there are no great differences between them in terms of the average number of words and clauses per sentence, and only small differences with respect to the use of simple, complex and multiple sentences, Dutch native speakers do seem to prefer more complex sentences, whereas French native speakers opt for slightly more multiple sentences than complex ones. Quite surprisingly, the two native corpora contain similar proportions of subordinate clauses. However, the French native speakers’ data have a higher proportion of non-finite clauses than the Dutch data. As far as clause-linking is concerned, no clear-cut patterns appear. The Dutch data are relatively paratactic and display a fair amount of coordination. However, they also contain a large number of subordinate links, although not a lot of juxtaposition. In the French

276

Annemie Demol and Pascale Hadermann

data, coordination prevails in complex sentences, and subordination in multiple sentences. Comparatively speaking, these data also contain quite a lot of juxtaposition and secondary predication. Overall, it is not so much the degree of packaging nor the use of subordinate clauses that distinguishes French and Dutch in this study, but the frequency of non-finite clauses and more specifically the use of secondary predication in the form of present participles separated from the main clause by a comma. 6.2

Indications of transfer and interlanguage features

The learners’ narratives are considerably shorter than those of native speakers. The FR L2 learners also put more emphasis on the girl in the story than the other subjects do, but they are unable to give as much detail in the later part of their accounts as they do at the beginning. These findings suggest some difficulties at the planning level and with respect to the cognitive treatment of information, as may still be expected at a lower advanced level (see Bartning & Kirchmeyer 2003). On average the sentences written by the learners do not contain many fewer words or clauses than those written by native speakers. Proportionally, the learner corpora display more simple sentences, but nevertheless they frequently contain complex and multiple sentences. Our data do thus not fully corroborate Bartning & Kirchmeyer’s (2003) findings, where learners produced only a small number of multiple sentences, and showed a clear preference for simple sentences. The differences could be due to the different composition of the learner groups, since Bartning & Kirchmeyer’s group contained not only lower advanced-level but also intermediate learners. The overall proportion of subordinate clauses in the learner data does not differ from that in the native data. No transfer-related phenomena are observed in this respect, since there are no important differences between the two native datasets. Our French-speaking learners of Dutch thus do not display the same tendency as Lambert et al.’s (2003) French-speaking learners of English, who use many more subordinate clauses than native speakers of English (in line with the custom in their native language). Moreover, our Dutch-speaking learners of French (FR L2) juxtapose clauses more often than the native speakers of Dutch (NL L1), as do the native speakers of French (FR L1). The FR L2 learners have not yet acquired the specific organisation principle of French that consists of secondary predication through the linking of participial clauses to main clauses with a comma. However they do use gerunds. The NL L2 corpus, in turn, is characterised by a high percentage of complement clauses, as compared to both the target and (especially) the source language, a phenomenon which has yet to be explained. It thus seems that there is no clear organisational pattern typical of interlanguage at a (lower) advanced level. No real prototypical treatment consisting of a simplification of the task for reasons to do with cognitive cost is visible in our data. In this respect, our results differ from those of Lambert et al.

An exploratory study of discourse organisation

277

(2003). The impact of target and source language in our data differs according to the feature being examined. 6.3

Some methodological implications for future research

We would like to conclude by formulating some remarks concerning the use of the integrated approach. However promising for the future, it should be pointed out that the integrated contrastive model can only function properly when all the data being compared are of a similar nature. We cannot help but wonder if the particular kind of data used in this project can be expected to present the same characteristics as the original and translated data used by Cosme (2008) and by the authors she refers to. Even though we have worked with native speakers who are probably as highly educated as the professional writers involved in the corpora used by others, and even though the majority of them have a “language profile”, their writing is likely to differ from that of trained professionals. Moreover, the specific character of narrative writing sets it somewhat apart from the novels, newspaper articles or editorials and parliamentary texts that make up the corpora used by Cosme (2006, 2008) and the authors she refers to. This means that the comparison of our findings with those obtained within the CAframework cannot be conclusive. It also shows that the preference for languagespecific organisation principles should be tested over a range of different text types and genres. We therefore think that it is necessary to continue compiling representative and diversified corpora of all kinds (different source and target languages, different translated languages, different interlanguages), as well as to stimulate the methodological debate about the ways in which this should happen and about the actual contrastive analyses. In addition, our contribution illustrates the need for a clear terminology and well-defined concepts in the study of discourse organisation in order to increase knowledge in this field. Not only has the comparison of our results with those obtained by others proved to be difficult at times, but the diverging results in other CIA-studies (for instance those on the use of subordinate clauses in German reported by Chini [2003] and Lambert et al. [2003]) are probably related to different conceptions of a particular phenomenon. The differences between Chini’s (2003) findings and Lambert et al.’s (2003), for instance, are more likely to be due to a different conception of subordination than to a problem of text-type comparability, since the two studies are based on comparable oral narratives.20 Notes 1

The existence of learner corpora of the International Corpus of Learner English (ICLE) type has favoured comparisons of the interlanguage of learners who have different native languages, but who are acquiring the same foreign language. See Section 2.2.2 for examples.

278

Annemie Demol and Pascale Hadermann

2

Cosme refers to the work carried out by Vanderauwera (1985), Chuquet & Paillard (1987) and Hannay & Mackenzie (1996). See also her own study (Cosme 2006) for a comparison of French and English.

3

We would like to express our sincere gratitude to Christelle Cosme for having provided us with this manuscript.

4

The term “subordination” under the heading “degree of integration” refers to the use of finite subordinate clauses, and therefore does not include infinitival and participial clauses.

5

This movie has been used in various parts of the European Science Foundation (ESF) project (Perdue 1993) and the Structure of Learner Varieties (SLV) project (Hendriks 2005). We employed one of its episodes in our own study.

6

Remember that Lambert et al. (2003) only observed 11.2% of subordinate clauses in the narratives of their German native speakers. The difference between their findings and Chini’s (2003) is puzzling.

7

We are thinking here of all kinds of zero pronouns and ellipses.

8

We use the term “gerund” to refer to the verb form called a “gérondif” in French. It is an invariable form consisting of the verb stem and the ending –ant, identical to the present participle except that it is preceded by the preposition en. When used in sentence-initial position, both the “gérondif” and the present participle have the same subject as the main verb in normative grammar. According to Riegel et al. (1994: 342), they also share the same temporal value: they express simultaneity with respect to the main verb. Syntactically speaking, the “gérondif” resembles an adverbial expression as it takes the role of an adverbial complement expressing simultaneity and sometimes also causality, condition or opposition.

9

Combettes (1998, quoted in Rossi-Gensane 2006) describes detached constructions as “une série d’expressions, différentes d’un point de vue formel (adjectifs, participes, [syntagmes] introduits par une préposition, constructions absolues), mais qui présentent des propriétés communes dont les principales sont : la liberté de position dans la phrase […], la valeur d’un prédicat secondaire qui vient s’ajouter à la prédication principale, la présence d’un référent sous-jacent auquel la construction détachée doit renvoyer” (a series of different expressions (adjectives, participles, [phrases] introduced by a preposition, absolute constructions), which share properties such as: relatively free sentence position […], secondary predication added to the main predication, the presence of an implicit referent the detached construction refers to).

10

The rare cases of detached adjectives were not taken into consideration:

An exploratory study of discourse organisation

279

Une jeune femme, l’air triste, marche dans la rue, le long d’un trottoir animé. (FR L1). ‘A young woman, sad, walks down the street, mingling in an animated crowd.’ 11

In Flanders, French is a compulsory subject in the last two years of primary school and throughout secondary school. In Wallonia, Dutch is not compulsory, as pupils can choose between Dutch, English and German. Therefore, not all the learners of Dutch will necessarily have taken Dutch courses for the same length of time.

12

This means that they were studying (Romance, Germanic, EasternEuropean, Eastern or African) languages.

13

We would like to thank Michel Berré and Carola Henn (Mons), AnneRosine Delbart (ULB) and Liesbeth Degand (UCL) for their help with the collection of the data and for welcoming us into their classrooms.

14

A closer look at the data reveals that some students showed very high frequencies of the elements studied and others very low frequencies. These neutralised each other to a considerable extent. To conduct research on the individual writing style of subjects (cf. Chini 2003), it would be necessary to also consider the learners as native speakers and compare their writings in their mother tongue to their narratives in their L2. This approach has one major drawback, however, which we judged to be possibly disruptive for our study: the subjects would have to see the movie, and perform the writing task, twice. After the second viewing they would be more familiar with the plot, which could influence their selection and planning process, and consequently the complexity of their writing.

15

The degree of complexity can be linked to the amount of detail presented by the visual stimuli and by the simultaneous involvement of many characters in the events.

16

The Dutch-speaking learners of French (FR L2) refer more frequently to the girl than do the other subjects. References to the girl constitute 34% of their references to all five characters, whereas they only make up 27% of the references in the FR L1 corpus, 26% in the NL L1 corpus and 27% in the NL L2 corpus.

17

Bearing in mind Chini’s (2003) observation that an insufficient mastery of referential devices (i.e. an overuse of full NPs and an underuse of pronouns) can lead to a more local treatment of information, which in turn results in a more paratactic discourse organisation, we also conducted a preliminary study on the types of referential devices used. We can conclude from this that, although the data diverge in some specific respects, there was no major overuse of explicit referential expressions

280

Annemie Demol and Pascale Hadermann (suggesting difficulties in the global treatment of information, see also Hendriks 2001) in the learner corpora.

18

The rather small proportion of juxtaposition suggests that Dutch is not really characterised by a choppy style.

19

The Dutch translation of example (2d), repeated below for convenience, seems a little formal and perhaps even slightly archaic. The juxtaposition of independent clauses would come across as more vivid and fast-paced in modern Dutch. Laissant vides chacun des plats sur la table, il se dirige vers la sortie où le récupère le gendarme. (FR L1) ‘Leaving every one of the dishes on the table empty, he goes to the exit where the policeman catches him again.’ ‘Elke schotel leeg achterlatend op de tafel, begeeft hij zich naar de uitgang waar de politieman hem opwacht.’

20

It should be noted, however, that these data were obtained with the help of different visual stimuli.

References Asher, N. and L. Vieu (2005), ‘Subordinating and coordinating discourse relations’, Lingua, 115: 591-610. Bartning, I. (1997), ‘L’apprenant dit avancé et son acquisition d’une langue étrangère. Tour d’horizon et esquisse d’une caractérisation de la variété avancée’, Aile, 9: 9-50. Bartning, I. and N. Kirchmeyer (2003), ‘Le développement de la compétence textuelle à travers les stades acquisitionnels en français L2’, Aile, 19: 9-39. Carroll, M. and M. Lambert (2003), ‘Information structure in narratives and the role of grammaticised knowledge: A study of adult French and German learners of English’, in: C. Dimroth and M. Starren (eds) Dynamics of First and Second Language Acquisition. Amsterdam: John Benjamins Publishing Company. 267-287. Chini, M. (2003), ‘Le phénomène de la jonction interpropositionnelle dans la narration en italien L2 : entre agrégation et intégration’, Aile, 19: 71-106. Chuquet, H. and M. Paillard (1987), Approche linguistique des problèmes de traduction anglais > < français. Paris & Gap: Ophrys. Coppen, P.A., W. Haeseryn, and F. de Vriend (2004), E-ANS (1.1). Katholieke Universiteit Nijmegen. Available online at www.ru.nl/e-ans/content/eans/index.php3 (last accessed on November 28, 2006). Cosme, C. (2005), ‘A corpus-based view on information packaging across languages’. Paper presented at the SPRIK seminar, 10 November 2005,

An exploratory study of discourse organisation

281

University of Oslo, Department of Literature, Area Studies and European Languages. Cosme, C. (2006), ‘Clause-combining across languages: a corpus-based study of English-French translation shifts’, Languages in Contrast, 6(1): 71-108. Cosme, C. (2008), ‘A corpus-based perspective on clause-linking patterns in English, French and Dutch’, in: C. Fabricius-Hansen and W. Ramm (eds) ‘Subordination’ vs. ‘Coordination’ in Sentence and Text from a Crosslinguistic Perspective. Amsterdam: John Benjamins Publishing Company. 89-114. Fabricius-Hansen, C. (1996), ‘Informational density: A problem for translation and translation theory’, Information Structure: A Key Concept for Translation Theory, Special issue of Linguistics, 34(3): 521-565. Fabricius-Hansen, C. (1999), ‘Information packaging and translation: Aspects of translational sentence splitting (German – English/Norwegian)’, in: M. Doherty (ed.) Sprachspezifische Aspekte der Informationsverteilung. Berlin: Akademie Verlag. 175-214. Gilquin, G. (2000/2001), ‘The Integrated Contrastive Model. Spicing up your data’, Languages in Contrast, 3(1): 95-123. Granger, S. (1996), ‘From CA to CIA and back: an integrated approach to computerized bilingual and learner corpora’, in: K. Aijmer, B. Altenberg and M. Johansson (eds) Languages in Contrast. Papers from a Symposium on Text-based Cross-linguistic Studies. Lund 4-5 March 1994. Lund: Lund University Press. 37-51. Hannay, M. and J. L. Mackenzie (eds) (1996), Effective Writing in English: A Resource Guide. Groningen: Martinus Nijhoff. Hendriks, H. (2001), ‘Using nouns for reference maintenance: a seeming contradiction in L2 discourse’, in: A.G. Ramat (ed.) Typology and Second Language Acquisition. Berlin: Mouton de Gruyter. 291-326. Hendriks, H. (ed.) (2005), The Structure of Learner Varieties. Berlin: Mouton-de Gruyter. Kirchmeyer, N. (2003), Etude de la compétence textuelle des lectes d’apprenants avancés. Aspects structurels, fonctionnels et informationnels. PhD dissertation, University of Stockholm, Cahiers de la recherche 17. Lambert, M. (1997), ‘En route vers le bilinguisme’, Aile, 9: 147-172. Lambert, M., M. Carroll and C. von Stutterheim (2003), ‘La subordination dans les récits d’apprenants avancés francophones et germanophones de l’anglais’, Aile, 19: 41-69. Lehmann, C. (1988), ‘Towards a typology of clause linkage’, in: J. Haiman and S.A. Thompson (eds) Clause Combining in Grammar and Discourse. Amsterdam & New York: John Benjamins Publishing Company. 181-225. Perdue, C. (ed.) (1993), Adult Language Acquisition: Cross-linguistic Perspectives. 2 volumes. Cambridge: Cambridge University Press. Riegel, M., J.-C. Pellat and R. Rioul (1994), Grammaire méthodique du français. Paris: Presses Universitaires de France.

282

Annemie Demol and Pascale Hadermann

Rossi-Gensane, N. (2006), ‘À propos des constructions détachées en rupture comportant une forme verbale non finie, et plus particulièrement des participes’, Travaux linguistiques du Cerlico, 19: 131-146. Vanderauwera, R. (1985), Dutch Novels Translated into English: The Transformation of a “Minority” Literature. Amsterdam: Rodopi. Watorek, M. (1996), ‘Le traitement prototypique : définition et implications’, Toegepaste Taalwetenschap in Artikelen, 55: 187-200. Wilmet, M. (1997), Grammaire critique du français. Louvain-la-Neuve: Duculot.

Involvement features in writing: do time and interaction trump register awareness? Annelie Ädel English Language Institute University of Michigan Abstract Speaking is typically characterised by linguistic markers of involvement, and writing by detachment, as noted by Chafe (1982). However, differences between speech and writing should not be understood as a dichotomy, but rather as a continuum along which texts, whether spoken or written, vary. Sometimes the degree of “writtenness” or “spokenness” varies inappropriately – a phenomenon often witnessed in foreign language learning. Research on the SWICLE, the Swedish subcorpus of the International Corpus of Learner English (ICLE), has shown that learners tend to produce an overly involved style, the stated reason being poor awareness of register (Altenberg 1997). However, through intercorpus comparison, it was found that, instead, this is primarily due to task setting (time available) and intertextuality (access to secondary sources). The findings are connected to previous observations that time available and possibilities for interaction profoundly influence linguistic output (Chafe 1986). Involvement markers were examined in the timed and untimed sections of the SWICLE and in the Uppsala Student Essay (USE) corpus of untimed English essays by Swedish speakers with access to secondary sources. Thus, the examined corpora differed precisely in task setting and intertextuality. This three-way comparison showed that the learners exhibit more involvement in timed than in untimed essays, but less if they have access to other texts. The paper closes by addressing the important question of corpus comparability and by making recommendations for testing practices in language teaching.

1.

Introduction

One of the classic topics in linguistics concerns the difference between spoken and written language. Linguists such as Chafe (1982), Halliday (1985) and Biber (1986, 1988) have studied specific grammatical features that mark a stretch of connected language as speech or writing, considered from the point of view of the general communicative and situational conditions that affect speech and writing. Biber (1988) compares speech and writing in terms of involved versus informational production, situation-dependent versus elaborated reference, and non-abstract versus abstract style, while Halliday (1985) “contrasts the prosodic features and grammatical intricacy of speech with the high lexical density and grammatical metaphor of writing” (Roberts & Street 1997: 168). Of special interest here is the work of Chafe (1982), Chafe & Danielewicz (1987) and Tannen (1982), describing speech and writing as differing in terms of involvement and detachment. According to these researchers, informal speaking is typically

36

Annelie Ädel

characterised by linguistic markers of involvement, such as first person reference and emphatic particles, and formal writing by markers of detachment, such as passive constructions and abstract subjects. Speakers, according to Chafe (1982: 45), typically exhibit involvement with their audience because they tend to have face-to-face contact. This results in shared knowledge about the environment of the conversation; the speaker being able to monitor effects of the discourse and the listener being able to signal understanding or request clarification; and the speaker being concerned with experiential richness rather than consistency or comprehensibility at a different time and in a different place. Writers, on the other hand, typically have a fundamentally different relation to their physically absent audiences, which, according to Chafe, leads to different linguistic effects: primarily, an increased use of detachment features. This is not to say that differences between speech and writing should be understood as a dichotomy; they should rather be conceived of as a continuum along which texts, whether spoken or written, vary. Consider, for example, the place on the continuum of a patent text versus that of an electronic chat, although these are both written forms. Similarly, consider the place on the continuum of a face-to-face conversation between friends versus that of an academic lecture. Instead of portraying speech and writing as static systems encoded by different media, Chafe draws attention to the influence of contextual factors, such as shared time and space, in all language production. Table 1 gives a short and non-exhaustive list of some typical types and features of involvement. It is taken from Altenberg (1997), but is ultimately based on Chafe (1982). Table 1. Types and features of involvement (examples from the SWICLE) 1st pers sg pronouns Disjuncts Questions Exclamations Discourse markers

I, me, my, mine frankly, obviously, of course, naturally, perhaps, maybe, unfortunately e.g. Does this all sound utopian to you?; And that is not what we want, is it? e.g. Terrible, I know!; Look at society!; But let’s hope for peace in Europe! well, you see

Involvement features can be described as ways in which speakers/writers show involvement with their audience. First person singular pronouns mark the presence of the speaker/writer and explicitly show that she is the agent behind the discourse.1 Questions tend to be explicitly audience-oriented and generally contribute to making a discourse more interactive. Exclamations can be divided into addressee-oriented directives, primarily used to instruct someone to do something, and addresser-oriented expressive exclamations, primarily used for

Involvement features in writing

37

expressing the extent to which she is impressed by something (see Biber et al. 1999: 202). The list of disjuncts includes hedges (perhaps, maybe), emphatic markers (obviously, naturally, of course) and opinion markers (frankly, unfortunately), displaying the speaker/writer’s degree of certainty, and her attitude and stance taken towards what is said. The short list of discourse markers (well, you see) includes expressions that are very frequent in conversation. This paper will summarise a study of these involvement features in written texts. My immediate reason for choosing these particular parameters to measure stylistic variation is that they have been used in previous research (Altenberg 1997), on which the present work builds. 2.

Communicative competence in a foreign language

Speakers and writers learn how to manage features of involvement and detachment as part of their communicative competence, which includes knowledge of style and genre variation.2 Features of involvement are learnt first, while features of detachment are adopted at a later stage. Language users typically do not learn to produce features of detachment in any great numbers or in a systematic way until they develop school-level literacy. In actual language use, the degree of “writtenness” or “spokenness” sometimes varies inappropriately; this phenomenon is often witnessed in the context of foreign language (FL) and second language writing. Many studies have shown that learner writing tends to exhibit high frequencies of linguistic features that are more characteristic of the spoken language rather than the written language of native speakers (e.g. Altenberg 1997, Granger & Rayson 1998, Petch-Tyson 1998, Hinkel 2003). In the following, I will focus on the use of involvement features in FL writing, because the question of how learners manage this variation is an important one, seeing as how this can be a challenge even to native speakers, especially during the development of academic literacy. It hardly needs to be pointed out that inability to vary one’s language use according to context is likely to result in considerable difficulties in communication in all domains, not to mention low assessment scores in the educational domain. Managing linguistic variation is an especially burning issue in the case of formal writing. As pointed out by Howard (2004: 146), learners’ linguistic output is influenced by factors like type of classroom input and authentic target-language contact outside the classroom. However, the type of language to which students are exposed may vary widely along the formal-informal continuum, even in the classroom, depending on the pedagogical practices in use. For example, it has been argued that, in some countries, the pedagogical pendulum has swung too far toward the informal end in EFL instruction (English as a Foreign Language; Altenberg 1997). The type of learner in focus here is a relatively advanced university student of English. The difficulties encountered by such learners are likely to be

38

Annelie Ädel

mainly lexical and stylistic rather than grammatical (Granger 1993: 59). This study concentrates on argumentative writing by this learner group, since this is a highly important text type in many domains of society. The students are writing in English, one of the languages most widely used by non-native speakers. Although the material under consideration consists of written texts, the focus will be on the degree to which these texts exhibit features of involvement, which are characteristic of spoken language. Table 2 illustrates (1) a relatively successful positioning along the continuum of involvement versus detachment and (2) a less successful attempt. The examples come from two different collections of argumentative essays, and they both deal with the topic of abortion. Table 2. Successful versus less successful written examples by Swedish learners Successful attempt (1) Although the question of abortions cannot be regarded in just black or white, the reasons in favour of this issue clearly overweigh those against and the right to abortions should therefore be defended and supported. (USE 1097.b2)

Less successful attempt (2) Personally I think it is a woman’s right to have one, if she feels she can’t afford bringing it up or if she already has five children. However, I don’t think abortions […]. (SWICLE 339)

Let us begin with the less successful attempt, example (2), which is taken from an essay in the International Corpus of Learner English (ICLE; Granger et al. 2002). It represents a style typical of advanced learners of English whose first language is Swedish. 3.

The Swedish subcorpus of the ICLE

Since many of the other chapters of this volume describe the ICLE in some detail, it will be treated rather cursorily here. However, one fact about the corpus that needs to be emphasised is that, since the ICLE writers are all learners of English at an advanced level, the focus of research is primarily on overuse and underuse of linguistic phenomena as compared to equivalent native-speaker production, rather than on grammatical errors. The ICLE material makes it possible to pinpoint which linguistic features are used considerably more or less by learners than by native speakers.3 The general idea is that “[w]hen matched with comparable native-speaker texts, a learner language corpus provides the basis for revealing the characteristics of learner language” (Johansson 1998: 7), such as interference from the mother tongue (L1) or possible universal features of learner language. The ICLE material includes several subcorpora of argumentative essay writing by university students from more than a dozen different L1 backgrounds. The focus of this chapter is on the Swedish subcorpus of the ICLE (the

Involvement features in writing

39

SWICLE), which consists of 350 argumentative essays with a total of 205,000 words. The essays are all full-length, with an average size of 560 words. The student writers were in their third or fourth year of university-level English studies at three different universities in Sweden. Although they tend to have relatively high proficiency levels in English, most students studying English at Swedish universities do not get much systematic training in argumentative or academic writing, nor have they had much input before entering university.4 More details on the SWICLE relevant to the present study will be given in Section 7 below. Next, we will look more closely at the high degree of involvement found in the SWICLE. 4.

The Swedish learner style

Previous research based on the SWICLE (Altenberg 1997, Petch-Tyson 1998) has shown that learner writers, instead of a detached, “writerly” style typical of argumentative text, as in (1) above, tend to produce an overly involved style, as in (2) above. This trend has been found to be quite persistent by comparison with native-speaker control corpora of argumentative texts.5 Examples (3) and (4), excerpts from the concluding sections of two SWICLE essays on different topics, serve to illustrate further the high degree of involvement of the learner style: (3)

(4)

The presentation I have made with science and the dreams of ours walking hand in hand on a very exciting road towards a “happy ever after” kind of story, does seem a bit glorifying, I admit. But my general opinion, in conclusion, is the same one as I introduced in this little essay: No, I do not agree with the statement that science dominate our world too much, so that there will be no space for dreaming - there is plenty of space! (SWICLE 075) I am in favour of a school which includes six-year-olds. The reason is that I think that most children are ready for it by then. Of course, I don’t think this early start should be too harsh but a playful and gentle period which creates space for the children to adapt to the new world they are entering. There, they must be stimulated to learn, taught how to show respect to other people and be treated with care. It may sound like an utopia but I am convinced that it could work given the right resources. Finally I would like to say that I believe that a lot of the social problems of today have to do with people’s childhood. Feeling safe and cared for as a child, creates better opportunities in life and the responsibility lies on us as parents or as a society. (SWICLE 334)

In these examples, the features listed in Table 1 have been underlined to give an idea of how they are used in the SWICLE.6 In (3), first person singular pronouns are used repeatedly, with four occurrences of I and one of my, and the excerpt

40

Annelie Ädel

ends with an exclamation. In (4), there are six occurrences of first person singular pronouns, and one occurrence of of course. Although the underlined parts do not tell the whole story as far as “spoken” style is concerned, they do show that the density of involvement features is considerable. Altenberg (1997: 123), in a paper investigating precisely the involvement features listed in Table 1 in a subset of the SWICLE material, as well as some detachment features represented by passives and abstract subjects, found that the Swedish learner style is “more typical of informal spoken registers than of argumentative writing”. This “spoken style” was manifest at all linguistic levels: lexis, grammar and discourse, leading him to conclude that the “common basic deficiency among the Swedish learners [is] a style of writing that is greatly influenced by spoken language” (ibid. 130). The reason for this is taken to be poor awareness of register, caused by inadequate teaching and over-exposure to informal, spoken registers. 5.

Reasons for over-involvement

Having studied the SWICLE material in great detail myself, I had come to the same conclusion about the exaggerated level of involvement features in the students’ writing. I had used SWICLE for research on metadiscourse, or selfreflective language (Ädel 2006), only to find that Swedish learners overused personal types of metadiscourse considerably more than impersonal types of metadiscourse – which also confirmed the personal and informal style of the SWICLE. In the course of the research, I attempted to list other possible reasons for this tendency, assuming that there could be other factors accounting for this trend among SWICLE writers, and that Altenberg’s poor-register-awareness explanation might not be telling the whole story. In addition to (a) register awareness, other relevant factors were suggested to be (b) cultural conventions, (c) general learner strategies and (d) genre comparability (Ädel 2006). Table 3 below distinguishes among and sums up the four factors. The first column names the factors, the second column gives a short-hand label for the main argument behind each factor, and the third column briefly explains the source of the problem. The groups mentioned in the explanation refer to (i) learner writers in the SWICLE and (ii) a native-speaker control group based on the university student writing in LOCNESS (see footnote 5). In the case of involvement, factor (a), register awareness, has been claimed to explain the divergence of advanced Swedish learner writing from nativespeaker writing. There are other possibilities, however, one of which is factor (b). The SWICLE writers’ overuse of involvement features may be influenced by writing conventions specific to Swedish. Informality in writing has been observed in many different areas in Swedish, in particular in public discourse, and several studies have shown that written Swedish has become more similar to spoken Swedish (e.g. Mårtensson 1988, Svensson 1993, Lagerholm 1999). Another way of explaining the differences along the learner versus native-speaker axis is to

Involvement features in writing

41

attribute them to “learner language”. Factor (c) refers to the existence of universal strategies employed by those writing in a foreign language that may still “shine through” despite differences in instruction, input and L1 transfer. Since the design of the present study provides no evidence for factors (b) or (c), these must remain unresolved for the time being. Instead, this study will focus on the possible influence of factor (d), genre comparability. The question to be answered is whether different conditions of writing for the texts collected in different corpora makes comparing the corpora fundamentally problematic. Table 3. Overview of four explanations for linguistic differences at the discourse level between learner and native-speaker corpora (based on Ädel 2006) Factor

Argument

Explanation

(a) Register awareness

“instructional”

(b) Cultural conventions

“typological”

(c) General learner strategies

“universal claim”

(d) Genre comparability

“apples & oranges”

Differences found between groups depend on background knowledge, primarily acquired through instruction, but also through general exposure. Differences found between groups are due to differing conventions in the L1 and the L2, which cause interference in L2 output. Differences found between groups arise because, when writing in a foreign language, writers employ universal strategies to cope. Differences found between groups are due to the corpora under comparison not being comparable, and can be explained by external factors like time given for planning or producing a text.

6.

Genre comparability

There are two specific phenomena subsumed by the concept of “genre comparability” that can be shown to play a role here. They will be referred to as task setting (or time available), meaning whether an essay was timed or untimed, and intertextuality, meaning whether the writer had access to secondary sources (in the form of newspaper or journal articles) to use as a starting point for her argumentation and to refer to in her own text. How might task setting and intertextuality have an impact on writing style? Linguists have observed that, in language production, the amount of time available and the possibilities for interaction profoundly influence the linguistic output (see e.g. Chafe 1986). In fact, one of the main differences between writing and speaking is that “a writer has much more time than a speaker to deliberate on

42

Annelie Ädel

what is being said” (Chafe 1986: 262). In other words, whether or not language producers have ample time to plan and edit their linguistic production makes a big difference. We have to assume, then, that even when writing, the amount of time one has available can make a difference to the output. Language production is also affected by interaction; indeed, this is a basic tenet of sociolinguistics. A spoken conversation typically involves direct participation on the part of two or more speakers. In written discourse, by contrast, there is no time or space shared by the writer and the reader. However, the availability of dialogue should not be seen as a yes-or-no proposition; although the written medium does not accommodate synchronous interaction, features typical of dialogue still occur in written text. In a dialogic written text such as a letter to a friend, the writer typically depicts forms of interaction between writer and reader, even though flesh-and-blood readers are not present at the time of writing to make concrete contributions to the discourse. Indeed, even in seemingly monologic texts, such as research articles, recent research has shown that a great deal of attention is paid to the audience. Academic writing is characterised as fundamentally interactive, for example by Hyland (2005: 11), who argues that “[m]anaging social relationships is crucial in writing because a text communicates effectively only when the writer has correctly assessed both the readers’ resources for interpreting it and their likely response to it”. Furthermore, when interaction is involved in language production, whom one interacts with will also affect the output. Even though there is no interactant directly present in the case of writing, the writer will still have a specific audience in mind, consisting of actual or imagined readers. In the production of a written text, the actual reader is never present to submit an utterance or reply directly, but the imagined reader will still affect the discourse and leave his mark. Bhatia (1993: 9) uses the term “hypothetical reader” for the notion that, in written discourse, the writer assumes a hypothetical reader for whom she is writing. She anticipates the reactions of this reader and adjusts her writing accordingly, in order to facilitate communication. A writer always has an imagined reader in mind, which may even be herself, as in the case of a personal diary. In addition to the imagined reader, another influence that typically affects a writer’s output is the presence of secondary source texts, such as an editorial or a journal article. When an essay writer uses secondary sources for her text, she will react to and relate to these and to the writer personae of those texts. That is, the writer is likely to accommodate to these sources of input and adapt to their style, leading to a more professional-sounding essay. Another way in which other texts can influence written production is through imitation: a writer who is exposed to model texts will have a better idea of what to aim for. I will support here the idea that a student’s writing style is likely to benefit from access to other texts in the same register.

Involvement features in writing 7.

43

Inter-corpus comparison

Seeking to investigate whether these factors actually do have an effect on the overuse of involvement in Swedish learner writing, I was fortunate enough to locate another corpus that was fully comparable with the SWICLE, with the exception precisely of task setting and intertextuality. The other corpus is the Uppsala Student Essay corpus (USE; Axelsson 2000), a collection of essays by university students with L1 Swedish and advanced L2 English.7 For this study, I used a subset of the USE, a part that specifically consists of argumentative essays.8 Table 4 illustrates the main similarities and differences between the USE subset and the SWICLE. Table 4. Comparison between USE and SWICLE USE Home university of students Student level Topics Exam conditions Secondary sources

Uppsala University 2nd term students Set argumentative topics Untimed (2-3 week deadline, out of class) Access to written sources for facts

SWICLE Universities of Göteborg, Lund, Växjö 3rd and 4th term students Set argumentative topics Some timed sit-down exams, some untimed No input other than personal experience

Looking at the data from the point of view of Bhatia’s (1993: 23ff.) external criteria of genre, we can classify the included texts as belonging to the same genre. These criteria include temporal factors (time of composition and age of writers), sociological factors (educational background of writers), factors that depend on the writer and reader (intended audience, writer-reader relationship and goals of the writer), and comparability of topics.9 All essays were written within a span of about a decade, although the time of composition of the two corpora differs slightly (SWICLE from 1990-1999 and USE from 1999–2001). Although the USE students were somewhat less experienced, the range of student status is still relatively restricted: from second to fourth term students (i.e. a range of less than 1.5 years of study). The writers represented in the corpora were all full-time students in English departments at different (but comparable) Swedish universities. They all wrote on set argumentative topics, with the exception of a small group of SWICLE students who wrote on a literary or general cultural topic. Crucially, and fortunately for the present purposes, the main points of divergence are found precisely in access to secondary sources and in exam conditions. The USE writers were instructed to refer to editorials or newspaper articles dealing with topical issues (e.g. gay couples’ right to adopt children), and to present counter-arguments to views expressed in these sources. The sources

44

Annelie Ädel

were all in Swedish, which arguably made for a more demanding languageprocessing task than if they had used English-language sources. Since the SWICLE is a mixture of timed and untimed essays, it was split into two subcorpora on the basis of the timed vs. untimed criterion. These will be referred to as SWICLE Timed and SWICLE Untimed. In effect, this produced three different subcorpora involving argumentative essay writing in advanced L2 English by university students whose L1 is Swedish, enabling a three-way comparison. Table 5 summarises the sizes and the distinguishing features of the three corpora. Given the differences between the three subcorpora, we can expect them to display a stepwise pattern in the amount of involvement features they display, with SWICLE Timed having the most involvement features because the essays were written under strict time constraints, and with no access to other texts, followed by SWICLE Untimed, for which the time constraint is absent, and then with the USE having the fewest involvement features, because these essays were untimed and the writers had access to other texts. Table 5. Sizes and features of the three subcorpora

Untimed essay? Access to other sources? Number of essays Number of words 8.

SWICLE Timed NO NO 162 87,678

SWICLE Untimed YES NO 121 70,670

USE YES YES 57 50,342

Results

The same markers of involvement used by Altenberg (1997), listed in Table 1, were investigated in the three subcorpora. The frequencies of the different involvement features are shown in Table 6. Items marked with a dagger are not sufficiently frequent to be statistically meaningful. The statistical significance of the differences between the corpora is given in Table 7 below. If we consider first the first person singular pronouns, SWICLE Timed has almost a third more occurrences of I than either SWICLE Untimed or USE. Similarly, both me and my are considerably more frequent in SWICLE Timed than in SWICLE Untimed and USE. The totals for disjuncts, by contrast, show a different picture from the pronouns: although SWICLE Timed has the greatest frequency, there is a clear difference between SWICLE Untimed (at 22.35) and USE (at 13.51). Most of the frequencies for the individual words follow the general pattern, although there are exceptions (frankly, obviously and unfortunately). As with the disjuncts, we find a close match between the two SWICLE corpora with respect to both questions and exclamations, while the USE corpus has considerably lower frequencies. For the selected discourse markers, finally, we find very low overall numbers, suggesting that this is not a major problem area for Swedish learners.10

Involvement features in writing

45

Table 6. Frequencies of involvement features, normalised per 10,000 words

1st pers sg pron: I me my mine† Disjuncts: of course naturally perhaps maybe unfortunately obviously frankly† Questions Exclamations Discourse markers: well you see† Total

SWICLE Timed

SWICLE Untimed

USE

102.65 13.46 25.43 0.34

68.91 3.96 8.91 0.42

76.48 4.17 7.55 0

11.06 2.62 6.62 3.54 2.05 1.14 0 31.25 8.10

7.22 1.98 5.80 3.25 3.54 0.42 0.14 30.28 8.07

5.36 0.60 3.38 0.79 1.39 1.99 0 20.86 0

1.60 0.46 210.32

2.83 0.13 145.86

0.20 0 122.77

Let us next restrict the discussion to the totals for the four types of involvement features with the highest frequencies (that is, excluding the discourse markers), presented in Table 7 below. The arrows and grey highlighting indicate the differences between the corpora that are significant (p<0.001) according to the chi-squared “goodness-of-fit” test. Dashes, by contrast, indicate differences that are not statistically significant. Table 7. Total types of involvement features, normalised per 10,000 words, and indications of statistical significance of differences between corpora

1st person singular Disjuncts Questions Exclamations Total

SWICLE Timed 141.88 27.03 31.25 8.10 208.26

Stat. sig. Ù -

SWICLE Untimed 82.21 22.35 30.28 8.07 142.91

Stat. sig. Ù Ù Ù

USE 88.20 13.51 20.86 0 122.57

The “Total” row in Table 7 displays the general stepwise pattern predicted: SWICLE Timed has the highest frequency of involvement features, while

46

Annelie Ädel

SWICLE Untimed has a lower frequency, and USE has the lowest. Overall, the learners exhibit less involvement in untimed essays, and less involvement when they have access to other texts. If we look more closely at Table 7, we see that the time and intertextuality factors actually have different influences. When we vary only the timed-untimed factor, only the use of first person singular pronouns is affected significantly. In other words, the less time the Swedish learners had when writing, the more first person pronouns they used, although the use of the other involvement features was not affected. On the other hand, when we vary only the intertextuality factor, the reverse pattern is found: the use of first person pronouns is not affected, but the use of disjuncts, questions and exclamations is affected significantly. In other words, when the Swedish learners had access to other texts, they used significantly fewer involvement features, with the exception of first person pronouns. It would appear, then, that the two factors examined have somewhat different effects on writing, although they both affect features that fall under the general heading of involvement features. More research is needed to determine whether these effects hold true for other groups of writers in other contexts. Furthermore, a word of caution concerning the potential influence of instruction is in order: the possibility exists that the USE writers were instructed to avoid features like disjuncts, questions and exclamations, such that instruction rather than (or in addition to) access to other texts is responsible for the less frequent use of involvement features. Only with more controlled studies will we be able to draw firm conclusions about the exact influence of the various factors. To sum up the findings so far, we began with the observation that Swedish learners represented by the SWICLE fail to produce the appropriate formal, written style in their argumentative writing. Instead, these writers display high levels of involvement features in their argumentative essays. The findings indicate that the degree of involvement in learner writing is reduced considerably (i) when learners are given untimed essay tasks, and (ii) when they are given secondary sources as input.11 At this point, I would like to point out that the successful example (1) in Table 2 above was taken from the USE corpus. This illustrates that, given the right circumstances, Swedish learners are able to produce relatively successful pieces of argumentative discourse. Interestingly, although this needs to be corroborated by future studies, the findings also suggest that the types of involvement feature used depend differently on the two factors, such that (i) untimed essay tasks reduce the number of first person pronouns, and (ii) tasks that include input from secondary sources reduce the number of disjuncts, questions and exclamations. The conclusion to be drawn is that the overuse of involvement features cannot just be attributed to lack of register awareness, although our Swedish advanced learners of English no doubt have a great deal to learn in this domain. As demonstrated in Altenberg’s study (1997: 129), these involvement features are present in native-speaker writing to a considerably lower degree than in the SWICLE, and the results of the present study demonstrate that even the USE

Involvement features in writing

47

writers use such features more frequently than native writers. With the exception of exclamations, which occur neither in the LOCNESS nor in the USE, the involvement types listed in Table 1 above are more frequent in the USE. This suggests that the register awareness hypothesis is still valid to some degree. In addition to acknowledging the register awareness factor, however, we must conclude that the extreme rate of involvement found in the SWICLE is due to the fact of the writers not having enough time to make the text formal and “writtenlike” and not having access to model texts, or other writers’ assertions, facts, or expert opinions, as a source to draw on and react to in their own writing. 9.

Task setting and implications for corpus comparison

The findings presented here have serious implications for comparison between corpora. First of all, they highlight the importance of corpus comparability in general. While it is understandable that a researcher may make do with an existing corpus that is “in the ballpark” instead of compiling a new corpus to specification, we need to be fully aware that it can make a tremendous difference not to control fully for aspects such as register and conditions of writing. Secondly, the findings have implications for comparability within the ICLE with respect to the task setting factor. On the corpus CD, the ICLE team at Louvain-la-Neuve, fortunately, provides a great deal of information about the texts as well as the writers. By using the ICLE software, users can build any kind of available metadata into their searches; one of the available parameters is timed versus untimed. If we examine this parameter across the ICLE subcorpora, the general picture that emerges is that there is a great deal of variation across the subcorpora. In fact, three different groups can be discerned, as shown in Table 8. Table 8. Percentages of untimed texts in the ICLE subcorpora Percentage of untimed texts Group A

80-100%

Group B Group C

~ 65% ~ 35%

Countries represented Austria, Belgium, Bulgaria, Czech Rep., Netherlands, Russia, Spain, Switzerland Finland, Poland Germany, Italy, Sweden

In Group A, the majority of the essays (80-100%) are untimed, while in Group C, a small part of the essays (35%) are untimed. Group B, in the middle, has approximately 65% untimed essays. The fact that the SWICLE is in the group with the smallest proportion of untimed essays suggests that we should be cautious in making comparisons between it and subcorpora from the other groups.12 In fact, having seen that time setting can have a considerable impact on involvement features in writing, at least on first person singular pronouns, we should be careful when comparing any similar features across the three groups.

48 10.

Annelie Ädel A previous study seen in new light

Having gathered all these data, I took a second look at an excellent study by Petch-Tyson (1998) on writer/reader visibility in four of the ICLE subcorpora: those with L1 Swedish, Finnish, Dutch and French. Writer/reader visibility, which partially overlaps with involvement, is defined by Petch-Tyson (ibid. 108) as a set of features in writing “used to express personal feelings and attitudes and to interact with readers”.

Figure 1. Writer/reader visibility ranking matched to timed/untimed groups Petch-Tyson’s results were that L1 Dutch and French speakers had by far the lowest degree of writer/reader visibility, Finnish ended up in between, and Swedish had by far the highest degree. If we match these results to the three different groups based on timed/untimed essay proportion, an interesting pattern emerges, as shown in Figure 1. The patterning is exactly as we might predict based on the results of this study: whether an essay is timed or untimed influences the presence and proportion of certain linguistic features. 11.

Intertextuality

Intertextuality, i.e. whether secondary sources (not counting language reference tools, such as dictionaries) are drawn upon and referred to in a text, appears to be uniformly absent from the ICLE subcorpora. The general instructions given by the ICLE team were that the texts included should be “entirely the students’ own work, i.e. they should not draw on other articles, books for the essay”.13 This means that intertextuality cannot account for the differences across ICLE

Involvement features in writing

49

subcorpora. It could, however, partly account for the great differences found between the SWICLE and the LOCNESS native-speaker control corpus (see Altenberg 1997: 129). Although the native-speaker corpus is rather undocumented in this respect, it is clear from the texts themselves that writers received some input on the topic prior to the composition and had access to other sources, such as magazines and newspaper articles, while writing. Furthermore, it is hard to imagine how the lack of secondary sources could not have influenced the ICLE writing. As noted in Ädel (2006: 143), one of the effects of the SWICLE writers’ lack of access to supporting materials is that the argumentative worlds of the learner writers ended up being very much limited to themselves and their own experiences, which tended to produce an informal and narrative style rather than a formal and argumentative one. As for the differences between the SWICLE and the USE, the results presented here suggest that intertextuality accounts, at least in part, for the discrepancies found. The fact that the USE students did have access to and did refer to other texts kept their levels of involvement considerably lower than the SWICLE Untimed for most of the features examined. More research is needed to clarify the important issue of what the linguistic effects are of having secondary sources to refer to and draw on during a writing task. 12.

Pedagogical recommendations

An important pedagogical question raised by Altenberg’s study is how we can help Swedish learners – or any group of students, as the case may be – to refrain from overusing involvement features in their argumentative writing, and make their texts stylistically more target-like. The present study has made it clear that not only do students need better instruction on and experience with register, but they also need to be given the right setting when producing texts. It is hardly fair to give students a timed writing task without any other texts for input or modelling, and then expect them to perform like professionals or native speakers. Judging from the present results, two specific recommendations can be made for testing practices in language teaching. The first is that, for any formal writing task, untimed is preferable to timed. In addition to giving the writers time to work on their style and overall composition, this would also create a more “authentic” or “naturalistic” setting. Another argument supporting untimed writing is found in a study by Kroll (1990), showing that ESL writers’ (i.e. writers for whom English is a Second Language) rhetorical competency was 50% higher when writing at home compared to writing in class. Of course, it needs to be acknowledged that giving an untimed writing test creates a host of problems having to do with authenticity and plagiarism; it is impossible to monitor individual students and prevent them from consulting unauthorised sources. Perhaps the ideal solution for composition teachers would be to navigate between the Scylla of fear of plagiarism and the Charybdis of nonnaturalistic timed writing by giving a combination of timed and untimed tests.

50

Annelie Ädel

Researchers into writing should then provide support for the grading teacher as to how to best compare and assess timed and untimed products. The second recommendation is that essay writers should be given other texts as input and to serve as a point of departure for their writing. There is a body of recent research stressing the fact that “academic writing rarely occurs as an isolated act but rather is often in response to the content of a course” or other material (Cumming et al. 2000: 9). Furthermore, if essay writers were allowed to draw on other texts and quote secondary sources, their general argumentation and their writing style would be likely to improve. This would reduce the number of essays with a very high involvement factor and help writers to use a more argumentative and less narrative style. Successful argumentative writing should be “polyphonic”, in the sense that arguments and opinions by others, whether quoted or rephrased, are expected to be brought in. Without references to other sources, which bring in other voices (such as facts and arguments from a different perspective), an argumentative text risks giving the impression of having been written in a vacuum, drawing exclusively on the single writer’s experience. Perhaps the role of intertextuality is more important to successful argumentative writing than we have previously assumed? Notes 1

Throughout this chapter, speakers/writers will be referred to with feminine pronouns, and hearers/readers with masculine pronouns.

2

The term “communicative competence”, coined by Dell Hymes (see e.g. Hymes 1992), refers to the notion that speakers of a language need to know not just grammatical forms and structures, but also how language is used by members of a speech community to accomplish their purposes.

3

What kind of native-speaker corpus to use as a basis for comparison, however, is not a straightforward matter (see e.g. Ädel 2006: 205ff.).

4

See http://www.englund.lu.se/corpus/corpus/swicle.html (last accessed in August 2006), where the SWICLE is described in more detail.

5

Both Altenberg’s and Petch-Tyson’s control corpora were the LOCNESS, short for the Louvain Corpus of Native English Essays, compiled by the Centre for English Corpus Linguistics. It consists of mainly argumentative essays written by native-speaker students in the US and in the UK.

6

There are, of course, several other features in examples (3) and (4) that contribute to an overly involved style (e.g. kind of, a bit, this little essay), but only the underlined elements were taken into account for the present study.

Involvement features in writing

51

7

The USE was compiled by Margareta Westergren Axelsson and Ylva Berglund in the English Department at Uppsala University between 1999 and 2001.

8

This subset is referred to as “b2” by the compilers.

9

With respect to the last criterion, it must be said that even small differences in prompts or essay topics could have an effect on the texts, as shown by Hinkel (2002: 162-164). However, examining prompts used in these corpora and their possible influence on involvement features is beyond the scope of the present study.

10

Suffice it to report here that, in the case of well, the difference between SWICLE Untimed and the USE reaches statistical significance.

11

Although it is possible that the USE writers cheated and received input from highly experienced writers or native speakers of English, it is unlikely that a large enough proportion of the students did so to affect the entire sample. Furthermore, having read the essays myself, I do not see any evidence of plagiarism in the corpus.

12

Needless to say, this also goes for the German and Italian subcorpora. It would be interesting to see a future comparison of the use of involvement features in the other two subcorpora of Group C, the German and Italian subcorpora, to see how their results compare to those of the SWICLE.

13

Information retrieved from http://cecl.fltr.ucl.ac.be/Cecl-Projects/Icle/ icle.htm (last accessed in August 2006).

References Ädel, A. (2006), Metadiscourse in L1 and L2 English. Amsterdam & Philadelphia: John Benjamins Publishing Company. Altenberg, B. (1997), ‘Exploring the Swedish component of the International Corpus of Learner English’, in: B. Lewandowska-Tomaszcyk and P.J. Melia (eds) PALC’97: Practical Applications in Language Corpora. Lódz: Lódz University Press. 119-132. Axelsson, M.W. (2000), ‘USE – The Uppsala Student English Corpus: An instrument for needs analysis’, ICAME Journal, 24: 155-157. Bhatia, V.J. (1993), Analysing Genre: Language Use in Professional Settings. London: Longman. Biber, D. (1986), ‘Spoken and written textual dimensions in English: Resolving the contradictory findings’, Language, 62(2): 384-414. Biber, D. (1988), Variation Across Speech and Writing. Cambridge: Cambridge University Press. Biber, D., S. Johansson, G. Leech, S. Conrad and E. Finegan (1999), The Longman Grammar of Spoken and Written English. London: Longman.

52

Annelie Ädel

Chafe, W. (1982), ‘Integration and involvement in speaking, writing and oral literature’, in: D. Tannen (ed.) Spoken and Written Language: Exploring Orality and Literacy. Norwood, NJ: Ablex. 35-53. Chafe, W. (1986), ‘Evidentiality in English conversation and academic writing’, in: W. Chafe and J. Nichols (eds) Evidentiality: The Linguistic Coding of Epistemology. Norwood, New Jersey: Ablex. 261-272. Chafe, W. and J. Danielewicz. (1987), ‘Properties of spoken and written language’, in: R. Horowitz and J. Samuels (eds) Comprehending Oral and Written Language. San Diego: Academic Press. 83-113. Cumming, A., R. Kantor, D. Powers, T. Santos and C. Taylor (2000), TOEFL 2000 Writing Framework: A Working Paper. TOEFL Monograph Series MS-17. Princeton, NJ: Educational Testing Service. Granger, S. (1993), ‘The International Corpus of Learner English’, in: J. Aarts, P. de Haan and N. Oostdijk (eds) English Language Corpora: Design, Analysis and Exploitation. Amsterdam: Rodopi. 51-71. Granger, S., E. Dagneaux E. and F. Meunier (2002), The International Corpus of Learner English. Handbook and CD-ROM. Louvain-la-Neuve, Belgium: Presses Universitaires de Louvain. Granger, S. and P. Rayson. (1998), ‘Automatic profiling of learner texts’, in: S. Granger (ed.) Learner English on Computer. London & New York: Longman. 119-131. Halliday, M.A.K. (1985), Spoken and Written Language. Oxford: Oxford University Press. Hinkel, E. (2002), Second Language Writers’ Text. Mahwah, NJ: Lawrence Erlbaum. Hinkel, E. (2003), ‘Simplicity without elegance: Features of sentences in L1 and L2 academic texts’, TESOL Quarterly, 37(2): 275-301. Howard, M. (2004), ‘Sociolinguistic variation and second language acquisition: A preliminary study of advanced learners of French’, SKY Journal of Linguistics, 17: 143-165. Hyland, K. (2005), Metadiscourse: Exploring Interaction in Writing. London & New York: Continuum. Hymes, D. (1992), ‘The concept of communicative competence revisited’, in: M. Pütz (ed.), Thirty Years of Linguistic Evolution. Amsterdam: John Benjamins Publishing Company. 31–57. Johansson, S. (1998), ‘On the role of corpora in cross-linguistic research’, in: S. Johansson and S. Oksefjell (eds) Corpora and Cross-Linguistic Research: Theory, Method, and Case Studies. Amsterdam: Rodopi. 3-25. Kroll, B. (1990), ‘What does time buy? ESL student performance on home versus class compositions’, in: B. Kroll (ed.), Second Language Writing: Research Insights for the Classroom. Cambridge: Cambridge University Press. 140-154. Lagerholm, P. (1999), Talspråk i skrift. Om muntlighetens utveckling i svensk sakprosa 1800–1997 [Speech in writing: On the development of orality in Swedish prose from 1800-1997]. Lund, Sweden: Studentlitteratur.

Involvement features in writing

53

Mårtensson, E. (1988), ‘Den familjära myndigheten. Intimiseringen av det offentliga språket’ [Familiar authorities: The intimization of public discourse], in: O. Löfgren (ed.) Hej, det är från försäkringskassan! Informaliseringen av Sverige. Stockholm, Sweden: Natur och Kultur. 105127. Petch-Tyson, S. (1998), ‘Writer/reader visibility in EFL written discourse’, in: S. Granger (ed.) Learner English on Computer. London & New York: Longman. 107-118. Roberts, C. and B. Street (1997), ‘Spoken and written language’, in: F. Coulmas (ed.) The Handbook of Sociolinguistics. Blackwell: London. 168-186. Svensson, J. (1993), Språk och offentlighet: Om språkbruksförändringar i den politiska offentligheten [Language in the public domain: On changes in language use in public politics]. Lundastudier i nordisk språkvetenskap A:47. Lund, Sweden: Studentlitteratur. Tannen, D. (ed.) (1982), Spoken and Written Language: Exploring Orality and Literacy. Norwood, NJ: Ablex.

Creativity in the use of verb + noun combinations by Chinese learners of English John Cross and Szilvia Papp University of Portsmouth Abstract In this paper we examine verb + noun combinations as used by Chinese learners of English and compare them with Greek and German learners as investigated by Giovi (2006) and Nesselhauf (2003), respectively. We show that the building blocks of learner language in Chinese interlanguage are memorised chunks that learners use to represent their ideas. This learner strategy seems to be more prominent in Chinese-English interlanguage than in Greek-English or German-English interlanguage, which is manifest in the Chinese learners’ more frequent use of semi-fixed expressions (i.e. verb + noun combinations), the higher error rate with which they use them, and the difference between the error types they produce and those of Greek and German learners of English. The factors that contribute to the ease or difficulty of learning and using verb + noun combinations by Chinese learners will be considered. These factors include languagerelated and environment-related factors, such as a) the type of input Chinese learners receive, mainly the textbooks they use, and the amount and nature of target English they are exposed to; b) features specific to the mother tongue, Chinese; and c) the influence of education in the Peoples’ Republic of China (PRC). We argue that all these factors will have an impact on the amount and type of creativity that Chinese learners of English exhibit in their language production.

1.

Introduction

In this paper we would like to contribute to the growing literature that draws on Contrastive Analysis (CA) and Contrastive Interlanguage Analysis (CIA) (Granger 1996, Gilquin 2000/2001) by comparing the use of verb + noun combinations in three interlanguage systems (Chinese-English, Greek-English and German-English). More particularly, we investigate the use of verb + noun combinations in two new comparable learner corpora: a Chinese learner corpus (Papp forthcoming) and a Greek learner corpus collected and analysed by Giovi (2006), and compare the results with a previous study on verb + noun combinations by German learners of English carried out by Nesselhauf (2003). We offer an account of the similarities and differences in the use of verb + noun combinations by these three groups of students. We propose several explanations, some already presented in the literature (such as the role of mother tongue [L1], teaching-induced factors and the input learners are exposed to), and add a new one, that of the variable role of creativity in language use among the different groups of learners.

58

John Cross and Szilvia Papp

It is usually accepted that knowledge of vocabulary, and more particularly, knowledge of the lexico-grammatical features of a second language (L2), is an important component of general language proficiency. However, language is not a random collection of words; there are systematic tendencies for words to cooccur (Sinclair 1991). Formulaic and phraseological knowledge is considered a prerequisite for native-likeness and fluency (Biskup 1992, Cowie 1992, Howarth 1998a, b). In fact, it has been estimated that formulaic language makes up 70% of native language production (Wray & Perkins 2000). Cowie (1998: 1) pointed out that “native-like proficiency in a language depends crucially on a stock of prefabricated units – or ‘prefabs’ – varying in complexity and internal stability”. Collocations, i.e. typical word combinations, form a part of prefabricated language that native speakers use, and their acquisition by L2 learners has been investigated by a number of studies which have used learner data among other types of data (e.g. Granger 1998, Howarth 1998a, b, Lewis 2000, Nesselhauf 2003, 2005). Since collocations are word combinations which involve arbitrary restrictions, their learning presents a problem (Cowie 1994, 1998, Weinert 1995, Howarth 1998a, b, Wray 1999). When someone “knows” a language, they have a knowledge of what is formally possible (usually referred to as grammatical competence), what is psycho-linguistically feasible (as language processing is constrained by memory capacities), what is socio-linguistically appropriate (context-related constraints governed by situation, participants, etc), but they also have a knowledge of what is actually performed with high probability/frequency (knowledge of typicality) (Stubbs 2001). All these aspects of language knowledge are necessary but not sufficient conditions for communicative competence. It is not enough to possess all these aspects of knowledge, learners need to be able to put them to use. Vocabulary teaching tends to take a key position in L2 pedagogy, as both teachers and students usually believe that L2 vocabulary needs to be taught and learnt explicitly. The learning of L2 collocations is an interesting area of research, not only from a pedagogical, but also a theoretical point of view, since collocations are arbitrarily restricted typical word combinations, and therefore both explicit and implicit types of learning (see DeKeyser 2003) are required for their mastery. Moreover, since collocations are word combinations that occur frequently together in the target language, but at the same time are characterised by some limitations on their combinatorial possibilities, language users have to work within the boundaries of possible variation, enabling creativity in language use. Some linguists believe that all language learning, including syntactic and lexical aspects, is driven by input alone (e.g. Hoey 2004, 2005). In this conceptualisation of language learning, learners are thought to use only general cognitive processes, such as analogy and generalisation, based on the language exposure they receive. This is what Hoey (2004, 2005) has called “lexical priming”. In his theory, every time a lexical item is encountered by the learner, it is primed by the lexical items surrounding it as well as the syntactic and textual patterns it appears in. It is clear that both explicit and implicit learning processes

Creativity by Chinese learners of English

59

need to be employed by L2 learners in the learning of word combinations, since they can be placed on a continuum according to several criteria: their frequency of co-occurrence, the strength of the relationship between their elements, and their internal variability. In other words, learners need to find out how much these word combinations are used, how much they like the company of each other, and what other partners either of them can have. In Cowie’s words (1998: 15), “it is typically not a matter of learning fixed units but of knowing when, and how far, the elements of a collocation are able to recombine with other items”. Therefore, the role of the input and memorisation is just as paramount as other general learning processes such as analogy and generalisation in their mastery. Research has shown that restricted collocations constitute the bulk of word combinations in various native corpora. Howarth (1996: 120) has found, for instance, that one-third of verb + noun combinations can be classified as restricted collocations.1 They have been found to present “the most formidable learning difficulty” (Cowie 1992: 3-4, see also Cowie 1998: 15 and Howarth 1998a: 31). This is because knowing how words are used in connection with other words requires sufficient exposure to them in context. This fact points to the paramount importance of the appropriate amount and nature of input in L2 learning. As a result of the qualitatively and quantitatively different input that L2 learners are usually exposed to, learners are left to create what Hoey (2005: 160) has termed “incomplete, inconsistent and leaky, but nevertheless workable semantic and grammatical systems” which are even more “incomplete”, “inconsistent” and “leaky” than those of native speakers. Learners not only need to find out what combinatorial possibilities exist, but also what restrictions need to be placed on them. These incomplete, inconsistent and leaky systems, including non-native-like word combinations, as pointed out by Nesselhauf (2005: 2), can distract the recipient and draw the attention away from the message. In the next sections we are going to outline types of word combinations (Section 2) and the way they have been analysed by other researchers, before we list the criteria for the analysis of verb + noun combinations in our corpus of Chinese learner English (Section 3). In Sections 4 and 5 we report our study and compare our results with the other two learner corpora, i.e. Greek and German learners of English. In Section 5 we list the sources of non-nativeness in the Chinese learner corpus and give some representative examples for each. In Section 6 we turn our attention to the role of creativity in the production of Chinese learners, and finally in Section 7 we outline some theoretical issues for a possible theory of creativity in language use by native and non-native speakers. 2.

Types of word combinations

Word combinations can be placed on continua or clines according to several criteria. Corpus linguists such as Sinclair (1991) and Stubbs (1995) usually define word combinations (and among them collocations) based on frequency and cooccurrence relations. Others, mainly within the phraseological tradition, such as

60

John Cross and Szilvia Papp

Cowie (1994), define them according to semantic and syntactic criteria. It is usually agreed that lexical and grammatical word combinations can have graded idiomaticity, i.e. semantic transparency and restricted substitutability, contributing to the notion of continua. The continua are defined by criteria such as: a) frequency (from rare to frequent), b) strength of co-occurrence of elements (statistically weak to strong relationship), c) restrictedness/internal stability (from variable to fixed), d) complexity (from simple to complex), as well as e) utility/intelligibility in L2 communication (from highly transparent, and therefore useful, to semantically opaque and thus, arguably, unintelligible to L2 learners).2 At one end of the continua of idiomaticity are free combinations and at the other end are fixed idioms making up a scale of idiomaticity. In the middle of the continua we find the category of collocations. Collocations are usually fairly frequent, lexico-grammatically complex phrases, whose elements have a strong relationship. In addition, collocations are to some extent variable, but the variation is restricted in an arbitrary way. They are semi-idiomatic, in that one element may have a non-literal meaning, but they are at the same time semantically transparent due to the compositionality of meaning of their elements. Finally, they usually have high utility in L2 communication, unlike idioms which, according to Seidlhofer (2004), do not. 3.

Analysis of word combinations

In this section we review some of the criteria that we have used in the analysis of learner data in this paper: 1) frequency, 2) transparency, 3) commutability, and we shall add our own criterion, 4) creativity. The first criterion that we have used is frequency. It is not difficult to see how purely frequency-based analyses of word combinations could be criticised, since they largely depend on the size of the corpus being analysed (Howarth 1998a: 26), raising questions, as a result, about the reliability of frequency findings. Secondly, frequency-based analyses may sometimes lead to invalid conclusions as far as competence is concerned as they put unnecessary emphasis on frequently attested collocations with strong collocational strength, such as have children, which may occur frequently as a result of the subject matter in a specific corpus (Howarth 1998a: 26-27). It is clear that frequency of cooccurrence cannot be the only criterion that is used in the definition of types of word combinations. Linguists within the phraseological tradition have used two other basic criteria for the delimitation of word combinations. A note of caution is due here since each criterion has been labelled differently by different scholars. For

Creativity by Chinese learners of English

61

instance, the second criterion that we have used, that is, transparency, has been variably referred to as opacity, compositionality or literalness. The third criterion of commutability is also known as substitutability, restrictedness or collocability. It is important to define what we mean by each criterion here. Transparency refers to the compositionality of meaning of the elements making up a combination. Wray (1999: 215) has termed compositionality “the extent to which the structure or meaning of the whole is a direct reflection of the sum of the parts”. Commutability, on the other hand, refers to the degree of restriction in the substitution of the elements of the combination. As a result of the criteria of transparency and commutability, word combinations can: (1) (2)

(3) (4)

be freely combined and have a literal meaning (e.g. blow a trumpet); have one component which is used in a special, often figurative sense only in the context of a limited number of collocates (e.g. blow a fuse, which means “lose your temper”, but is sometimes used in its literal sense); have a metaphorical meaning, but can sometimes be literally interpreted (e.g. blow your own trumpet); be completely semantically and syntactically opaque (e.g. blow the gaff) (examples from Cowie 1994 and Howarth 1998a).

Some authors have combined all three criteria of frequency, transparency and commutability, and define a collocation as a restricted, transparent and frequent word combination (e.g. Nation 2001). Others, such as Nesselhauf (2003, 2005) and Poulsen (2005), have singled out the criterion of restrictedness: Restricted collocations like foot/pick up/settle the bill are seen as occupying an intermediate position on a continuum ranging from “idioms”, which are typically defined as noncompositional multiword units like fill/fit the bill, to free combinations like discuss/tear up the bill, which are seen as fully compositional. (Poulsen 2005: 18) The fourth criterion that we propose for the delineation of various types of word combinations is the potential for creativity and experimentation in their use. For the purpose of this paper, we shall define creativity as the deliberate use of non-standard, unconventional, atypical word combinations by users of a language, both native and non-native, which are still within the boundaries of permissible (syntactic and semantic) variation within a language. Thus, the application of criteria of transparency, commutability and creativity enables the following classification of word combinations in Table 1 (based on Cowie 1994, Howarth 1998a, Nesselhauf 2005 and Giovi 2006):

62

John Cross and Szilvia Papp

Table 1. Classification of word combinations (based on Cowie 1994, Howarth 1998a, Nesselhauf 2005, examples from Giovi 2006) Free combinations wear a coat, pay money, carry a gun

Restricted collocations wear a smile, pay attention/respect/ compliment/call/visit, carry a burden

Figurative idioms wear the trousers, pay lip service to / no heed, carry on

Transparency

All the elements are used in a literal sense.

Commutability

Substitution is based on semantic grounds. Creativity is allowed and is only restricted semantically.

One element is used in a non-literal sense, but there is an element with a literal meaning; the whole combination is transparent. Partial substitution is possible; arbitrary limitations exist.

The combination has a figurative meaning, but preserves a current literal interpretation. Substitution is almost impossible.

Creativity is restricted by arbitrary limitations, but violations are possible.

Creativity is severely restricted. While deliberate creativity is possible, it is highly marked.

Examples

Creativity

Pure idioms wear your heart on your sleeve, he who pays the piper calls the tune, carry the can The combination cannot be literally interpreted. Substitution is impossible.

No creativity is allowed; only highly deliberate breaking of rules for special stylistic effect.

We have defined our fourth criterion, creativity, as a notion closely connected to (in fact, the opposite of) the idea of normality or typicality. If we probe this criterion further, we can see that there are two dimensions of typicality or normality along which word combinations can vary: grammaticality and idiomaticity, as seen in Figure 1 below. + grammaticality 4

2

3

1

- idiomaticity

+ idiomaticity

- grammaticality Figure 1. Dimensions of typicality Thus, if a language user violates either of the dimensions of grammaticality or idiomaticity, they stretch the boundaries of normality or typicality. Let us now exemplify the different sectors of these two dimensions.

Creativity by Chinese learners of English

63

In category 4 (grammatical but unidiomatic phrases), we find unidiomatic, foreign-sounding combinations typically used by advanced language learners, such as: (5)

*raise the question about, *reach an aim, *pass one’s judgements, *make one’s homework, *give a solution to, *take one’s task (from Nesselhauf 2003); *conduct a task, *perform a survey (from Cowie 1986, cited in Cowie 1994); *perform a project, *pay effort, *reach findings (from Howarth 1998a)

Some of these can be creative according to our definition above, some are, however, the result of a lack of knowledge of the correct phrase. In category 3 (ungrammatical, unidiomatic language), we would find undecipherable language, such as: (6)

We exit at on the earth, the person is the son of the nature, but we can’t see only the person make the natural conqueror. (from Papp forthcoming)

In category 2, we would expect to find typical, normal word combinations produced by native and near-native speakers (grammatical, idiomatic phrases). However, in this category, if the grammaticality and idiomaticity are taken to the extreme, and as a result of repeated use, some phrases can become too familiar, banal or clichéd, such as the example: (7)

This is guaranteed to meet your special requirements.

Finally, in category 1 (idiomatic, but ungrammatical phrases), we find incidentally or deliberately creative, unique word combinations produced by native or non-native speakers, such as the lyric by Bob Dylan (8), or the deliberately creative alternative uses of idioms such as wear your heart on your sleeve, which can be found by a simple search on Google, e.g. (9). (8) (9)

Today’s the day I’m gonna grab my trombone and blow. wear your faith / patriotism / influence / pride / politics on your sleeve, wear your heart on your wrist / shirt / feet3

Thus, we could expect creative use of language in three of these categories, but not in category 2. 4.

Our study: methodology

The purpose of our study was to compare the use of verb + noun combinations by three learner groups to see what similarities and differences we find in their use. Our particular interest was to see which learner group produced more non-native word combinations, and to see whether there were some qualitative differences in

64

John Cross and Szilvia Papp

the non-native word combinations among the three groups. If we were to find quantitative and qualitative differences among the learner groups, we might attribute them to some factors, such as the influence of the L1, the effect of the type of input learners receive, or other environment-related factors. We will argue that all these factors may have an influence on the amount of creativity learners exhibit in their use of verb + noun combinations. Details of the three learner corpora can be seen in Table 2. We analysed 48 essays by Chinese learners of English from the Portsmouth Chinese-English learner corpus (Papp forthcoming), which is part of the International Corpus of Learner English (ICLE) project (Granger et al. forthcoming). The code of the Portsmouth Chinese learner corpus within the ICLE database is CNUK. The German corpus which is also part of the ICLE project (Granger et al. 2002) is described in Nesselhauf (2003). The Greek corpus is described in Giovi (2006) and was collected in a similar way to the ones represented in the ICLE project. The three corpora are comparable as the learners are similar in educational background and length of exposure to English. According to these external criteria, they can be categorised as advanced learners of English within the ICLE project. The task type the learners were asked to perform is also similar, that is, academic essays. Since the sizes of the learner corpora are not comparable (see Table 2), in the results section we report relative percentage data. Wherever possible, we ran statistical tests on the differences in proportions that we found. Table 2. Corpora used in the study L1

Type

No of essays

No of words

German Greek Chinese

NNS NNS NNS

32 28 48

16,000 8,500 25,884

No of verb + noun combinations 1,072 538 1,596

Percentage of verb + noun combinations 13.4% 12.6% 12.3%

From the last column in Table 2 we can see that all three groups have used similar amounts of verb + noun combinations. This can be shown through statistical tests: there is no significant difference between the total number of verb + noun combinations used by the three learner groups relative to the size of the corpora as displayed in Table 2 (Ȥ2 = 4.17, df = 2, p = 0.20). The Chinese corpus was corrected by two native speakers and fully errortagged with the ICLE error tagset.4 First, all verb + noun combinations were extracted, including passives and participial reduced relative clauses. Then, in order to identify non-native combinations, a native speaker of English made judgements as to their acceptability in context along a continuous Likert scale of 1-5 (where 1 = clearly acceptable, 2 = largely acceptable, 3 = unclear, 4 = largely unacceptable, and 5 = clearly unacceptable).5 As a third step, the non-native combinations (those judged as “largely unacceptable” or “clearly unacceptable” by the native speaker) were classified according to the type of mistake(s)

Creativity by Chinese learners of English

65

occurring in them. Here our objective was not to distinguish idioms from collocations and other free combinations, but to determine the extent of nonnative-likeness and the possible causes of errors in our learner data.6 5.

Results

Table 3 displays the overall results of the three separate data analyses including all types of verb + noun combinations (i.e. free word combinations, restricted collocations and idioms): Table 3. Overall degree of acceptability of verb + noun combinations in the three learner corpora (German data based on Nesselhauf 2003, Greek data based on Giovi 2006) Degree of acceptability

Clearly and largely acceptable

Unclear

Largely and clearly unacceptable

Total no of verb + noun combinations

German

787 (73.4%)

30 (2.8%)

255 (23.8%)

1,072

No of non-native (i.e. unacceptable) verb + noun combinations per 100 words 1.59

Greek

394 (73.2%)

6 (1.1%)

138 (25.6%)

538

1.62

Chinese

770 (48.2%)

10 (0.6%)

816 (51.1%)

1,596

3.15

We can see from these results, especially from the relative frequencies of nonnative verb + noun combinations per 100 words produced by learners, that German and Greek learners produce similar numbers of errors in verb + noun combinations (1.59 and 1.62 per 100 words, respectively), whereas Chinese learners produce almost double the number of non-native verb + noun combinations (3.15 per 100 words). There is a statistically significant difference if we compare the number of non-native verb + noun combinations among the three learner groups (Ȥ2 = 122.89, df = 2, p<0.001). In Table 4 we give a breakdown of the types of lexical mistakes in nonnative (i.e. unacceptable) verb + noun combinations in the three learner corpora.

66

John Cross and Szilvia Papp

Table 4. Types of lexical errors in verb + noun combinations in the three learner corpora (categories and examples based on Nesselhauf 2003)7

Verbrelated

Nounrelated

Category

Example

Occurrences in German corpus

Occurrences in Greek corpus

Occurrences in Chinese corpus

Wrong choice of verb (or nonexistent, missing or redundant verb)

carry out races $hold races$

24 (40.68%)

29 (51.79%)

147 (34.19%)

Verb preposition

fail in one’s exams $fail one’s exams$ close lacks $close gaps$

3 (5.08%)

1 (1.79%)

86 (20%)

14 (23.73%)

13 (23.21%)

80 (18.60%)

4 (6.78%)

3 (5.36%)

10 (2.33%)

2 (3.39%)

0 (0%)

5 (1.16%)

11 (18.64%)

7 (12.50%)

90 (20.93%)

1 (1.69%)

3 (5.36%)

12 (2.79%)

Wrong choice of noun (or non-existent, missing noun) Noun preposition

Usage

Combination exists but is not used correctly Combination does not exist and cannot be corrected by exchanging single elements

Structure

Syntactic structure wrong

raise the question about $raise the question of$ to take notice $to notice$ hold children within bounds $show children where the boundaries lie$ make sb. friends $make friends with sb.$

From Figure 2, we can see that verb-related errors are different in the Greek and Chinese learner groups when compared to the German group: approximately 54% of errors produced by Greek and Chinese learners are verbrelated errors as opposed to 46% in the German learner data. However, if we compare the results in Table 4 for the Greek and Chinese students, we see that the Chinese learners do not seem to experiment with the wrong choice of verb as

Creativity by Chinese learners of English

67

much as Greek and German learners do (only 34% of lexical errors are due to wrong choice of verb in Chinese learners, as opposed to 41% in German learners and 52% in Greek learners). The Chinese learners, however, do use the wrong verb preposition much more often than the other two groups (20% by Chinese learners as opposed to 5% and 2% by German and Greek learners, respectively). The proportion of noun-related errors is similar in the German and Greek corpora but different from the Chinese (31% of German errors and 29% of Greek errors as opposed to 21% of Chinese errors are noun-related). This is, again, because German and Greek learners seem to experiment with lexis by choosing the wrong noun more than Chinese learners do (24% German vs. 23% Greek vs. 19% Chinese).

60 50 Total verb-related errors

40

Total noun-related errors

30

Total usage + structure errors

20 10 0 Occurrences in German corpus

Occurrences in Greek corpus

Occurrences in Chinese corpus

Figure 2. Summary of verb-related, noun-related and usage + structure errors in the three learner corpora In the usage + structure category, German and Chinese learners produce similar numbers of combinations (24% and 25%), whereas usage + structure errors accounted for only 18% in Greek learners’ non-native combinations. This is due to the fact that German and Chinese learners produced a similar proportion of usage errors where the combination does not exist and cannot be corrected by exchanging single elements (Chinese 21%, German 19% vs. Greek only 12.5%). Statistical tests show that the only difference that is statistically significant among error types within non-native word combinations is in the number of wrong choice of verb among all three groups (Ȥ2 = 7.058, df = 2, p<0.05). The same is not true of wrong choice of nouns, where the difference is not statistically significant (Ȥ2 = 1.371, df = 2, p = 0.504).

68 6.

John Cross and Szilvia Papp Sources of non-nativeness in the Chinese corpus

We now wish to address the issue of why the written English of our Chinese learners shows more non-native verb + noun combinations than that of German and Greek learners, a finding reported in Table 3.8 We argue that although the influence of the L1, Chinese, is clear, as shown below, it does not entirely account for the quantity and different types of non-native features in the Chinese corpus, and that some of the most characteristic errors are the result of input in the teaching and learning process as well as notions regarding the production of English. Such input-related errors take several forms, although the most noticeable one may be the inappropriate or inaccurate use of memorised chunks of language. In what follows we categorise and provide examples of errors induced by language-related factors, and then suggest categories and provide examples of errors induced by other environment-related factors. In our qualitative discussion of different categories of errors, it should not be assumed that we are suggesting each error may be neatly assigned to one category or another. Apart from the fact that without access to the mind of the learner at the time of production, judgements about errors are speculative, it is also clear that word combination errors may show elements from different categories. 6.1

Language-related errors

Errors induced by language-related factors may be subdivided into three subcategories: those influenced by grammatical and structural characteristics of the L1 (Chinese), those influenced by direct translation of the L1, and those influenced by lack of exposure to the target language, English. Naturally, there is some overlap, and these categories represent tendencies rather than absolutes. 6.1.1 Errors influenced by the grammar and structure of L1 Non-native features influenced by the grammatical and structural features of Chinese verb + noun combinations include non-standard or inaccurate word forms (L1 reason: Chinese does not inflect either the verb or the noun for tense, plurality, etc), omission or inaccurate use of articles (L1 reason: Chinese does not have articles), omission or inaccurate use of prepositions (L1 reason: Chinese has fewer words functioning as prepositions and they are used less frequently than in English), omission of conjunctions (L1 reason: Chinese often requires no conjunction or uses a comma in places where a conjunction is required in English), and attempts to recreate typical Chinese sentence structure, in particular the use of topic-first sentences (L1 reason: topicalised sentences are common in Chinese). Examples of inaccurate verb forms include (10) and (11). More complex examples are not infrequent and include (12) and (13).

Creativity by Chinese learners of English (10) (11) (12) (13)

69

Military service, in some country of the word is force […] (CNUK1006) No one can resist the temptation of know the big new happen nowadays immediately. (CNUK1021) Even the common man might be possible to meet the unexpectable accident. (CNUK1006) We live in a peaceful world and we needn’t a huge army. (CNUK1022)

An example of both inaccurate use and omission of articles is (14). Use of the statistics is not always incorrect, but this sentence starts a paragraph and statistics are not previously referred to in the discourse. An example of the omission of prepositions is (15), whereas an example of inaccurate use of prepositions is (16). (14) (15) (16)

According to the statistics, the chance to have Ø boy or girl are half and half […] (CNUK1061) Secondly, people look Ø capital punishment […] (CNUK1078) People can look at about execution in their society […] (CNUK1090)

An example of the omission of a conjunction is (17). In this case English requires and, whereas in Chinese a comma would indicate the link between the two phrases. (17)

Spending much time on watching TV not only do harm to your health, waste your time, but also […] (CNUK1025) Sentences that seem to use the topic-comment pattern include:

(18)

(19)

About capital punishment people usually discuss it and some people agree should be carried out in all cases for murder, some people disagree the opinion. (CNUK1078) Talking about ‘East Enders’, which just like cartoons of Disney, traditional dance of Japan and ‘hip top’ in Africa, nobody doesn’t know in Britain. (CNUK1093)

The sentence structure “topic + is + adjectival phrase” is common in Chinese and seems to be replicated in example (20): (20)

I don’t think the capital punishment is China has necessary to be abolish”. (CNUK1205)

The equivalent of this sentence in Chinese would be: shì zhǀngguó yƯnggƗi fèichú de. The sentence in (21) seems to use the structure ràng rén nán + verb, literally ‘make people difficult’, meaning ‘make it difficult for people (to do something)’, with the additional feature that nán is mistranslated as hard rather than difficult.

70 (21)

John Cross and Szilvia Papp It’s something like mental opium that made people hard to stop”. (CNUK1025)

6.1.2 Errors influenced by direct translation from L1 Non-native features influenced by direct translation may include individual words, whole phrases, including fixed sayings, and even whole structures, including topic-first sentences, which all potentially bear on the use of verb + noun combinations. An example of a direct translation of a Chinese collocation is: (22)

When the people after their work, they will choose to open the television”. (CNUK1076)

The Chinese for ‘switch on a television’ is kƗi diànshì, literally ‘open television’. Another example is: (23)

TV can make people’s life more colourful […] (CNUK1044)

This is very likely to be a translation of fƝngfùduǀc΁i, literally ‘abundant many colours’, but meaning ‘rich and varied’. The Chinese word wán is usually translated as play in English, and there are contexts where this would be acceptable (such as the children play outside), but more frequently a different word is required in English as in example (24): (24)

[…] instead of working or playing with their families in the limited time. (CNUK1036)

The Chinese word xìnggé can be translated as temperament, but the word character would be a better choice in example (25): (25)

[...] deciding children’s temperament may result in personality singleness among people. (CNUK1113)

Example (25), unlike open the television exemplified in (22), suggests misuse of a dictionary, as does (26), where culprit has been chosen instead of criminal. The choice of adjudge is more interesting in that verbs with similar meanings, such as sentence, cannot fit into the existing structure (a nation sentences a criminal to capital punishment), and verbs that fit, such as mete out, have a different meaning, as well as being infrequent. The misuse of an electronic dictionary is strongly suggested in example (27). Dacoit is an Indian English word meaning ‘bandit’, ‘brigand’, from which dacoity meaning ‘banditry’ is derived.

Creativity by Chinese learners of English (26) (27)

71

When a nation adjudges capital punishment to serious culprit […] (CNUK1078) […] commit a crime of young people, like, dacoity, murder and so on. (CNUK1097)

6.1.3 Errors influenced by lack of exposure to target language Limited exposure to authentic language norms and lack of feedback are likely to be a factor in preventing learners from making accurate judgements about the combinatorial possibilities of collocations. Such inaccuracies may include words or phrases imperfectly acquired, and, for example, confused with words with similar sounds or similar meanings. For example, the sentence in (28) suggests the similar-sounding respect was intended. Another example in (29) seems to link insist or possibly consider with consist. The sentence in (30) seems to use full fill for the similar-sounding but semantically different fill, as well as irate for the phonetically and semantically similar verb irritate. Example (31), where deny seems intended, indicates a transfer into English of Chinese sounds by a learner whose first language is Cantonese, in which l and n are allophones. The sentence in (32) implies that the writer has been exposed to (or looked up) the archaic word bethink but has not fully acquired its correct use, and, possibly, uses it to add a more formal quality to the writing. Example (25), repeated for reference in (33), where the context strongly suggests uniformity is intended instead of singleness, may be a case where the writer has imperfectly understood the meaning of input, linking singleness to ‘all being the same one’. (28) (29) (30) (31) (32) (33) 6.2

[…] everyone should expect to each other […] (CNUK1205) […] most people consist on that university degrees are useful in reality. (CNUK1072) […] advertisement full fill the margin time to irate the audience […] (CNUK1093) I can’t dely that I make a certain progress […] (CNUK1001) Thus, some people bethink of corruption, bribery and so on. (CNUK1078) […] deciding children’s temperament may result in personality singleness among people. (CNUK1113) Environment-related errors

Errors induced by environment-related factors include both those related to specific language teaching and learning as well as those related to general education and, possibly, to concepts widely held in Chinese society. We examine these in turn and give indicative examples of non-native-like verb + noun combinations for each. Until very recently most learners of English in secondary and tertiary institutions in the Peoples’ Republic of China (PRC) were given relatively limited exposure to authentic native language, and textbooks tended to use English from a very wide range of genres, registers and historical periods without explanation

72

John Cross and Szilvia Papp

or contextualisation, as well as many example sentences created by Chinese users of English in the PRC. Overt language teaching and learning in the PRC still lays relatively heavy emphasis on the memorisation of language (Watkins & Biggs 1996, 2001), including chunks. This observation is based on one of the authors’ (Cross) experience of learning and teaching in the PRC, and supported by research in the educational field. Chunks may be misused either by being employed in an inappropriate context, or by being inaccurately recalled. In terms of collocations, the more restricted the type, the more damaging to comprehension its inappropriate or inaccurate use will be (cf. Nesselhauf 2003, 2005). Limited exposure to authentic language norms is likely to be a factor in preventing learners making accurate judgements about the combinatorial possibilities of collocations. Language teaching and learning in the PRC not only influences the input that is made available to learners and the way learners learn and acquire, but also influences attitudes about the nature of language learning and about the nature of non-Chinese languages, English in particular. These attitudes seem to rate translation into Chinese as the key language skill and consider the purpose of the production of English, especially written, to be the demonstration of (memorised) knowledge of structures and vocabulary chunks. Communication and comprehensibility seem to be relatively less important. Furthermore, it may be that the L2, English, is not considered as a code which represents reality, but rather as a code which re-represents the reality of the L1, Chinese (cf. Daniels 1995). This view will almost certainly have an influence on comprehensibility. Of course, most Chinese learners of English in the PRC, certainly the majority in our corpus, are bi- or multilingual in Chinese languages, and English might be more accurately called their L3 or L4. However, there seems to be a gulf between the view of various Chinese languages and that of foreign languages, of which English is the most prominent; a view which is likely to influence the learning and production of English, and, in particular, lead to increased tolerance of ambiguity. That is to say, learners either do not notice deviation from native norms because of limited authentic input or, because of prevailing attitudes towards the production of English, they are relatively unconcerned about conforming to previously encountered norms. Examples of memorised chunks used inappropriately include: (34)

How time flies. Some to think of it […] (CNUK1001)

The first phrase in (34) is grammatically and semantically correct, although the register may be inappropriate. The second phrase is inaccurately reproduced, some being used for the phonetically similar come, which in any case would not make sense in this context. It seems to be an imperfectly recalled memorised chunk employed at the expense of meaning. Numerous examples in the corpus seem to show an attempt to start an essay or paragraph with a memorised set phrase, all somewhat inaccurately. The following are some representative examples:

Creativity by Chinese learners of English

(35) (36) (37) (38) (39)

73

As we can see, following economy increase rapidly […] (CNUK1011) As the rapid development of science and technology […] (CNUK1005) As the development of science and technology […] (CNUK1018) With the development of the society, […] (CNUK3062) We can with the development of our modern society […] (CNUK1076)

In addition, it may be that the English production of learners in the PRC is influenced by general concepts widely held in that society. Certain ideas that are taken as undisputed but which are not shared elsewhere, when combined with language errors, are likely to create confusion in the mind of the reader. When faced with serious comprehension difficulties, the non-PRC reader may not be able to guess what the writer’s intention was based on shared fundamental ideas – what might be called ideology. These concepts cannot be vaguely called “Chinese culture” in that Chinese learners of English outside the PRC may not share them. There may be some link here with research by Li, Chen & Yu (2006) showing that, because of differences in education, Chinese students were overconfident in opinions compared with ethnic Chinese students from Singapore. Examples of belief in fixed ideas include (40) to (42). (40) (41) (42)

Known by every Chinese citizen that […] (CNUK1004) As the sentence we all know ‘we’re not too old to learn’ […] (CNUK1004)9 In Japanese, most of woman, who hold bachelor or above degree, will not go out to work after having a baby; they draw all attention on the family and children. (CNUK1042)

The example in (42) refers to a commonly-held view treated as fact by many university-educated people in the PRC. Other examples, such as the one in (43), are difficult to interpret because it is not clear whether they are language or factual errors: (43)

Nowadays, not every country has death penalty, such as America, Canada. (CNUK1037)

It is difficult to know whether such as refers to examples of “country has death penalty” or to examples of “not ... has death penalty”. In any case, America and Canada belong to different categories. Another example in (44) refers, presumably, to socialist views of rational “scientific” living, work and thought: (44)

Using your time scientially. Don’t waster your limited time in that kind of nonsense. (CNUK1044)

Again it is difficult to know for certain and comprehension is seriously impeded. The difficulty for the reader is the combination of beliefs not shared by those

74

John Cross and Szilvia Papp

outside the PRC with language errors related to an overall strategy in which communication does not seem to be the main purpose. In the next section we argue that these language- and environment-related factors bear on the increased use of memorised chunks and therefore a limited use of creativity in Chinese interlanguage. 6.3

The effect of creativity in learner language

In this section, we would like to discuss the issue of creativity in interlanguage in relation to our Chinese learner data. Nesselhauf (2005: 5), in her survey of the literature on L2 collocation use, recounts the difference found by Biskup (1992) between Polish and German learners of English as follows: Biskup (1992) describes a study in which 34 Polish and 28 German advanced learners were asked to translate 23 collocations into English. Clear differences between the two groups emerged. The Polish students produced more collocations than the German students, but they also much more frequently gave no answer at all. The German students more frequently tried to paraphrase the intended meaning without using a collocation but made more mistakes. According to Biskup, this emphasis on accuracy on the part of the Polish students and creative strategies on the part of the German students can probably be put down to the different emphases in foreign language teaching in the two countries. She also observes that the L1 influence on non-native forms is greater with the Polish than with the German students, and that different types of transfer are preferred by the two groups. Nesselhauf (2003: 233) herself attributes German learners’ attested tendency to use the wrong verb or noun in restricted collocations to their excessive creativity motivated by teaching practices which encourage originality and expressiveness. She writes: Combinations of the type RC1 [collocations involving a lot of restriction, where the verb can, in the given sense, only combine with a limited number of nouns] are more often acquired and produced as wholes, whereas combinations of the type RC2 [collocations involving a little restriction in which the verb cannot, in the given sense, be used with every noun that would be syntactically and semantically possible] are more creatively – and sometimes too creatively – combined by learners. These are very interesting observations about the creative use of English by German learners. The possibility arises that we are dealing with two types of learner groups in general: some learners who rely more on memorisation and recall (Chinese) and others who take liberties and play around with the L2 more

Creativity by Chinese learners of English

75

freely, creatively and innovatively (German and Greek). This would seem to be supported by the numerous claims in the educational literature as to the tendency of Chinese learners to use repetitive learning and memorisation to ensure accurate recall (e.g. Biggs 1996, Marton et al. 1996, for a summary see Papp 2007). In addition, perhaps as a result of the typological differences between English and Chinese and the ensuing consequences for their learnability, as well as perceived purposes of English use, learners from Mainland China tend to resort to random guessing as a compensatory strategy more than other L2 learners of English (cf. Oxford 1990), rather than creatively and experimentally trying to convey their intended meaning. 7.

Towards a possible theory of creativity in language use

Even if Chinese learners are indeed shown to be less creative than, for instance, German or Greek learners in their use of the L2, the theoretical question remains: where do we draw the line between learner error and legitimate creative use of a language? Collocations provide an excellent testing ground for this, as they occupy a range on a continuum of restrictedness of combinability. In cases where the restriction is not so strict and native speaker judgements are not so categorical and determinate, learner creativity is given free leeway. As a first step in theory making, the distinction between creative language and error can be usefully defined by the distinction between convergent and divergent on the one hand and merely incomplete representations on the other (cf. Sorace 1993). Convergent representations are native-like uses of word combinations, which are readily attested by corpora of native speakers, while divergent usages are those idiosyncratic word combinations which are substantially different from native usage, as evidenced by non-occurrence in native production (see Kenny 2001). In addition to this, divergent knowledge representations are categorical, that is, if asked why they used a certain word combination, creative learners’/language users’ judgements would be determinate and their choice deliberate and intentional. To put it differently, creative learners’/language users’ knowledge state is confirmed, not critical, that is, their choices are characterised by stability and integrity within their language system (cf. Klein 1986). Their language use is firmly based on knowledge of the combinatorial possibilities and the legitimate violations of the restrictions placed on word combinations. The determination of legitimate violations is contingent upon the preservation of some of the original formal and semantic restrictions. Linguistic creativity in this sense can theoretically result in language change if other users of the language adopt the new word combination. Incomplete knowledge of word combinations, on the other hand, results in word combinations that are the product of indeterminate judgements on the part of the user. It is important to note that some non-native speakers can come up with idiosyncratic word combinations within the limits of the grammatical/ subcategorisational and semantic/selectional restrictions of the target language

76

John Cross and Szilvia Papp

even if their proficiency is relatively low. This could be the case, for instance, with the example of reach an aim from (5) above. Conversely, even native speakers can violate restrictions to the extent that their production would seem non-native-like. Our theoretical question about the dividing line between learner error and deliberate creative use of language needs an empirical answer. We need to be able to measure creativity and define its limits. As long as new innovative collocations are structurally, semantically and pragmatically motivated and justified, that is, they are entrenched in old conventional patterns and are recognised as creative uses of the language that were produced for some special purpose (for instance, for purposes of informality, irony or humour, see e.g. Carter 2004, Carter & McCarthy 2004), they can be accepted as legitimate combinations. However, when there is no such feeling that the non-conventional usage was based on conventional patterns and produced for a particular purpose, it will always be conceived of as a learner error and will serve as a marker of non-nativeness and taken as evidence of lack of control that the learner has over the L2. Our argument in this paper can be summarized as follows. We have been interested in three types of behaviour. First, Dylan Thomas, in the phrases cited in note 3, is making a deliberate choice, is sure of the rules, and is motivated by a desire to communicate in a new way. Second, some German users of English, such as in some of the phrases cited by Nesselhauf (2003) (see (5) e.g. raise the question about, reach an aim), are making deliberate choices, are sure of grammar rules, but not sure what is usual usage, and motivated by a desire to communicate their meaning. Third, some Chinese users of English are (in example sentence (28), everyone should expect to each other) not making a deliberate choice, not entirely sure of the rules, and motivated by a desire to demonstrate language knowledge. To recap: some German learners use what they know to communicate but are “divergent” when they do so, sometimes in a deliberately creative way, whereas some Chinese learners try to reproduce memorised chunks but do so inaccurately, demonstrating “incomplete” knowledge of the L2. 8.

Conclusion

In this study we set out to compare the use of verb + noun combinations by Chinese learners of English with those of Greek and German learners at university level. Our aim was to account for the quantitatively and qualitatively different verb + noun combinations Chinese learners of English produce. We found that while Greek and German learners seem to use English in an experimental and innovative way, Chinese learners tend to not avail themselves of the creative opportunities that the use of verb + noun combinations offers. That is, instead of trying to exploit the combinatorial possibilities of the language, they tend to use memorised chunks which are sometimes inaccurately recalled and inappropriately used. We accounted for this tendency by considering some language and environment-related factors, including the effect of the L1, but

Creativity by Chinese learners of English

77

more importantly the influence of particular language teaching and learning notions and practices shared by some Chinese teachers and learners. According to extensive research in L2 acquisition, native-standard L2 proficiency is most probably an unattainable aim. Teaching therefore should aim at a functional and effective L2 proficiency to approximate native-likeness and improve fluency in learners. We pointed out that the learning of verb + noun combinations requires a large amount of exposure, and their combinatorial possibilities and restrictions need to be learnt explicitly. On the other hand, idioms are fixed and thus need to be memorised as invariable chunks. They can present a difficulty in L2 learning, as they may lead to problems in intelligibility when unfamiliar to learners (Seidlhofer 2004, Jenkins 2006: 47). Their teaching should probably be restricted to those which are necessary for the learners’ needs. Since creativity as a learner strategy is only applicable in the case of collocations and free word combinations, not idioms, learners should be provided with sufficient input containing word combinations that they need for their own purposes of language use, and the possibilities of and restrictions on their combinability should be pointed out to them. Moreover, students should be encouraged in the innovative, experimental use of their L2 as a potentially effective communicative strategy. A final suggestion relating to research methodology in this area is in order. Learner production data should be complemented by experimental and intuitional data to capture aspects of competence as well as performance. Spontaneous production should be investigated along with cloze tests, translation tasks and acceptability judgements to validate any results on learners’ use of word combinations. Also, if one were to carry out a full CA/CIA study, one would need to compare the NNS results with NS data from comparable corpora of novice and professional/expert writing. Acknowledgements The authors would like to thank Gaëtanelle Gilquin, María Belén Díez-Bedmar, Carol O’Sullivan, Mario Saraceni, John Wrigglesworth, Sue Wright and Fiona Barker for their helpful comments on an earlier draft of this paper. Notes 1

The proportion of restricted collocations and idioms, which has been viewed as an index of collocational “density” and degree of stylistic conventionality, has been shown to vary according to register: collocational density in social science texts is 38%, against 40-50% in journalistic writing (Howarth 1998a: 35-36).

2

For an excellent review of definitions of categories along these continua, see Nesselhauf (2005).

78

John Cross and Szilvia Papp

3

Or the innovative phrases all the sun long, all the moon long and a grief ago by the poet Dylan Thomas, even though these are not verb-noun combinations.

4

Please note that throughout this paper the terms “errors” and “non-native features” are used interchangeably.

5

In Nesselhauf’s (2003) study the notation on the 5-point scale was slightly different. However the categories were the same both in Giovi’s (2006) and our study.

6

The Greek corpus was annotated only for the purposes of the study of verb + noun combinations (see Giovi 2006). Once the verb + noun combinations were identified, their occurrence and combinability were checked in three on-line dictionaries, in line with Nesselhauf’s (2003, 2005) methodology. The dictionaries used were the Oxford Advanced Learner’s Dictionary (OALD 2000), the Collins COBUILD English Dictionary (CCED 1995) and the Longman Dictionary of Contemporary English (LDOCE 1995). The objective of these analyses was to set apart free combinations and restricted collocations from idioms, in order to see in which category the learners’ L1 played the principal role.

7

Although not all the categories in Table 4 are lexical, we have kept the ones relating to usage and structure for ease of comparison of the three learner datasets.

8

For a review of non-native features in Chinese-English interlanguage, see for instance Papp (2007).

9

Both (40) and (41) come from the same essay, but as many as 14 essays contain the phrase we all know.

References Bahns, J. (1993), ‘Lexical collocations: A contrastive view’, ELT Journal, 47: 5663. Biggs, J. (1996), ‘Western misconceptions of the Confucian-heritage learning culture’, in: D. Watkins and J. Biggs (eds) The Chinese Learner: Cultural, Psychological, and Contextual Influences. Hong Kong: University of Hong Kong. 45-67. Biskup, D. (1992), ‘L1 influence on learners’ renderings of English collocations: a Polish/German empirical study’, in: P.J.L. Arnaud and H. Béjoint (eds) Vocabulary and Applied Linguistics. Basingstoke: Macmillan. 85-93. Carter, R. (2004), Language and Creativity: The Art of Common Talk. London: Routledge. Carter, R. and M. McCarthy (2004), ‘Talking, creating: interactional language, creativity, and context’, Applied Linguistics, 25(1): 62-88.

Creativity by Chinese learners of English

79

Cowie, A.P. (1992), ‘Multiword lexical units and communicative language teaching’, in: P.J.L. Arnaud and H. Béjoint (eds) Vocabulary and Applied Linguistics. Basingstoke: Macmillan. 1-12. Cowie, A.P. (1994), ‘Phraseology’, in: R.E. Asher (ed.) The Encyclopedia of Language and Linguistics. Oxford: Pergamon. 3168-3171. Cowie, A.P. (1998), ‘Introduction’, in: A.P. Cowie (ed.) Phraseology: Theory, Analysis, and Applications. Oxford: Clarendon. 1-20. Dagneaux, E., S. Denness, S. Granger and F. Meunier (1996), Error Tagging Manual Version 1.1. Centre for English Corpus Linguistics. Louvain-laNeuve: Université catholique de Louvain. Daniels, H. (1995), ‘Psycholinguistic, psycho-affective and procedural factors in the acquisition of authentic L2 phonology’, Speak Out!, 15: 80-85. DeKeyser, R. (2003), ‘Implicit and explicit learning’, in: C.J. Doughty and M.H. Long (eds) The Handbook of Second Language Acquisition. Oxford: Blackwell. 313-348. Gilquin, G. (2000/2001), ‘The Integrated Contrastive Model. Spicing up your data’, Languages in Contrast, 3(1): 95-123. Giovi, C. (2006), The use of verb-noun collocation by Greek advanced learners of English using corpora analysis and some implications for teaching. Unpublished MA dissertation, University of Portsmouth. Granger, S. (1996), ‘From CA to CIA and back: an integrated approach to computerized bilingual and learner corpora’, in: K. Aijmer, B. Altenberg and M. Johansson (eds) Languages in Contrast. Papers from a Symposium on Text-based Cross-linguistic Studies. Lund 4-5 March 1994. Lund: Lund University Press. 37-51. Granger, S. (1998), ‘Prefabricated patterns in advanced EFL writing: collocations and formulae’, in: A.P. Cowie (ed.) Phraseology: Theory, Analysis, and Applications. Oxford: Clarendon. 145-160. Granger, S., E. Dagneaux and F. Meunier (eds) (2002), The International Corpus of Learner English. Handbook and CD-ROM. Louvain-la-Neuve: Presses Universitaires de Louvain. Granger S., E. Dagneaux, F. Meunier and M. Paquot (eds) (forthcoming), The International Corpus of Learner English. Handbook and CD-ROM. Version 2. Louvain-la-Neuve: Presses Universitaires de Louvain. Hoey, M. (2004), ‘The textual priming of lexis’, in: G. Aston, S. Bernardini and D. Stewart (eds) Corpora and Language Learners. Amsterdam: John Benjamins Publishing Company. 21-41. Hoey, M. (2005), Lexical Priming: A New Theory of Words and Language. London: Routledge. Howarth, P. (1996), Phraseology in English Academic Writing: Some Implications for Language Learning and Dictionary Making. Tübingen: Max Niemeyer. Howarth, P. (1998a), ‘Phraseology and second language proficiency’, Applied Linguistics, 19(1): 24-44.

80

John Cross and Szilvia Papp

Howarth, P. (1998b), ‘The phraseology of learners’ academic writing’, in: A.P. Cowie (ed.) Phraseology: Theory, Analysis, and Applications. Oxford: Clarendon. 161-186. Jenkins, J. (2006), ‘Current perspectives on teaching world Englishes and English as a lingua franca’, TESOL Quarterly, 40(1): 157-181. Kenny, D. (2001), Lexis and Creativity in Translation: A Corpus-based Study. Manchester: St. Jerome Publishing. Klein, W. (1986), Second Language Acquisition. Cambridge: Cambridge University Press. Lewis, M. (2000), Teaching Collocation: Further Development in Lexical Approach. London: Language Teaching Publications. Li, S., W.-W. Chen and Y. Yu (2006), ‘The reason for Asian overconfidence’, The Journal of Psychology, 140(6): 615-618. Marton, F., G. Dall’Alba and T.L. Kun (1996), ‘Memorizing and understanding: The keys to the paradox?’, in: D. Watkins and J. Biggs (eds) The Chinese Learner: Cultural, Psychological, and Contextual Influences. Hong Kong: University of Hong Kong. 69-83. Nation, I.S.P. (2001), Learning Vocabulary in Another Language. Cambridge: Cambridge University Press. Nesselhauf, N. (2003), ‘The use of collocations by advanced learners of English and some implications for teaching’, Applied Linguistics, 24(2): 223-242. Nesselhauf, N. (2005), Collocations in a Learner Corpus. Philadelphia, PA: John Benjamins Publishing Company. Oxford, R.L. (1990), Language Learning Strategies: What Every Teacher Should Know. Boston, MA: Heinle & Heinle. Papp, Sz. (2007), ‘Inductive learning and self-correction with the use of learner and reference corpora’, in: E. Hidalgo, L. Quereda and J. Santana (eds) Corpora in the Foreign Language Classroom. Amsterdam: Rodopi. 207220. Papp, Sz. (forthcoming), ‘Portsmouth Chinese-English learner corpus’, in: S. Granger, E. Dagneaux, F. Meunier and M. Paquot (eds) The International Corpus of Learner English. Handbook and CD-ROM. Version 2. Louvainla-Neuve: Presses Universitaires de Louvain. Poulsen, S. (2005), Collocations as a language resource: a functional and cognitive study in English phraseology. Unpublished PhD dissertation, University of Southern Denmark. Seidlhofer, B. (2004), ‘Research perspectives on teaching English as a lingua franca’, Annual Review of Applied Linguistics, 24: 209-239. Sinclair, J. (1991), Corpus, Concordance, Collocation. Oxford: Oxford University Press. Sorace, A. (1993), ‘Incomplete vs. divergent representations of unaccusativity in non-native grammar of Italian’, Second Language Research, 9: 22-47. Stubbs, M. (1995), ‘Collocations and semantic profiles: on the cause of the trouble with quantitative studies’, Functions of Language, 2: 23-55.

Creativity by Chinese learners of English

81

Stubbs, M. (2001), Words and Phrases: Corpus Studies of Lexical Semantics. Oxford: Blackwell. Watkins, D. and J. Biggs (eds) (1996), The Chinese Learner: Cultural, Psychological, and Contextual Influences. Hong Kong: University of Hong Kong. Watkins, D. and J. Biggs (eds) (2001), Teaching the Chinese Learner: Psychological and Contextual Perspectives. Hong Kong: University of Hong Kong. Weinert, R. (1995), ‘The role of formulaic language in second language acquisition: a review’, Applied Linguistics, 16(2): 180-205. Wray, A. (1999), ‘Formulaic language in learners and native speakers’, Language Teaching, 32(4): 213-231. Wray, A. and M.R. Perkins (2000), ‘The functions of formulaic language: an integrated model’, Language and Communication, 20(1): 1-28.

Postverbal subjects at the interfaces in Spanish and Italian learners of L2 English: a corpus analysis Cristóbal Lozano Universidad de Granada Amaya Mendikoetxea Universidad Autónoma de Madrid Abstract The purpose of this article is to characterise the production of postverbal subjects in two ICLE subcorpora (Italian and Spanish). The question has been dealt with before in the literature with emphasis on the production of ungrammatical inversion structures in L2 English of speakers from a variety of L1s, but in quite a scattered, unsystematic and rather intuitive fashion. Our approach seeks to identify the conditions under which learners produce inverted subjects. Based on previous research findings and our review of the theoretical literature, we hypothesise that for Spanish and Italian learners of L2 English, there is a tendency for subject inversion to occur when: the verb is unaccusative (H1), the subject is long or “heavy” (H2), and the subject is new (or relatively new) information or “focus” (H3). While H1 has found confirmation in the L2 literature, H2 and H3 have, to our knowledge, been untested and the facts they describe gone unnoticed in previous research. Our results show that the three conditions are met in the writing of Spanish and Italian L2 speakers of English, despite errors in the syntactic encoding of the structures concerned. Thus, a full account of the production of inverted subjects in L2 English must look at properties which operate at (i) the lexicon-syntax interface, (ii) the syntaxphonology interface, and (iii) the syntax-discourse interface.

1.

Introduction

In this paper we analyse in detail the production of postverbal subjects (V(erb)S(ubject) order) in the writing of Spanish and Italian speakers of English as represented in the two relevant ICLE subcorpora (International Corpus of Learner English; Granger et al. 2002). Our aim is to see if the properties that govern the occurrence of postverbal Ss in native English, as currently analysed in the theoretical and descriptive literature, are the same as those operating in the non-native grammars of Spanish and Italian speakers.1 This study is part of a larger project which seeks to gain an insight into the grammatical processes underlying word order in non-native grammars (L2), as compared to both the mother tongue (L1) and the target language. Thus, we incorporate some of the fundamental tenets of what is known as Contrastive Interlanguage Analysis (CIA) (see e.g. Granger 1996 and Gilquin 2000/2001), which establishes comparisons between: (a) native and non-native speakers (NS vs. NNS), and (b) different non-native speakers (NNS vs. NNS). The former involves a detailed analysis of linguistic features in native and non-native corpora

86

Cristóbal Lozano and Amaya Mendikoetxea

to uncover and study non-native features in the speech and writing of (advanced) non-native speakers. This includes errors, but it is conceptually wider as it seeks to identify overuse and underuse of certain linguistic features and patterns. As for (b), by comparing learner data from different L1 backgrounds, we can gain a better understanding of interlanguage processes and features, such as those which are the result of transfer or those which are developmental, common to learners with different L1s (see Granger 2002: 12-13 and references cited therein). Here, we will be comparing the production of inverted subjects in the writing of advanced Italian and Spanish learners of L2 English, as in (b); NS vs. NNS comparison is addressed in Lozano & Mendikoetxea (in preparation). We first examine in detail the properties of VS order in English vs. Spanish/Italian as presented in the current theoretical literature (Section 2). In Section 3, we look critically at previous L2 studies on postverbal Ss. Our hypotheses are presented in Section 4. Section 5 describes the method used to extract and code data from the corpus. The statistical analysis of the data is presented in Section 6, where we compare (i) production of postverbal Ss both with unaccusative and unergative Vs (6.1), and (ii) the nature of both postverbal and preverbal Ss in terms of “heaviness” (roughly, number of words) (6.2) and discourse status (topic or focus) (6.3). Results are discussed in Section 7, where it is shown that our data confirm previous findings in that postverbal Ss are only found with unaccusative Vs (with some significant differences between L1 Italian and Spanish speakers) and, more interestingly, they show that inversion is the result of two other factors not previously discussed in the L2 literature: S must be syntactically heavy, as well as display new information or focus. Section 8 is the conclusion, where we discuss suggestions for further research. 2.

VS order in English vs. Romance languages

English has been characterised as a “fixed” word order language, as opposed to Italian and Spanish where word order is said to be “free”. In English the occurrence of VS order is highly restricted. It is found mostly in two types of constructions: there-constructions and so-called “inversion” constructions of the type XP-V-S, where (i) XP is an adverbial element, typically expressing time or place and linking the sentence to the prior discourse (in italics in [1]), (ii) V is an intransitive verb, typically expressing existence or appearance on the scene, and (iii) S is often syntactically/phonologically “heavy”, consisting of a noun and a variety of pre- and/or postmodifiers, which introduce new information in the discourse (in bold in [1]). These properties are illustrated in (1) – a corpus example from Birner (1994: 254). (1)

Michael puts loose papers like class outlines in the large file-size pocket. He keeps his checkbook handy in one of the three compact pockets. The six pen and pencil pockets are always full and [PP in the outside pocket] [V go] [NP-SUBJECT his schedule book, chap stick, gum, contact lens solution

Postverbal subjects at the interfaces in Spanish and Italian learners

87

and hair brush]. (Land’s End March 1989 catalog. p. 95, quoted in Birner 1994: 254) Given the interplay of factors involved, the properties of VS order have to be analysed at different levels: (a) the lexicon-syntax interface, to account for the lexico-semantic properties of V and their interaction with the grammatical properties of the structure; (b) the syntax-discourse interface, to account for the discourse status (topic or focus) of the preverbal and the postverbal elements and their interaction with the syntactic properties of the structure; and (c) the syntaxphonology interface, to account for the grammatical/phonological properties of the postverbal S along a “heaviness” scale. We deal with each of these issues in turn. 2.1

2.1.1

Postverbal Ss in English and Spanish/Italian at the lexicon-syntax interface: the Unaccusative Hypothesis Postverbal Ss and the Null Subject Parameter

Both Modern English and Spanish/Italian show canonical SV(O(bject)) order as illustrated in (2), but while in English this order is “fixed”, Spanish and Italian allow combinations like those in (3), where S appears in postverbal2 position (VSO) without a special intonation pattern.3 (2)

(3)

a. Maria has phoned the president. b.María ha telefoneado al presidente. c. Maria ha telefonato al presidente.

Spanish Italian

a. *Has phoned Maria the president. b.Ha telefoneado María al presidente. c. Ha telefonato Maria al presidente.

Spanish Italian

Intransitive sentences in Spanish and Italian may also allow postverbal subjects as in (4), while the corresponding sentences in English are ungrammatical (5). (4)

a. i. Ha hablado Juan. b. i. Ha parlato Gianni.

ii. Ha llegado Juan. ii. E’ arrivato Gianni.

(5)

i. *Has spoken John.

ii. *Has arrived John.

Spanish Italian

The grammaticality of postverbal Ss in languages like Spanish/Italian vs. English has often been attributed to the N(ull) S(ubject) P(arameter) (see inter alia Chomsky 1981, Rizzi 1982, Burzio 1986, Fernández Soriano 1989, Jaeggli & Safir 1989, and more recently, Rizzi 1997, Luján 1999, Zagona 2002 and Eguren & Fernández Soriano 2004). “Free inversion” is among the cluster of properties

88

Cristóbal Lozano and Amaya Mendikoetxea

that distinguish languages that allow null Ss (i.e. are positively marked for the NSP), like Spanish and Italian (6), from languages which do not allow null Ss, like English (7).4 (6)

a. i. Ha hablado. ii. Ha llegado. Spanish b.i. Ha parlato. ii. E’ arrivato. Italian ‘(He/She) has spoken.’ ‘(He/She) has arrived.’

(7)

i. *Has spoken.

ii. *Has arrived.

In the generative grammar literature, the possibility of having null Ss has been associated with “rich” (verb) agreement features (e.g. Rizzi 1982), which allows for the recovery of the “content” of a phonologically null pronominal element (pro) in the subject position in (6), while languages which do not allow null Ss have “poor” V agreement and lack pro as part of their lexical inventory (e.g. French [see note 4] and English).5 As for sentences with postverbal Ss, a null element is also postulated for the preverbal subject position of structures like those in (3b, c), (4) and (8a), which we take to be the specifier of the Inflectional Phrase, <Spec, IP> (cf. note 2): expletive pro (proexpl). This is the null equivalent of the overt expletives in languages negatively marked for the NSP, such as French il (8b) or English there (8c). (8)

a. [IP proexpl b. [IP Il c. [IP There

[VP llegaron tres chicas]]. [VP est arrivé trois filles]]. [VP arrived three girls]].

Spanish French

Like its non-overt counterpart in Italian/Spanish, expletive there is involved in the licensing of the postverbal S in English: proexpl and there enter a coindexing relation with the postverbal NP subject, which involves sharing of Case as well as (Person and Number) agreement features between the expletive and the associate NP (not for French, though, as arrivé in [8b] is singular).6 Notice, however, that the presence of there in English does not render the structure in (3a) with a transitive V grammatical (as opposed to [3b, c] with proexpl), as shown in (9a), and only certain intransitive Vs allow there-insertion: (9b), which is the result of adding there to (5i) with speak, is ungrammatical, as opposed to (8c) with arrive. (9)

a. *There has phoned Maria the president. b. *There has spoken John.

This is so because there-constructions show certain constraints that are not found in constructions with proexpl in Spanish and Italian. For a proper account of the distribution of there we have to refer to what is known as the Unaccusative Hypothesis.

Postverbal subjects at the interfaces in Spanish and Italian learners 2.1.2

89

Postverbal Ss and the Unaccusative Hypothesis

Since Perlmutter (1978) first formulated the Unaccusative Hypothesis, it is commonly assumed in the generative grammar literature that there are two classes of intransitive Vs, which differ in the position occupied by their only argument: unergative Vs have an external argument but no internal argument, while unaccusative Vs have an internal argument, but no external argument. The internal argument occupies the position of complement of V, while the external argument is in the preverbal subject position (<Spec, IP>) (see Chomsky 1981, Burzio 1986). More recently, after the introduction of the VP-internal subject hypothesis (see Koopman & Sportiche 1991), by which all arguments are generated internal to the VP, external arguments are generated in the position of the specifier of the VP (<Spec, VP>), as in (10), and subsequently move to <Spec, IP>.7 Thus, unergative Vs appear in initial structures (D[eep]-Structures) like (11a), while unaccusatives appear in initial structures like (11b). (10)

(11)

a. unergative ‘John spoke’

b. unaccusative ‘Three girls arrived’

(D-Structure)

90

Cristóbal Lozano and Amaya Mendikoetxea

The syntactic distinction between the two classes of Vs is systematically related to a semantic distinction: roughly, unergatives typically denote activities controlled by an agent (speak in [11a] and also cry, cough, sweat, jump, run, dance, work, play, etc), while unaccusatives have themes (or patients) as their only argument (arrive in [11b] and also blossom, appear, exist, deteriorate, arrive, come, etc). In the course of the derivation, that is, in the mapping from D-Structure to S(urface) Structure, the NPs in (11) move to <Spec, IP> in order to satisfy their Case requirements (i.e. to be assigned nominative Case) and/or the requirement that <Spec, IP> in English must be occupied by an overt element (roughly, Chomsky’s [1981] Extended Projection Principle; see note 6), as in (12). (12)

a. [IP [NP Johni] [VP [NP ti] [V’ spoke]]]

Unergative

b. [IP [NP Three girlsi] [VP [V arrived] [NP ti]]]

Unaccusative

While this movement is obligatory for external arguments like John, as in (12a), the internal argument of an unaccusative V may, however, remain in its initial position under certain conditions, as in there-constructions like (13) and inversion constructions like (1), now reproduced as (14), with there and PP in <Spec, IP>, respectively. (13)

[IP [NP There] [VP [V arrived] [NP three girls]]]

(14)

[IP [PP in the outside pocket] [VP [V go] [NP his schedule book, chap stick, gum, contact lens solution and hair brush]]]

Characteristically, inversion constructions involve an opening XP adverbial in <Spec, IP>: typically a locative element as in (14) and (15a) below, which are examples of locative inversion, but also time adverbials, as in (15b), as well as other types like the with-PP in (15c).8 (15) a. On one long wall hung a row of Van Goghs. b. Then came the turning point of the match. c. With incorporation, and the increased size of the normal establishment came changes which revolutionized office administration. (corpus examples from Biber et al. 1999: 912-913) There is a considerable overlap between Vs appearing in thereconstructions and typical locative inversion structures like (14) and (15a). For both types of constructions there is the requirement that the V be unaccusative, but this is not by itself sufficient to characterise the class of Vs appearing in these

Postverbal subjects at the interfaces in Spanish and Italian learners

91

structures.9 Two main semantic classes of unaccusative Vs have been distinguished in approaches which seek to characterise unaccusativity at the lexicon-syntax interface in order to establish a relation between the semantic properties of the predicate and the syntactic properties of the constructions they enter: (i) Vs of change of state and (ii) Vs of existence and appearance (Levin & Rappaport Hovav 1995). Only the latter are found in this construction. That is, together with be, we find Vs such as appear, arise, ascend, emerge, exist, come and so on, but not unaccusative Vs of change of state like those in (16): (16)

a. *There opened the door. b. *There broke a window.

(= The door opened) (= A window broke)

It was with these considerations in mind that the Vs for this study were carefully chosen (see Section 5.2 and Table 2). In contrast, as we have seen, null expletives in Spanish and Italian may be found wherever we have a postverbal S, regardless of whether V is unaccusative or unergative. Their occurrence is not determined by properties operating at the lexicon-syntax interface. Rather, Romance “free” inversion appears to be governed by features that operate at the syntax-discourse interface, at least in informationally non-neutral contexts. This could, in principle, lead to the expectation that Spanish and Italian learners of English may overgeneralise and produce postverbal subjects with all verb classes. However, we will see that there are reasons to predict that VS structures in the NNS of our learners are found with exactly the same subset of Vs as in English NS. This issue will be addressed in our discussion of previous L2 research below (Section 3). 2.2

2.2.1

Postverbal Ss in English and Spanish at the syntax-discourse interface: inversion as focalisation Inversion as focalisation

It has been argued recently that information structure notions, such as topic and focus, appear to play a crucial role in the position of S in free-order languages, with preverbal Ss usually analysed as topics or given information, and postverbal Ss as (presentational/informational) focus or new information (see, among others, Vallduví 1990, Fernández-Soriano 1993, Liceras et al. 1994, Picallo 1998, Zubizarreta 1998, 1999, Belletti 2001, 2004b, Domínguez 2004, Lozano 2006). The following question-answer pairs from Italian and Spanish clearly indicate that the postverbal S can be interpreted as (non-contrastive) focus, i.e. new information. When S is preverbal, it is interpreted as old information (topic), which explains why the (c) examples in (17) and (18) are pragmatically-odd (marked as #).10

92

Cristóbal Lozano and Amaya Mendikoetxea

(17)

a. ¿Quién ha llegado/hablado? who has arrived/spoken? b. Ha llegado/hablado Juan. has arrived/spoken Juan c. #Juan ha llegado/hablado. Juan has arrived/spoken

Spanish

(18)

a. Chi è arrivato/ha parlato? who has arrived/spoken? b. É arrivato/Ha parlato Gianni. has arrived/spoken Gianni c. #Gianni è arrivato/ha parlato. Gianni has arrived/spoken

Italian

Additionally, Belletti (2004b) points out that in Italian, (18b) is also an appropriate answer to neutral, out-of-the-blue questions like (19), i.e. when the whole clause is new information, as opposed to the sentence with the preverbal S in (18c) (examples from Belletti 2004b: 22). (19)

Che cosa è sucesso? What has happened?

Empirical studies on Spanish native speakers’ acceptability preferences show, however, that verb choice may determine whether the answer to questions like (19) may contain either a preverbal or a postverbal S. In particular, preverbal Ss are favoured for unergative Vs and postverbal Ss in unaccusative contexts (Hertel 2003, Lozano 2003, 2006) (see also Pinto 1997 for similar observations for Italian), as in (20). (20)

a. i. Juan ha hablado. Juan has spoken b. i. #Juan ha llegado. Juan has arrived

ii. #Ha hablado Juan. has spoken Juan ii. Ha llegado Juan. has arrived Juan

Though this issue requires further analysis, it is largely irrelevant for the present study as our data is extracted from written pieces of work in which we are not likely to find neutral or out-of-the-blue contexts (even the initial sentence is contextualised by the title of the essay). Thus, independently of whether the preverbal S can also be focus or new information in Romance Null S languages, what is crucial for our purpose is that in these languages inversion serves the purpose of focalising the subject. Notice also that in languages like Italian, Spanish and English, new information in clause initial position can be interpreted as contrastive focus (see Rizzi 1997), as shown in (21) in capitals.

Postverbal subjects at the interfaces in Spanish and Italian learners (21)

93

a. IL TUO LIBRO ho comprato (non il suo). b. TU LIBRO he comprado (no el suyo). c. YOUR BOOK I have bought (not his).

From this, it follows that preverbal subjects are not necessarily topics or old information, but may also be analysed as contrastive focus. Non-contrastive (i.e. informational or presentational) focus, of the type we are looking at here, occurs postverbally (see also note 10). Let us now turn to English, which makes a much more restricted use of inversion mechanisms, as we have seen. Postverbal Ss in English thereconstructions and locative inversion structures are also often analysed as (presentational) focus (see, among others, Bolinger 1977, Rochemont 1986, Bresnan 1994). In Bresnan’s (1994) analysis of locative inversion, the referent of the postverbal NP is introduced, or reintroduced, on the scene referred to by the preverbal PP: for instance in (15a) on one long wall provides the scene onto which a row of Van Goghs is introduced, which is characterised as a new discourse entity. A different approach is adopted by Birner (1994, 1995), who has argued on the basis of a large corpus of naturally-occurring tokens that the discourse function of all inversion constructions is that of “linking relatively unfamiliar information to the prior context through the clause-initial placement of information that is relatively familiar in the discourse” (Birner 1995: 238). Thus in (14), in the outside pocket is relatively more familiar than the material in postverbal position, as in the preceding context the speaker has mentioned the different pockets in which things are placed (see [1] above). The felicity of an inversion depends on the notion of relative discourse-familiarity: the preverbal constituent may not present “newer” information in the discourse than the postverbal constituent. This implies that the postverbal constituent need not always be discourse-new; it may, on the contrary, represent quite familiar information, given the appropriate discourse conditions, i.e. as long as it is relatively less familiar than the preverbal constituent. For Birner (1995), this cannot be accounted for by an analysis in which the postverbal NP is characterised as presentational focus and, therefore, discourse-new. Given that both in English and in Italian/Spanish, inversion is used as some sort of focalisation device, we would expect the inverted Ss in the NNS of our study to be discourse-new or focus. It has to be stressed, however, that Italian and Spanish make use of this device with all verb types, whereas in NS English inversion appears to be restricted mostly to unaccusative Vs of existence and appearance, which in principle could lead to the expectation that our learners will use postverbal subjects in syntactic contexts in which it is not found in native English, as pointed out at the end of Section 2.1. We will return to this issue in Section 3 below. 2.2.2

Brief remarks on the definition of focus and topic

Focus and topic have been used here as common labels for new and old information, respectively, but these labels are just a first approximation to name

94

Cristóbal Lozano and Amaya Mendikoetxea

concepts and processes which may be adequate to express the relation between information structure and syntactic configurations (see Belletti 2004b: 42, note 1, for a similar comment). Regarding the old vs. new information dichotomy, Prince (1981, 1992) distinguishes between hearer status (hearer-old. vs. hearer-new) and discourse status (discourse-old vs. discourse-new) and argues that only the latter distinction is relevant for subjecthood. There is also a third status for an entity in discourse: inferrables, defined by Prince (1992: 312) as “NPs evoking entities which were not previously mentioned and which I as the reader had no prior knowledge of, but whose existence I could infer on the basis of some entity that was previously evoked and some belief I have about such entities”. Compare the following examples used by Prince (1992: 305) to characterise inferrables: (22)

a. He passed by the door of the Bastille and the door was painted purple. b. He passed by the Bastille and the door was painted purple.

In (22a) the door is discourse-old, as it has been mentioned before, the hearer is assumed to have a mental representation of it. In contrast the door in (22b) is not strictly speaking discourse-old, but it is treated as though it were already known to the hearer. Indeed, the hearer knows that the speaker is talking about the door of the Bastille, so it is not quite discourse-new. As Prince (1992) notices, inferrables have a double status as hearer-old (and discourse-old) entities, in that they rely on the earlier presence of another entity (e.g. the Bastille) triggering the inference, and as hearer-new (and discourse-new), in that the hearer is not expected to already have in his or her head the entity concerned. The question that arises is whether they can be collapsed with one or the other category, whether they form a separate category, or whether we have some sort of information continuum, with inferrables somewhere in the middle. The continuum or “gradience” approach is adopted by Kaltenböck (2005) in his corpus analysis of it-extraposition structures in English. He defines old (or given) information as “retrievable” from the preceding co(n)text and new information as not thus derivable (see Geluykens 1991 and Firbas 1992 on the concept of retrievability – or recoverability). Retrievability is a relative concept: some entities are more retrievable than others, depending on whether they can be derived directly (evoked) from the context (as in [22a]) or indirectly via inferences (inferrable), where context comprises both the verbal “co-text”, as well as the situational context. This approach looks at the finished product of verbal interaction, the corpus text, rather than focusing on speakers’ assumptions (referred to as the “psychologist’s view”) (see Kaltenböck 2005: Section 4.1). In our study, both evoked and inferrable entities are considered to be topics, on the basis of Prince’s (1992) analysis of Ss and Birner’s (1994, 1995) findings that both entities are treated alike (as discourse-old) in inversion structures. As for new or “irretrievable” information, it can either be brand-new or new-anchored (an irretrievable state of affairs or entity, which is in some way linked to – “anchored in” – the previous context). We consider these as focus, together with what Prince (1992) refers to as unused (discourse-new, hearer-old

Postverbal subjects at the interfaces in Spanish and Italian learners

95

entities). An example of an unused entity is given in (23a) (from Prince 1992), where the public is discourse-new but the reader is expected to know of its existence. As for new-anchored, notice that the postverbal subject in the corpus example in (23b) is new in the sense that the idea it introduces has not been mentioned before, but it is anchored in that this work refers to what is mentioned in the preceding context. (23)

a. … from reporters eager to tell the public about … (from Prince 1992: 312) b. So should scientists thus be condemned for allowing it to become possible for these women to have babies? Certainly there exists a demand for this work to be done. (from the LOCNESS corpus, file ICLE-ALEV0014.8)

We thus consider topic and focus as concepts which encompass a variety of notions which are best analysed in terms of a gradience.11, 12 2.3

Postverbal Ss at the syntax-phonology interface

Together with discourse notions such as “newness” or “focus”, choice of ordering is influenced by properties to do with the phonetic realisation of the strings of words generated by the grammar. Thus, at the syntax-phonology interface, there can be operations affecting the linear ordering of constituents which are not triggered by syntactic features. This is presumably what happens in structures involving “Heavy NP-Shift”, where a “heavy” NP appears to have been “displaced” to the end of the sentence.13 Thus, while (24a) shows canonical VNP-PP, the non-canonical V-PP-NP order in (24b) is also possible with a “heavy” NP. (24)

a. I bought [NP a book written by a specialist in environmental issues] [PP for my sister]. b. I bought [PP for my sister] [NP a book written by a specialist in environmental issues].

The generalisation that “heavier” constituents should follow “lighter” constituents is essentially what is known as the “end-weight” principle (Quirk et al. 1972). Heaviness can be defined simply as a matter of string length (number of words) or on the basis of more sophisticated criteria to do with grammatical complexity (see Arnold et al. 2000 for a review of these two approaches). In fact, the two concepts are difficult to separate neatly, as revealed by Wasow’s (1997) corpus study, which shows high correlations among the various characterisations of heaviness. In sum, long and complex elements tend to be placed towards the end of the clause, an operation which reduces the processing burden and, thus, eases comprehension by the receiver (see Hawkins 1994). Since long and complex grammatical elements typically also carry new information, the endweight principle and the discourse principle by which new information tends to

96

Cristóbal Lozano and Amaya Mendikoetxea

be placed towards the end of the clause (focalisation) appear to reinforce each other (see Biber et al. 1999: 898). As pointed out by Arnold et al. (2000: 34) “items that are new to the discourse tend to be complex, and items that are given tend to be simple”. The end-weight principle appears to be in operation also in structures involving postverbal Ss. Culicover & Levine (2001) claim that certain structures which appear to be the result of locative inversion become a good deal less natural with “lighter NPs”, as shown in (25) (highlighting is ours), and analyse such structures as involving Heavy-NP shift. (25)

a. *Into the room walked carefully Robin. b. Into the room walked carefully the students in the class who had heard about the social psych experiment that we were about to perpetrate. (from Culicover & Levine 2001: 291)

Our own analysis of the corpus examples used by Levin & Rappaport Hovav (1995) (L&RH) in their study of locative inversion reveals, indeed, that overwhelmingly the postverbal S is heavy. When it is a proper noun or a relatively light NP, it is normally followed by material in apposition, as in (26) (highlighting is ours). (26)

a. And when it is over, off will go Clay, smugly smirking all the way to the box office, the only person better off for all the fuss. (R. Kogan, “Andrew Dice Clay Isn’t Worth ‘SNL’ Flap” 4, cited in L&RH: 221) b. Above it flew a flock of butterflies, the soft blues and the spring azures complemented by the gold and black of the tiger swallowtails. (M. L’Engle A Swiftly Tilting Planet, 197, cited in L&RH: 257)

The gradience approach adopted for information status is also adopted in our study for “heaviness”: the heavier an NP, the more likely it is to be placed in clause-final position. This approach has the additional advantage that it can be used to measure “relative” weight, which has been shown to be relevant in studies involving two constituents which may appear in different orders, such as the direct and indirect object of ditransitive Vs (Arnold et al. 2000). The relatively “free” word order of Spanish and Italian means that the principle of end-weight may be less noticeable in these languages. Given that its general purpose appears to be related to easing the processing burden on the receiver, by placing long, complex (and new) elements towards the end of the clause, we will assume that this is a universal processing principle (Frazier 2004). Thus (27a), which shows canonical word order in Spanish with the adjunct PP following the NP complement, appears to be less “natural” than (27b), where the heavy object follows the PP adjunct.

Postverbal subjects at the interfaces in Spanish and Italian learners (27)

97

a. Vi [NP los libros que tanto gustaban al padre de tu vecino] [PP en la librería]. I saw [NP the books that your neighbour’s father liked so much] [PP in the bookshop] b. Vi [PP en la librería] [NP los libros que tanto gustaban al padre de tu vecino].

The conclusion, then, is that long and complex information tends to be placed at the end in both English and Spanish/Italian. Therefore, we expect learners to produce postverbal subjects which are long and complex, as a reflection of this general processing mechanism. As we have seen, the principle of end-weight interacts with information structure principles which operate at the syntax-discourse interface, by which (discourse-) new information tends to be placed towards the end of the clause. Thus, Ss which are focus, long and complex tend to occur postverbally in those structures which allow them in both English, on the one hand, and Spanish and Italian, on the other hand. This is also the prediction made for the learners in our study.14 3.

Previous L2 findings

Little is known about the production of postverbal Ss in L2 English with intransitive Vs, apart from the fact that they are found with unaccusative Vs. White’s (1985, 1986) studies on subject-verb inversion, as a feature of the NSP, which are based on grammaticality judgements, are rather inconclusive, due mostly to verb choice, as pointed out by Oshita (2004: 106). Two production studies, Zobl (1989) and Rutherford (1989), support the hypothesis that learners with Spanish L1 produce VS only with unaccusatives, but are rather small in scale. Zobl (1989) reports from a small corpus that learners of L2 English with Null S L1s (mostly Japanese, Spanish and Arabic) occasionally produce postverbal Ss with unaccusatives, as shown in (28) (highlighting is ours). (28)

a. I was just patient until dried my clothes. (L1 Japanese) b. Sometimes comes a good regular wave. (L1 Japanese) (Zobl 1989: 204)

Similarly, Rutherford (1989) reports that L1 Spanish and Arabic speakers produce postverbal Ss with unaccusatives in a written corpus of L2 English, as in (29) (highlighting is ours), but never with unergatives.

98 (29)

Cristóbal Lozano and Amaya Mendikoetxea a. In the town lived a small Indian. (L1 Spanish) b. On this particular place called G… happened a story which now appears on all Mexican history books…. (L1 Spanish) c. After that, they’ll be lead to their house, and with that comes the end of the wedding. (L1 Arabic) d. The bride was very attractive, on her face appeared those two red cheeks… (L1 Arabic) (Rutherford 1989: 178-179)

Zobl (1989) and Rutherford (1989), however, differ in their explanation of why VS order is found with unaccusative verbs. For Zobl (1989), this word order results from the syntactic S of an unaccusative sentence being actually the logical or semantic object of V or the theme,15 what we have been referring to as its internal argument (see Section 2.1.2). This word order preserves at S-Structure the position occupied by the internal argument at D-Structure (see [12b] above) and therefore we do not expect unergatives to appear in VS structures since their only argument is an external argument (see [12a]). VS structures are just one type of solution to the problem of how learners map logical relations to surface structure configurations in languages like English, where Ss and Os are defined in terms of their structural position (see examples [1]-[6] in Zobl [1989: 204] for other solutions learners come up with). For Zobl (1989), the production of VS structures is developmental and precedes a stage in which learners are able to determine the canonical alignment between semantic roles and syntactic structure. Rutherford (1989), however, argues that VS order in the learner English of Spanish and Arabic L1 speakers is the direct result of transfer. His main hypothesis is that “the tendency for canonical word order permutation in written ILs [interlanguages] will correlate directly with the propensity of the learner’s native language to permute its own canonical constituents” (Rutherford 1989: 166). Crucially, Rutherford’s prediction is that not all word order differences will lead to transfer. Transfer is expected when the learners’ L1 has what Rutherford (1989) refers to as “Pragmatic Word Order” (PWO), but not when L1 and L2 show different “Grammatical Word Orders” (GWO). Thus, Spanish and Italian are basically SVO languages, like English, and, as was discussed in Section 2 above, variations on this word order are mostly due to properties operating at the syntax-discourse interface, showing what Rutherford refers to as PWO. On the other hand, Japanese and English differ in their canonical word order: SOV vs. SVO. While transfer is expected in the L2 English of Italian and Spanish (as well as Arabic) L1 speakers, this is not the case for Japanese L1 speakers. Rutherford’s predictions are borne out by his results. While no violations of word order were found in the written production of 21 Japanese L1 learners of English (but see Zobl 1989), the 59 compositions produced by Spanish L1 learners of English contained 20 instances of violations of English canonical word order and in virtually all cases the resulting order was (X)VS (see Rutherford 1989: 168). For Rutherford (1989), this is the result of transfer from PWO L1s. However, he offers no explanation as to why (X)VS order in the IL of Spanish (and Arabic) L1

Postverbal subjects at the interfaces in Spanish and Italian learners

99

speakers is restricted to a definable class of lexical verbs: those which we have referred to as unaccusative verbs of existence and appearance, following Levin & Rappaport Hovav (1995). It has to be mentioned that both Zobl’s (1989) and Rutherford’s (1989) conclusions on VS production by Spanish L1 learners of English are based on a relatively small number of VS instances. This is roughly 20 instances in Rutherford (1989). As for Zobl’s (1989) study, 13 examples of VS order are found in the written work of 114 informants from different language backgrounds, of which 90 are Japanese and 10 Spanish L1 speakers; 10 of those examples were produced by Japanese speakers, which is used by Zobl (1989) to rule out transfer, contrary to Rutherford’s (1989) results. Though these studies point out that the unergative-unaccusative distinction is psychologically real for learners, not enough data is provided to support their claims and very little information is provided about the learners, sample size and so on. Oshita’s (2004) is the first study to use a large electronic corpus: Longman Learners’ Corpus (version 1.1, March 1993, LLC). His purpose is to investigate the psychological reality of null expletives, for which he extracts 941 token sentences on 10 common unaccusative verbs and 640 token sentences with 10 common unergative verbs from compositions written by speakers of Italian (684), Spanish (1,079), Japanese (1,363) and Korean (236). For VS order, he extracts sentences with preverbal overt expletives (i.e. it, there) and null expletives. L1 Spanish and Italian learners produced postverbal Ss only with unaccusatives (never with unergatives), as in (30) (highlighting is ours). Both groups’ production ratios were similar: 14/238 (6%) Spanish; 14/346 (4%) Italian. (30)

a. …it will happen something exciting... (L1 Spanish) b. …because in our century have appeared the car and the plane… (L1 Spanish) c. …it happened a tragic event… (L1 Italian) d. One day happened a revolution. (L1 Italian) (Oshita 2004: 119-120)

Additionally, in a study of existential there-V-S constructions, PalaciosMartínez & Martínez-Insua (2006) report that Spanish learners of English (ICLE corpus) use the verb be, except for 5 tokens with the unaccusative exist, which represents 1.1% of all existential there-constructions. While this study does not address the issue of unaccusativity and postverbal Ss, it reveals that the only verb, other than be, that appears in these postverbal S structures is an unaccusative verb. All these studies show a remarkably consistent pattern in which unaccusative and unergative verbs are treated differently by learners of English regarding the occurrence of postverbal Ss. This adds to other type of evidence, provided in Oshita (2004), which points towards the fact that the Unaccusative Hypothesis, that is, the unaccusative-unergative distinction, is psychologically real in L2 acquisition, as demonstrated by studies on learners’ auxiliary selection

100

Cristóbal Lozano and Amaya Mendikoetxea

(Sorace 1993, 1995), the production of “passivised” unaccusative structures (Zobl 1989, Oshita 2000) and learners’ reluctance to accept SV order with unaccusatives (Oshita 2002). This is despite the fact that English lacks overt marking for unaccusatives, rendering the unergative-unaccusative distinction inaccessible: unaccusative Vs overwhelmingly appear in SV constructions and there is no auxiliary or morphological marker for unaccusative Vs. That is, although inversion structures with unaccusatives are found in English, as discussed at length in Section 2 above, the rarity of the construction makes it unlikely that such VS order is sufficiently represented in the input data to learners to count as positive evidence (see, by way of illustration, the percentages given in Biber et al. [1999: 945] for existential clauses with Vs other than be). Without entering the debate of whether VS order is developmental or the result of transfer, our study should be considered as part of the growing body of research which takes the Unaccusative Hypothesis to be psychologically real for L2 learners of English from speakers of Null Subject languages like Italian and Spanish, which is the reason for predicting that VS order will be found only with unaccusative verbs in our corpus (see Section 4 below). There are two aspects in which our study differs considerably from previous L2 studies: (i) we intend to show that unaccusativity is a necessary but not sufficient condition for VS order to occur, as it is not just the nature of V (unaccusativity), but crucially the features of S (heaviness and focus) that trigger inversion, and (ii) unlike previous L2 studies, we do not focus on errors, but rather on the conditions under which postverbal subjects are found. 4.

Hypotheses

As a general hypothesis, we expect Spanish and Italian advanced learners of L2 English to produce postverbal subjects in the same contexts in which inversion takes place in native English, given that, except for the condition on unaccusativity, the native English contexts are substantially the same as those for Spanish and Italian, as we have seen. We do not expect significant differences between the groups of Spanish and Italian learners. In particular, in accordance with previous L2 findings and the theoretical analysis in Section 2, we can hypothesise the following, (31-33). (31)

(32)

(33)

H1. Lexicon-Syntax interface: As reported in previous research, both groups of learners will produce postverbal Ss with unaccusatives only, but never with unergatives, as in L1 English. H2. Syntax-Phonology interface: Given our theoretical analysis, we predict that, in those contexts where inversion is allowed, both groups of learners will tend to place Ss in postverbal position when S is heavy. H3. Syntax-Discourse interface: Given our theoretical analysis, we predict that, in those contexts where inversion is allowed, both groups of learners will tend to place Ss in postverbal position when S is the focus.

Postverbal subjects at the interfaces in Spanish and Italian learners 5.

Method

5.1

Corpora

101

We used the Spanish and Italian subcorpora of the International Corpus of Learner English, ICLE (Granger et al. 2002), which consists of 11 subcorpora of academic essays written by advanced L2 English learners of 11 different L1s. In total, nearly half a million words were used in our analyses (Table 1). Table 1. Corpora Corpus ICLE Spanish ICLE Italian TOTAL 5.2

Number of essays 251 392 643

Number of words 200,376 227,085 427,461

Analysis of concordances

Following Levin (1993) and Levin & Rappaport Hovav (1995), we constructed an inventory of unaccusative (n = 32) and unergative (n = 41) lemmas in English (Table 2). The unaccusative Vs selected belong to those semantic classes that can be found in inversion structures in English (see Section 2.1.2). Searches were then performed with WordSmith Tools 4.0 (Scott 2002) on all verbal forms of the lemmas. All possible forms of the lemma (both native English and possible misspelt learner forms) were queried, e.g. for the lemma APPEAR: appear, appears, appearing, appeared, appeard, apear, apears, apearing, apeared, apeard. Additionally, the sentences under analysis were both grammatical and ungrammatical in native English since, as stated previously, we are not interested in ungrammaticality but rather in the conditions under which postverbal Ss appear. The concordances output by WordSmith were filtered manually according to 51 criteria to discard those structural contexts in which inversion in English is not possible, regardless of the nature of V. Approximately three quarters of the concordances output by the software turned out to be unusable since they did not meet the filtering criteria. For conciseness, we present here only the main filtering criteria, cf. (34) (see Lozano & Mendikoetxea in preparation for details on the filtering criteria). (34)

a. V must be intransitive (unaccusative or unergative, with either preverbal or postverbal S). We discarded (un)grammatical uses of transitive unaccusatives (e.g. parents grew their children) and letconstructions (e.g. let him develop).

102

Cristóbal Lozano and Amaya Mendikoetxea b. V must be finite. We discarded instances of to-infinitive clauses (e.g. He would like to leave), infinitival clauses (e.g. to avoid this happen again), gerundive clauses (e.g. Returning to the title of this paper). c. V must be in the active voice with(out) auxiliary or modal. We discarded case of passivised unaccusatives (e.g. This situation has already been happened). d. S must be a Noun Phrase. We discarded cases of the NP being a clause, as in extraposition (e.g. it happened that the countries which make the weapons are…).

Table 2. Inventory of unaccusative and unergative verbs Unaccusatives Unergatives Semantic class: Semantic class: Semantic subclass: Existence: exist, flow, Emission Light emission: beam, grow, hide, live, remain, burn, flame, flash rise, settle, spread, survive Sound emission: bang, beat, blast, boom, clash, Appearance: appear, crack, crash, cry, know, arise, awake, begin, ring, roll, sing develop, emerge, flow***, Smell emission: smell follow, happen, occur, Substance emission: rise*** pour, sweat Disappearance: die, disappear Inherently directed motion: arrive, come, drop, enter, escape, fall, go, leave, pass, rise***, return

Communication

Manner of speaking: cry*, shout, sing* Talk verbs: speak, talk

Bodily processes

Breathe verbs: breathe, cough, cry*, sweat** Nonverbal expressions: laugh, sigh, smile

Manner of motion

Run verbs: fly, jump, run, swim, walk, ride, travel, slide

Performance

Monadic agentives: dance, phone, play, sing, work

sleep Snooze Total unaccusatives: 32 Total Unergatives: 41 Notes: (*) see also sound emission; (**) see also substance emission; (***) see also existence.

Postverbal subjects at the interfaces in Spanish and Italian learners

103

Consider now the analysis of weight for the (pre- and postverbal) S of the concordances. Length in number of words has been typically used to measure heaviness (see Section 2.3). This measuring method is legitimate, yet it is unclear in previous studies where the dividing line between heavy/light lies, i.e. it is not known how long (in words) a constituent must be for it to be classified as heavy. Hence, we used three types of measurement to arrive at a satisfactory nominal heavy/light scale (Table 3): (i) a numeric scale (length in words, as in earlier research), (ii) an ordinal scale (ranging from 0 [lightest] to 3 [heaviest]) based on syntactic structure, as shown in the table, and (iii) a dichotomous nominal scale (light vs. heavy) based on the ordinal scale where the “light” category corresponds to ranks 0 and 1 in the ordinal scale, while “heavy” corresponds to 2 and 3. To our knowledge, there is no standardised scale linking syntactic structure to weight. We followed a basic (yet intuitive) principle to link syntactic structure to weight. The syntactic structure corresponding to “light” implies a basic structure with a pronoun (PRN) or a noun (N) as the head, with basic premodification and postmodification: either Ø, or a determiner (D) and/or an adjective (ADJ). The rest of the combinations were regarded as “heavy”, as shown in the table. Table 3. Scales for measuring weight Nominal scale

Ordinal scale

Syntactic structure

LIGHT

0

(D)

N PRN 1 (D) ADJ N HEAVY 2 (D) N PP (D) ADJ* N (D) (ADJ) N* 3 (D) N PP* (D) (ADJ) N AdjP* (D) ADJ N PP (D) N IP/CP (D) (ADJ) N* PP* (D) ADJ N* (PP*) Notes: (i) The asterisk (*) represents a complex (i.e. recursive) categorical or phrasal structure. (ii) Parentheses indicate the optional realisation of the bracketed category or phrase. To ensure that the nominal scale is a reliable measurement of the numeric scale, we performed a correlation test between the numeric and the ordinal scale. Results reveal that they are highly correlated [ȡ = 0.914, p<0.001]. Hence, it can be safely assumed that the division between ranks 0 and 1 (light) vs. ranks 2 and 3 (heavy) is a reliable dividing line between what we considered to be light vs. heavy subjects.

Cristóbal Lozano and Amaya Mendikoetxea

104

As for the analysis of the discursive status of S, each (pre- and postverbal) S in the selected concordances was coded as either topic or focus, according to our definition of these terms in Section 2.2.2, by which topic and focus are concepts encompassing a variety of notions which are best analysed in terms of a scale such as the retrievability scale in Kaltenböck (2004, 2005). Coding was performed manually, taking into consideration the preceding discourse and context to determine whether S was topic or focus. Both evoked and inferable entities were coded as focus. Finally, consider the analysis of preverbal unaccusative subjects. The total number of unaccusative SV concordances was 588 for the Spanish corpus and 599 for the Italian corpus. Given that coding manually these 1,147 concordances in terms of weight and discursive status is time-consuming and requires a considerable amount of human resources, we randomly sampled a relatively high number of concordances (minimum 80 for each corpus). The Vs pooled in the sample were the top 4 unaccusative inversion Vs: exist, appear, begin and come (Spanish corpus) and exist, emerge, live and come (Italian corpus), as we will see in Figure 4. The sampling process resulted in the following unaccusative SV concordances: n = 81 Spanish and n = 96 Italian. These amounts will be contrasted against the unaccusative VS concordances (n = 52 Spanish and n = 15 Italian) in terms of weight and discourse status (Sections 6.2 and 6.3). 6.

Results

6.1

H1: Postverbal Ss and unaccusativity

As Table 4 shows, the total raw frequency of usable concordances produced was: 153 unergatives plus 640 unaccusatives for the Spanish corpus, and 143 unergatives plus 574 unaccusatives for the Italian corpus. Crucially, postverbal Ss are produced only with unaccusatives (8.1% for the Spanish corpus vs. 2.6% for the Italian corpus), but never with unergatives (0% for both the Spanish and Italian corpora). This finding supports H1 and previous research. Table 4. Rate of postverbal Ss produced Subcorpus

V type

Spanish

Unergative Unaccusative Unergative Unaccusative

Italian

# postverbal Ss 0 52 0 15

# usable concordances 153 640 143 574

Rate (%) 0/153 (0%) 52/640 (8.1%) 0/143 (0%) 15/574 (2.6%)

Given that learners never produce VS with unergatives, in what follows we will leave aside unergative Vs and will focus on unaccusatives with pre- and postverbal Ss, i.e. unaccusative SV and VS orders.

Postverbal subjects at the interfaces in Spanish and Italian learners

105

Consider now the different rates of unaccusative SV vs. VS orders. The production of postverbal Ss (i.e. VS) is rather low (Figure 1), since the majority of structures produced are SV structures: 91.9% in the Spanish corpus and 97.4% in the Italian corpus. While the VS rates are relatively low, a statistical analysis reveals that the difference in VS production between the two corpora is significantly different: 8.1% in the Spanish corpus and 2.6% in the Italian corpus [Ȥ2 = 17.630, df = 1, p<0.001]. This result is unexpected since, according to our general hypothesis, we expect both groups to show similar rates, given that L1 Spanish and Italian behave similarly regarding the conditions under which VS order is produced. This is discussed below, but crucially note that, (i) as we have just seen, both groups behave similarly at the lexicon-syntax interface by producing VS with unaccusatives only and, (ii) as we will see in the next sections, both groups’ behaviour is similar for the rest of the interfaces under investigation (weight at the syntax-phonological form interface and information status at the syntax-discourse interface). Spanish 8.1%

Italian 2.6%

Unac SV Unac VS

91.9%

97.4%

Figure 1. Preverbal vs. postverbal subjects produced with unaccusatives We should bear in mind that, despite the proportion of postverbal subjects produced by our learners, unaccusative SV concordances represent the majority of productions with unaccusatives, namely 588 (640–52, i.e. 91.9%) in the Spanish corpus and 559 (574–15, i.e. 97.4%) in the Italian corpus, as just seen in Figure 1. The issue of grammaticality/acceptability inevitably arises when dealing with learner data. Recall that we are interested in postverbal subject constructions with unaccusative Vs, regardless of their grammatical status in native English.16 Overall, Figure 2 shows that most unaccusative postverbal Ss produced by our learners are ungrammatical ([35] vs. [36] below), the difference being more marked in the Spanish corpus (65.4% ungrammatical vs. 34.6% grammatical) than in the Italian corpus (53.3% vs. 46.7%), but the difference is not significant [Ȥ2 = 0.723, df = 1, p = 0.395]. In contrast, 100% of preverbal-subject structures with both unergatives and unaccusatives are grammatical in both corpora, as in (36) (codes in round brackets represent the ICLE filename: those filenames starting with an “s” belong to the Spanish subcorpus and those starting with an “i”, to the Italian subcorpus).

106

Cristóbal Lozano and Amaya Mendikoetxea Spanish

Italian 34.6% 46.7% 53.3%

Unac VS Gram Unac VS Ungram

65.4%

Figure 2. Grammatical vs. ungrammatical postverbal subjects (unaccusative VS) (35)

(36)

a. […] in the evolution of the human specie it would disappear the capacity of thought in a near future. (spm04006) b. It is almost disappearing the use of writing nice letters to friends […] (itrs1018) a. Then come the necessity to earn more […] (spm07023) b. Then came psychoanalysis. (itrs1010)

The production of a postverbal S here often implies the presence of preverbal material. Overall, six different types of XP-V-NPSUBJECT were produced, cf. (37)-(42). (37)

(38)

(39)

(40)

Ungrammatical it-insertion: a. I do believe that it will not exist a machine or something able to imitate the human imagination. (spm01007) b. […] and it still live some farmers who have field and farmhouses. (itb07001) Grammatical locative inversion: a. In the main plot appear the main characters: Volpone and Mosca […] (spal1002) b. Cesare Lombroso (1835/1909) criminologal, asserted that on the earth lived people which were born-criminal. (itrl1005) Insertion of any other type of phrase (XP-insertion), which is typically (but not exclusively) a PP: a. There exists a whole range of occ[a]sions in which we have had to be witness of how people from other nations usually fight abroad for foreign causes. (spm10015) b. […], there still remains a predominance of men over women. (itto4006) Ungrammatical Ø-insertion: a. Nevertheless exist other means of obtaining it [i.e. money] which are not so honourable, but quicker. (spm01013) b. Instead I think that exist factors which, on long term, can predispose human mind to that crime […] (itrl1010)

Postverbal subjects at the interfaces in Spanish and Italian learners (41)

(42)

107

AdvP insertion: a. […], and here emerges the problem. (spm01001) b. Later came a world of disorder, during and after the First World War […] (itrs1010) Grammatical existential there-insertion: a. […] and from this moment begins the avarice. (spm04048) b. [No instances of XP-insertion were found in the Italian corpus]

Ungrammatical it-insertion, as in (37), and grammatical locative inversion, as in (38), were the most frequent structures in both groups (Figure 3). Due to space limitations, we will leave aside the study of these different preverbal structures and will focus on the conditions under which post- and preverbal unaccusative Ss are produced. 100%

Group

95%

VS Spanish ICLE

90%

VS Italian ICLE

85% 80% 75%

Production rate (%)

70% 65% 60% 55% 50% 45% 40% 35%

38 33

30% 25%

27

20%

20

15%

15

15

10%

10

5%

12

13

7

on rti se in eer th

on rti se -in vP Ad

on rti se -in Ø

n io rs ve in

on rti se -in XP

c Lo

on rti se in it-

0%

10

Preverbal structure

Figure 3. Preverbal material in unaccusative VS structures (XP–V– NP) Consider, finally, the different types of unaccusative lemmas that trigger VS order. Figure 4 represents, for each unaccusative lemma, its frequency of inversion (in %) out of the grand total number of concordances. We can observe that in the Spanish corpus, out of the 8.1% of VS structures produced (recall Figure 1), the top four inversion unaccusatives are: exist (3.4%), appear (1.7%),

Cristóbal Lozano and Amaya Mendikoetxea

108 Ital

Frequency of inversion (%)

Spa 0.0 APPEAR

ARRIVE

0.0 0.0

AWAKE

0.0 0.0

0.0 0.2 0.2 0.2 0.0 0.0 0.3

ENTER

0.0 0.0

0.5

0.7

EXIST FALL

0.0

FLOW

0.0 0.0

FOLLOW

0.0 0.0

GO

0.0 0.0

GROW

0.0

3.4

0.2

0.2 0.0 0.0

HIDE

0.0 0.0

LEAVE

0.0 0.0

LIVE

0.0

0.3

0.0 0.3 0.0 0.0 0.2 0.2

REMAIN RETURN

0.6

0.2

ESCAPE

PASS

3.5

0.0

0.0 0.0

OCCUR

3.0

0.6 0.3

EMERGE

HAPPEN

2.5

0.0

DISAPPEAR DROP

2.0

0.3

COME

DIE

1.5 1.7

0.0

DEVELOP

1.0

0.0

ARISE

BEGIN

0.5

0.0 0.0

RISE

0.0 0.0

SETTLE

0.0 0.0

SPREAD

0.0

SURVIVE

0.0 0.0

0.2

Figure 4. VS structures produced with each unaccusative lemma begin (0.6%) and come (0.6%), while the rest contribute a small percentage to inversion. In the Italian corpus, the top four inversion unaccusatives are: exist (0.7%), emerge (0.5%), come (0.3%) and live (0.3%). We will return to these

Postverbal subjects at the interfaces in Spanish and Italian learners

109

results (Section 7), as they are crucial to interpret the difference in VS rates between the two corpora, Spanish (8.1%) and Italian (2.6%). 6.2

H2: Postverbal Ss and weight

Consider first the results on weight in the numeric scale (as measured in number of words). This will be crucial to see the differences in the spread of scores between both groups and the actual length of pre- vs. postverbal Ss, which will be the basis to later interpret the results on the nominal scale (heavy/light). The boxplot (Figure 5) represents the spread of weight of unaccusative pre- and postverbal Ss for both groups (Spanish and Italian), with circles representing outliers and asterisks representing extreme cases. While both heavy and light subjects appear in both preverbal and postverbal positions, as the contrast between (43) and (44) shows, a statistical analysis reveals that, for both groups, preverbal Ss are typically light (short), as in (43a, b), while postverbal Ss are typically heavy (long), as in (44a’, b’).

SV Italian ICLE



=

/



/

/

Group

SV Spanish ICLE

=

=

VS Italian ICLE

VS Spanish ICLE

=





1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

Weight (# of words)

Figure 5. Boxplot (with median and mean) of subject weight in number of words

110 (43)

(44)

Cristóbal Lozano and Amaya Mendikoetxea Preverbal unaccusative subjects: light vs. heavy a. […] for the first time, beggars appeared. (spm02003) a’. […] it was in that time when the utopian societies created by the [e]arly socialists appeared. (spm04019) b. Violence does exist […] (itto2034) b’. Nowadays, the differences between men and women should not exist any more […] (itto4006) Postverbal unaccusative subjects: light vs. heavy a. […] and from there began a fire, […] (spm04011) a’. ,[…] and thus began the period known as Restoration, which in literature ended in 1707 on the death of George Farquhar, the last mahor writer of the “Comedy of Manners”. (spm08005) b. We could call it the body language and through it, emerges the protagonists’ personality. (itrs1064) b’. This is conveyed in line 25 where by the expression, emerges the people’s ignorance in having prejudices. (itrs1065)

In particular, in SV structures, both the Spanish and Italian groups behave similarly: the median for both groups is 2, i.e. preverbal Ss containing 2 words divide the spread of weight scores. The mean (which is marked by a plus sign in Figure 5) is 3.2 in the Spanish corpus and 2.6 in the Italian corpus. This difference between both corpora is not significant [t = 1.430, df = 175, p = 0.155]. By contrast, the spread of postverbal Ss is different from preverbal Ss in both corpora. In particular, the median for postverbal Ss is 6 in the Spanish corpus and 7 in the Italian corpus, and the means are 7.5 in the Spanish corpus and 8.4 in the Italian corpus. The difference in means between both groups is not significant [t = -0.554, df = 65, p = 0.581]. To summarise, the results on the numeric scale reveal that for both learner groups preverbal Ss are short (around 2~3 words long), while postverbal Ss are long (around 7~8 words long). Consider now the same results in terms of the nominal (heavy/light) scale (Figure 6). The majority of unaccusative postverbal Ss are heavy: 78.8% in the Spanish corpus vs. 86.7% in the Italian corpus. The difference between the two corpora is non-significant [Ȥ2 = 0.455, df = 1, p = 0.50]. By contrast, as Figure 7 reveals, the majority of unaccusative preverbal Ss are light: 67.9% in the Spanish corpus vs. 78.1% in the Italian corpus. The difference is non-significant [Ȥ2 = 2.355, df = 1, p = 0.125]. To summarise, for both groups postverbal Ss tend to be heavy, while preverbal Ss tend to be light. Since the results of the nominal scale confirm the results of the numeric scale, these results support hypothesis 2.

Postverbal subjects at the interfaces in Spanish and Italian learners Spanish

111

Italian 13.3%

21.2%

Weight_nom Heavy Weight_nom Light

78.8%

86.7%

Figure 6. Heavy vs. light postverbal subjects (unaccusative VS) Spanish

Italian 21.9%

32.1%

Weight_nom Heavy Weight_nom Light

67.9% 78.1%

Figure 7. Heavy vs. light preverbal subjects (unaccusative SV) 6.3

H3: Postverbal Ss and information status

The last property of unaccusative Ss under investigation is information status.17 Figure 8 clearly shows that postverbal Ss are focus (evoked or inferable from the co(n)text): 98.1% in the Spanish corpus and 100% in the Italian corpus, the difference between corpora being non-significant [Ȥ2 = 0.293, df = 1, p = 0.588].18 Spanish 1.9%

Italian 0.0%

VS Top VS Foc

98.1%

100.0%

Figure 8. Topic vs. focus postverbal subjects (unaccusative VS)

112

Cristóbal Lozano and Amaya Mendikoetxea

The contexts in (45) illustrate this finding: postverbal subjects are brandnew information (focus), i.e. the subject has not been mentioned previously in the discourse. These examples are taken directly from the first paragraph of the ICLE essay to illustrate the point that the subject is a new entity in the discourse. (45)

a. In the world, dominated by science, technology and industrialisation, there is no a place for dreaming and imagination. Thanks to science and its consecuences, technology and insdustrialisation, appeared the big factories and the capitalism system. (spm03007) b. It seems impossible, but although we have now reached through technology a high standard of life, we are very pessimists. It seems as progress has stolen our imagination and therefore the love for small things. I can give few examples that such a fact: television is becoming lately the killer of conversation between parents and children; it is almost disappearing the use of writing nice letters to friends, since there is the telephone. (itrs1018)

There exist other contexts where the subject is new-anchored (to use Kaltenböck’s [2005] terminology), i.e. while part of the subject has been mentioned in the preceding context, typically a postmodifier PP, e.g. theatre in (46a) and men and women in (46b), the keyword in the NP is brand-new information, e.g. the decline and a predominance respectively. These have also been coded as focus, as has been done previously in the literature (see Kaltenböck 2004, 2005 and references therein). (46)

a. In the 2nd half of the 17th century we’ve got the Racionalism and so the comedy was the best mean to express its ideas. The most important for the restoration theatre was to know the human; so in its comedies appears the satire to know human behaviours. The theatre was highly professional because it was in the hands of actors, and so what they wanted was to make money the quickliest possible, so they represented what was more acceptable, so came the decline of the theatre because they repeated a lot of time the same themes. (spm06010) b. Nowadays, the difference between men and women should not exist any more, at least in our mind. It is true that, from the physical point of view, excluding particular situations (for example women who practice sports or do works which implies physical strength), there still remains a predominance of men over women. (itto4006)

Consider now the information status of preverbal material (XP) with unaccusative VS structures, i.e. unaccusative structures like XP-V-NPSUBJECT. Figure 9 shows that preverbal material is typically topic in both the Spanish and Italian corpora: 80% and 90% respectively. This difference is not significant between groups [Ȥ2 = 0.480, df = 1, p = 0.488]. The examples in (47) show preverbal XPTOPIC while those in (48) represent preverbal XPFOCUS.

Postverbal subjects at the interfaces in Spanish and Italian learners Spanish

113

Italian 10.0%

20.0%

XP Top XP Foc

80.0%

90.0%

Figure 9. Topic vs. focus preverbal material (unaccusative XP-V-S) (47)

(48)

a. The Victorian society drama is one of the 3 manifestations that emerged in the XVIII century. This kind of theatre represents the victorian values, which is the conservatism. It was developed in a drawing-room or an isolated set. That is why it is called the “drawing roam” drama. Against this society drama emerged an oposition headed by Oscar Wilde and Bernard Shaw. (spm08007) b. In the passage there are many points in which the so called [deleted word here] emerges, and this is conveyed by the many characters’ gestures. We could call it the body language and through it emerges the protagonists’ personality. (itrs1064) a. I think we have been created for to think and also for to dream and to imagine. Sometimes, when you have got a bad moment in your live, when you become depressed, to dream that maybe things will be better soon, to imagine that it can exist a better world can help to keep you alive in those hard days that everybody use to have. If some day we are not going to need dreaming nor imagining because everything will be done, then we won’t need either to be sad or hungry or to have problems because science will make something for not to exist problems, for our mind not to work about nothing. If this became so (and I say this arriving to a very extremed and exaggerated point of view) might someday babies will born without need to use their minds which means that in the evolution of the human specie it would disappear the capacity of thought in a near future. (spm04006) b. I think that the fashion of being vegetarian is lunked to this general going towards something more spiritual in an era in which we have got everything. Even if it could seem add after this introduction I gave up eating meat at about ) years ago. The reason of my choice concerned especially my personal tastes (I have never liked meat very much) but something strange that I can’t exactly explain happened to me one day 9 years ago. I refused to eat meat because it remind me about the animal alive. […] In my opinion a lot of vegetarian people have made this choice first to follow the fashion and not because of their own belief. In the U.S.A. for example, the country of contradictions, where there is the

Cristóbal Lozano and Amaya Mendikoetxea

114

highest number of obese persons but where the fitness mania involves every level of the population has recently spread out the fashion fo[r] natural and healthy food that of course doesn’t include meat. (itb13001) Contrary to what happens with postverbal Ss (cf. Figure 8), Figure 10 shows that preverbal unaccusative Ss are typically topic: 88.9% in the Spanish corpus and 90.6% in the Italian corpus. The difference is non-significant [Ȥ2 = 0.145, df = 1, p = 0.703]. The examples in (49) illustrate preverbal subjects which are topics (shown in italics), since they have previously been mentioned in the discourse (shown in underlined typeface). Note that personal pronouns are also topics, since by necessity their referent must have been mentioned in the prior discourse, cf. (50). Spanish

Italian 9.4%

11.1%

SV Top SV Foc

88.9%

90.6%

Figure 10. Topic vs. focus preverbal subjects (unaccusative SV) (49)

(50)

a. The approval of acting of women were something essential. Women started to perform female characters and this contribute to give a sexual and realistic atmosphere. […] Female characters appear with a stronger personality they really love these men. (spm08014) b. The idea of Europe doesn’t ignore these differences, but inglobes them, accept them and upon them construct its identity. […] If I think of the concept of Europe I cannot think of anything else that of a whole of different countries, but that all together produce the European identity. The differences have always existed in the Europe and for ages its peoples fought one against the other. (itrs1008) a. Also the records, movies have more success if they appear several days on the little screen. (spm07018) b. Violence does exist and it will always exist. (itto2034)

To summarise, in unaccusative VS structures, the postverbal S is typically focus and the preverbal material (in those cases where it appears) is topic. By contrast, in unaccusative SV structures, the preverbal S is typically topic. This confirms hypothesis 3.

Postverbal subjects at the interfaces in Spanish and Italian learners 7.

115

Discussion

As H1 predicted, both groups produce VS with unaccusatives only, yet their rates (8.1% in the Spanish corpus vs. 2.6% in the Italian corpus) differ statistically. Recall that this quantitative difference is unexpected given the similarities between Spanish and Italian (see Section 2). Additionally, Oshita (2004) did not find remarkable differences either (Spanish 6% vs. Italian 4%). A possible explanation for this may lie in the ratios of inversion with certain unaccusative Vs. As Figure 4 revealed, the Spanish results are somewhat inflated by the excessively high ratio of inversion with two of the unaccusatives, viz. exist (3.4%) and appear (1.7%). Both account for the majority of inversions (5.1% out of the total 8.1%). Additionally, recall that Palacios-Martínez & Martínez-Insua (2006) found that exist is the only unaccusative V in the Spanish ICLE corpus that appears with postverbal Ss in existential there-constructions. By contrast, the inversion rates with those two Vs in the Italian group are unexpectedly lower than in the Spanish group: 0.7% with exist and 0% with appear. A possible explanation for these discrepancies may come from the number of frequencies produced. We observed that in the Spanish group the raw frequency of concordances produced with those two Vs was very high: 58 concordances for exist (i.e. 9.1% of all concordances) and 60 concordances for appear (i.e. 9.4%). By contrast, the Italian counterparts are less frequent: 33 concordances for exist (i.e. 5.7%) and 13 concordances for appear (i.e. 2.3%). Follow-up analyses revealed that, for all verbs, there is a significant correlation between the frequency of concordances and the frequency of inversions in both corpora, Spanish (r = 0.397, p = 0.027) and Italian (r = 0.364, p = 0.044). This implies that the higher the number of concordances produced with unaccusatives, the higher the probability of inversion. So, it could be assumed that the unusually high proportion of inversions in the Spanish group with exist and appear is a consequence of their unusually high proportion of concordances. It could then be asked why there is such a high number of concordances for exist and appear in the Spanish corpus. Factors like composition topic and/or proficiency level could be the source of such differences. Regarding proficiency, it may be the case that the Italian group’s proficiency is higher than the Spanish group’s. Kaszubski (2001) notes that the ICLE Spanish subcorpus belongs to the upper-intermediate level, as opposed to the advanced level of the rest of the subcorpora. Additionally, the frequencies of grammatical vs. ungrammatical unaccusative VS produced suggest that this is the case (see Figure 2): the Italian group produces more grammatical structures (46.7%) than the Spanish group (34.6%). We should take this assumption provisionally, since, crucially, the ICLE does not provide an independent measure of proficiency for each participant. To summarise H1 (lexicon-syntax interface), the results support the hypothesis since both Spanish and Italian learners of English produced postverbal Ss with unaccusatives only, never with unergatives. This finding is in line with previous research (Rutherford 1989, Zobl 1989, Oshita 2004).

116

Cristóbal Lozano and Amaya Mendikoetxea

Regarding H2 (syntax-phonology interface), the results confirm the hypothesis that both Spanish and Italian learners of English tend to produce heavy Ss in postverbal position, but light Ss in preverbal position. To our knowledge, this finding has not been previously reported in the L2 literature. Consider now H3 (syntax-discourse interface). The results confirm the hypothesis that both Spanish and Italian learners of English consistently produce focus Ss in postverbal position, but topic Ss in preverbal position. To our knowledge, this finding has not been previously reported in the L2 literature either. To summarise all the results with unaccusative VS structures, the postverbal S is focus and it tends to be heavy, while preverbal material (in those cases where it appears) is topic. By contrast, with unaccusative SV structures, the preverbal S is topic and tends to be light. This is schematised in Table 5. Table 5. Summary of weight and information status with unaccusatives Construction Unaccusative postverbal S

Structure

Unaccusative preverbal S

NPSUBJECT Top Light

§ XP · ¨¨ ¸¸ © Top ¹

V Unac

NPSUBJECT Foc Heavy

V Unac

In conclusion, for learners of L2 English, unaccusativity is a necessary but not sufficient condition for Ss to be produced in postverbal position. In particular, other conditions must be met for S to appear postverbally. These are (in order of strictness): unaccusative verb, focus S, heavy S, cf. (51). (51)

Conditions on the production of postverbal Ss: (i) Unaccusative V (ii) Focus S (iii) Heavy S

8.

Conclusion

Inverted Ss are considered to be a feature of the interlanguage of L1 Spanish and Italian learners of L2 English. The Unaccusative Hypothesis (Perlmutter 1978) has been used to account for why inverted Ss appear to be produced only with unaccusative Vs. In recent years, the issue of unaccusativity has become part of a research agenda which seeks to establish a relation between the lexical semantics of Vs and the syntactic properties of the constructions they enter, in accordance with Perlmutter’s (1978) original intuition that unaccusativity is encoded in the syntax but is semantically determined. Within this framework, Levin &

Postverbal subjects at the interfaces in Spanish and Italian learners

117

Rappapport Hovav (1995) distinguish two major classes of unaccusative Vs: (i) Vs of change of state and (ii) Vs of existence and appearance. Only the latter appear in constructions with inverted Ss, such as locative inversion structures and there-constructions in native English. Our results from L1 Spanish/Italian learners confirm the possibility of inversion with unaccusative Vs of existence and appearance. While these results confirm that learners are sensitive to the unaccusative-unergative distinction, and thus provide support for the psychological reality of the Unaccusative Hypothesis, further research should show that unaccusative Vs of change of state do not trigger inversion, as in L1 English. Thus, learner data could be used to provide further evidence for the existence of these two verb classes, following the tradition of research which takes learner data “to make a unique and potentially significant contribution to theoretical linguistics” (Oshita 2004: 95-96) and the investigation of interlanguage “a valuable endeavour for any researcher interested in extending our knowledge about what language is” (Rutherford 1989: 164) (see Rutherford 1986 and Oshita 2004: Section 1 on the contribution of language acquisition studies to theoretical linguistics). The same can be said about the other two conditions identified for subject inversion (heaviness and discourse status), which support findings in the native English literature that inversion of S occurs, given the appropriate structural conditions, when S is syntactically heavy, as well as new (or newer) information or focus. Interestingly, there is an interrelation between these two factors (the newer an entity is, the more words are used to refer to it). Both conditions are designed to ease the processing burden on the receiver, by placing new and long constituents towards the end of a clause. The gradience approach adopted in some L1 studies to information status and heaviness turned out to be the most appropriate to code and explain our results concerning these two conditions, which, to our knowledge, have never been discussed in L2 studies of inversion. These studies have mostly focused on errors, while we have adopted a wider approach seeking to identify the conditions under which learners produce inverted Ss, regardless of problems to do with syntactic encoding. No significant differences were found between Italian and Spanish learners regarding the conditions under which inversion occurs, though, as was pointed out, Italian speakers produced significantly fewer inversions with unaccusative Vs. To provide a proper explanation of this fact, we would require precise information about the Italian and Spanish learners’ proficiency level. An issue that has not been mentioned but which could also prove extremely revealing would be the use of native Spanish and Italian corpora to determine to what extent the differences observed between the two groups of learners could be attributed to L1 properties (and thus to transfer). In fact, though we have been assuming throughout, on the basis of analyses provided in the theoretical literature, that Spanish and Italian are identical regarding the phenomenon known as “free inversion” (but see note 3), corpus studies may well reveal differences between the two languages. We are not aware of the existence of any such studies in either Italian or Spanish.

118

Cristóbal Lozano and Amaya Mendikoetxea

Similarly, a comparison between our L2 corpora and L1 English corpora would be crucial to identify overuse or underuse of inversion constructions, as part of the CIA approach mentioned in the introduction to this paper. Intuitively, it seems that at least Spanish learners appear to overuse postverbal subject structures. This intuition is confirmed by a preliminary comparison of the results obtained from the Spanish ICLE subcorpus with LOCNESS (Louvain Corpus of Native English Essays, CECL, Université catholique de Louvain, see http://cecl.fltr.ucl.ac.be), which does indeed reveal that this is the case. We leave a detailed analysis of these facts for further research (see Lozano & Mendikoetxea in preparation). Also for further research is the comparison of our results with Italian and Spanish NS data, as mentioned above, which would be crucial to make meaningful statements regarding transfer, as well as the analysis of L2 data from speakers of languages which do not allow free inversion. A preliminary analysis of the French ICLE subcorpus reveals a total of 2.3% of inversions with unaccusative verbs. This figure is similar to the rate of postverbal subjects found in the Italian subcorpus (2.6%) (see Figure 1 above). However, the inversion structures in the French subcorpus are mostly grammatical, as opposed to those found in the Italian and the Spanish subcorpora (see Figure 2 above). These facts require an in-depth examination. It is also our intention to broaden the empirical scope of this research into the production of inverted Ss to include the V be, which, according to Biber et al.’s (1999: 954) native English corpus findings appears in 95% of thereconstructions and about half of all locative inversion structures. Of the 1,778 tokens of the study reported in Birner (1994), 654 were instances of inversion around be, making it the most common V in inversion constructions. We also intend to combine corpus data with experimental data (e.g. acceptability judgements) in order to get a bigger picture of the mental processes and structures which underlie learners’ production of inversion structures (see e.g. Kennedy 1998 on the need to combine corpus data with other types of linguistic evidence). In sum, though the results obtained in this study are highly significant in that they reveal conditions for the occurrence of postverbal Ss in L2 English that have never been uncovered before, extending the scope of the research by comparing our results with native corpora and other learner corpora (different L2) and broadening the empirical scope by looking at inversion with the V be, as well as by using experimental language data, will no doubt allow us to gain a better understanding of the processes underlying this phenomenon in both native and non-native grammars. Notes 1

The research presented here has been supported by grants from the Spanish Ministry of Education (HUM2005-01278/FILO) and (jointly) from the Comunidad de Madrid and Universidad Autónoma de Madrid (09/SHD/016). Previous versions of this paper have been presented at the

Postverbal subjects at the interfaces in Spanish and Italian learners

119

4th International Contrastive Linguistics Conference in Santiago de Compostela and at TALC 7 in Paris, as well as in the Corpus Research Group at the University of Lancaster. We thank the audiences at all these events for their comments. We also thank all members of the WOSLAC project (see http://www.uam.es/woslac), with whom we have discussed many of the issues here. We also wish to thank the editors of this volume for their detailed comments on the first draft of this paper. All remaining errors are, of course, attributed to the authors themselves. 2

The term “postverbal” as used in this paper is purely descriptive: an S that occurs after V. Equally, “preverbal” refers to an S which occurs before V. We do not address here the question of what are the specific structural positions occupied by both the postverbal and the preverbal S, an issue which has recently received a lot of attention in the theoretical literature (see e.g. Hulk & Pollock 2001 for Romance subject inversion).

3

There are well-known differences between Spanish and Italian concerning VSO order: when S is an NP, as opposed to the PP in (2c), VSO is considered to be less acceptable in Italian. Italian does not allow VOS either. These restrictions are not found in Spanish, which seems to have a “freer” word order (see Belletti 2004b). Since we will be dealing with intransitive sentences like those in (4), in which Spanish and Italian behave alike, the differences observed in transitive sentences are not relevant here.

4

Among the Romance languages, French has a special status in that it is negatively marked for the NSP: it does not allow for NSs, nor for the type of “free inversion” found in other Romance languages. Thus, the French counterpart of the examples in (3b, c) and (4) is ungrammatical.

5

Languages like Chinese, Korean and Japanese may also have null Ss, but do not display “rich” agreement and appear to lack pro (see Huang 1984). For an account of these languages along different lines, within the context of L2 acquisition, see Oshita (2004).

6

Within the Minimalist Program, Chomsky (1995: Ch. 4) has attributed the presence of there in constructions like (8c) (and, presumably, proexpl) with the satisfaction of the Extended Projection Principle (EPP), by which clauses must have Ss (see Chomsky 1981). There are, however, proposals which seek to eliminate null expletives from the theory (for instance Picallo 1998 and Yusa 2002).

7

We are actually adopting a simplified version of the VP-internal subject hypothesis. In recent years more sophisticated proposals regarding the internal structure of the VP have been put forward in the literature (Chomsky 1995: Ch. 4 and references cited therein).

120

Cristóbal Lozano and Amaya Mendikoetxea

8

The inversion types in (15) are to be distinguished from subject-operator inversion, a construction found following negative opening elements (not, not only, hardly…), after so, etc.

9

There are, of course, additional requirements. For instance, for thereconstructions it has been claimed that the postverbal NP must be indefinite (see, for example, Safir 1985), although it is well known that this does not account for all the data and definite NPs can be found in these constructions (see e.g. Prince 1992: 299 and Biber et al. 1999: 967). Likewise, some indefinites are banned. In particular, as noted by Prince (1992: 299), plural generics are incompatible with there-constructions. Though Prince (1992) does not give reasons for this, the explanation has to be found in the semantics of the construction which requires indefinites to have existential meanings and not universal or generic meanings.

10

As pointed out by Belletti (2004b, Section 5), with the appropriate intonation pattern and the right pragmatic conditions, a postverbal S can also be interpreted as topic. This is illustrated in (i) for Italian (Belletti 2004b: 22) and the same can be said of parallel examples in Spanish. (i)

a. Che cosa ha poi fatto Giovanni? What has Gianni finally done? b. Ha parlato, Gianni. Has spoken, Gianni.

Examples like these are largely irrelevant for our purposes, since we are dealing with written corpora, in which examples like these are unlikely to occur. The other side of the coin concerns preverbal elements which are focus. These often receive a contrastive interpretation, which is marked by special intonation (see Rizzi 1997, 2004 for Italian and Domínguez 2004 for Spanish) (see also the examples in [21]): (ii)

a. GIANNI ha comprato il libro (non Maria). b. JUAN ha comprado el libro (no María). c. JOHN has bought the book (not Mary).

11

This is the approach taken in Chocano et al. (in preparation), in which the binary topic-focus distinction comprises both a retrievability scale, as well as a recency scale.

12

One attempt to introduce information structure notions such as topic and focus as part of syntactic structure is what is known as the “cartographic approach” to syntactic structure (see e.g. Rizzi 1997 and the papers contained in Belletti 2004a). Focalisation (and topicalisation) are analysed as involving movement of a phrase to a designated position either in the left periphery or in the right periphery. The interpretation of an element as focus or topic is derived automatically from the position occupied by that

Postverbal subjects at the interfaces in Spanish and Italian learners

121

element in clause structure: the interpretation is read off the structural configuration. 13

The term “Heavy NP Shift” is used here descriptively. We are not postulating a rightward movement rule of the type found in the transformational analyses of the 70s.

14

Though all the examples in this section involve instances of locative inversion structures, a parallel situation is observed for there-constructions which are often found with “heavy” postverbal elements.

15

As is commonly assumed in the Generative Grammar literature, we take theme to be the semantic role associated with the internal argument of a V. In other frameworks this is referred to as the patient.

16

It is important to note that we classify as “ungrammatical” only those postverbal Ss that are not possible in English. We are abstracting away from standard ungrammaticalities such as S-V agreement. For example, in the sentence Then come the necessity to earn more (spm07023), the postverbal-subject structure is possible in English, yet lack of S-V agreement renders it ungrammatical sensus stricto.

17

Apart from the weight and information status of S, we also analysed its definiteness as it has been traditionally claimed that postverbal focused Ss are indefinite, while preverbal topic Ss are definite. This has been recently confirmed in corpus studies (e.g. Prince 1992 and Biber et al. 1999; see also note 9 here). In order to test whether there was a relationship between the information status and definiteness of S, we coded every S as being definite or indefinite. With unaccusative VS structures, learners typically mark S as indefinite (59.6% Spanish vs. 53.3% Italian, non-significant [Ȥ2 = 0.189, df = 1, p = 0.664]), while in unaccusative SV structures, S is marked as definite (70.4% Spanish vs. 72.99% Italian, non-significant [Ȥ2 = 0.141, df = 1, p = 0.708]). In other words, our results follow the typical native English pattern, i.e. preverbal (topic) Ss tend to be definite, while postverbal (focus) Ss tend to be indefinite.

18

The 1.9% corresponds to only one case (out of 52) where the unaccusative postverbal S is a topic.

References Arnold, J.E., T. Wasow, A. Losongco and R. Ginstrom (2000), ‘Heaviness vs. newness: The effects of structural complexity and discourse status on constituent ordering’, Language, 76: 28-55. Belletti, A. (2001), ‘“Inversion” as focalization’, in: A.C. Hulk and J.-Y. Pollock (eds) Subject Inversion in Romance and the Theory of Universal Grammar. Oxford: Oxford University Press. 60-90.

122

Cristóbal Lozano and Amaya Mendikoetxea

Belletti, A. (2004a) (ed.) Structures and Beyond. The Cartography of Syntactic Structures. Vol 3. New York: Oxford University Press. Belletti, A. (2004b), ‘Aspects of the low IP area’, in: L. Rizzi (ed.) The Structure of CP and IP. The Cartography of Syntactic Structures. Vol 2. New York: Oxford University Press. 16-51. Biber, D., S. Johansson, G. Leech, S. Conrad and E. Finegan (1999), The Longman Grammar of Spoken and Written English. London: Longman. Birner, B. (1994), ‘Information status and English inversion’, Language, 70: 233259. Birner, B. (1995), ‘Pragmatic constraints on the verb in English inversion’, Lingua, 97: 223-256. Bolinger, D. (1977), Form and Meaning. London: Longman. Bresnan, J. (1994), ‘Locative inversion and the architecture of Universal Grammar’, Language, 70: 72-131. Burzio, L. (1986), Italian Syntax. Dordrecht: Kluwer. Chocano, G., C. Lozano, R. Jiménez, A. Mendikoetxea and S. Murcia (in preparation), ‘Retrievability, recency and the Topic-Focus distinction: a study of subjects’ (ms.). Universidad Autónoma de Madrid. Chomsky, N. (1981), Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. (1995), The Minimalist Program. Cambridge, MA: MIT Press. Culicover, P.W. and R.D. Levine (2001), ‘Stylistic inversion in English: a reconsideration’, Natural Language and Linguistics Theory, 19(2): 283310. Domínguez, L. (2004), Mapping focus: The syntax and prosody of focus in Spanish. Unpublished PhD dissertation, Boston University. Eguren, L. and O. Fernández Soriano (2004), Introducción a una sintaxis minimista. Madrid: Gredos. Fernández-Soriano, O. (1989), ‘Strong pronouns in null-subject languages and the Avoid Pronoun Principle’, MIT Working Papers in Linguistics, 11: 228-239. Fernández-Soriano, O. (1993), ‘Sobre el orden de palabras en español’, Cuadernos de Filología Hispánica, 11: 113-151. Firbas, J. (1992), Functional Sentence Perspective in Written and Spoken Communication. Cambridge: Cambridge University Press. Frazier, L. (2004), ‘(Default) Focus structure in sentence processing’. Paper presented at the workshop Information structure in language processing and language acquisition, University of Potsdam, October 2004. Geluykens, R. (1991), ‘Information flow in English conversation: a new approach to the given-new distinction’, in: E. Ventola (ed.) Functional and Systemic Linguistics: Approaches and Uses. Berlin: Mouton de Gruyter. 141-167. Gilquin, G. (2000/2001), ‘The Integrated Contrastive Model. Spicing up your data’, Languages in Contrast, 3(1): 95-123. Granger, S. (1996), ‘From CA to CIA and back: an integrated approach to computerized bilingual and learner corpora’, in: K. Aijmer, B. Altenberg and M. Johansson (eds), Languages in Contrast. Papers from a

Postverbal subjects at the interfaces in Spanish and Italian learners

123

Symposium on Text-Based Cross-linguistic Studies. Lund 4-5 March 1994. Lund: Lund University Press. 37-51. Granger, S. (2002), ‘A bird’s-eye view of computer learner corpus research’, in: S. Granger, J. Hung and S. Petch-Tyson (eds) Computer Learner Corpora, Second Language Acquisition and Foreign Language Teaching. Amsterdam & Philadelphia: John Benjamins Publishing Company. 3-33. Granger S., E. Dagneaux and F. Meunier (2002), The International Corpus of Learner English. Handbook and CD-ROM. Louvain-la-Neuve: Presses Universitaires de Louvain. Hawkins, J. (1994), A Performance Theory of Order and Constituency. Cambridge: Cambridge University Press. Hertel, T.J. (2003), ‘Lexical and discourse factors in the second language acquisition of Spanish word order’, Second Language Research, 19: 273304. Huang, C.-T.J. (1984), ‘On the distribution and reference of empty pronouns’, Linguistic Inquiry, 15: 531-574. Hulk, A.C. and J.-Y. Pollock (eds) (2001), Subject Inversion in Romance and the Theory of Universal Grammar. Oxford: Oxford University Press. Jaeggli, O. and K. Safir (eds) (1989), The Null Subject Parameter. Dordrecht: Kluwer. Kaltenböck, G. (2004), It-Extraposition and Non-Extraposition in English: A Study of Syntax in Spoken and Written Texts. Wien: Braumüller. Kaltenböck, G. (2005), ‘It-extraposition in English: A functional view’, International Journal of Corpus Linguistics, 10(2): 119-159. Kaszubski, P. (2001), ‘Tracing idiomaticity in learner language: the case of BE’, in: P. Rayson, A. Wilson, T. McEnery, A. Hardie and S. Khoja (eds) Proceedings of the Corpus Linguistics 2001 Conference. Lancaster: University Centre for Computer Corpus Research on Language. 312-322. Kennedy, G. (1998), An Introduction to Corpus Linguistics. London & New York: Longman. Koopman, H. and D. Sportiche (1991), ‘The position of subjects’, Lingua, 85: 211-258. Levin, B. (1993), English Verb Classes and Alternations: A Preliminary Investigation. Chicago: University of Chicago Press. Levin, B. and M. Rappaport Hovav (1995), Unaccusativity at the Lexical Semantics-Syntax Interface. Cambridge, MA: MIT Press. Liceras, J., B. Soloaga and A. Carballo (1994), ‘Los conceptos de tema y rema: problemas sintácticos y estilísticos de la adquisición del español’, Hispanic Linguistics, 5: 43-88. Lozano, C. (2003), Universal Grammar and focus constraints: The acquisition of pronouns and word order in non-native Spanish. Unpublished PhD dissertation, University of Essex. Lozano, C. (2006), ‘Focus and split intransitivity: The acquisition of word order alternations in non-native Spanish’, Second Language Research, 22: 1-43.

124

Cristóbal Lozano and Amaya Mendikoetxea

Lozano, C. and A. Mendikoetxea (in preparation), ‘Interface conditions on postverbal subjects: a corpus study of “inversion” in non-native grammars’ (ms.). Universidad de Granada/Universidad Autónoma de Madrid. Luján, M. (1999), ‘Expresión y omisión del pronombre personal’, in: I. Bosque and V. Demonte (eds) Gramática descriptiva de la lengua española. Madrid: Espasa-Calpe. 1275-1315. Oshita, H. (2000), ‘What is happened may not be what appears to be happening: a corpus study of “passive” unaccusatives in L2 English’, Second Language Research, 16: 293-324. Oshita, H. (2002), ‘Uneasiness with the easiest: on the subject-verb order in L2 English’, Second Language, 1: 45-61. Oshita, H. (2004), ‘Is there anything there when there is not there? Null expletives in second language data’, Second Language Research, 20: 95130. Palacios-Martínez, I. and A. Martínez-Insua (2006), ‘Connecting linguistic description and language teaching: native and learner use of existential there’, International Journal of Applied Linguistics, 16(2): 213-231. Perlmutter, D. (1978), ‘Impersonal passives and the unaccusative hypothesis’, Papers from the Annual Meeting of the Berkeley Linguistics Society, 4: 157-189. Picallo, C. (1998), ‘On the Extended Projection Principle and null expletive subjects’, Probus, 10: 219-241. Pinto, M. (1997), Licensing and Interpretation of Inverted Subjects in Italian. Utrecht: UiL OTS Dissertation Series. Prince, E.F. (1981), ‘Toward a Taxonomy of Given-New Information’, in: P. Cole (ed.) Radical Pragmatics. London: Academic Press. 223-255. Prince, E.F. (1992), ‘The ZPG letter: Subjects, definiteness and information status’, in: S. Thompson and W. Mann (eds) Discourse Description: Diverse Analyses of a Fund Raising Text. Amsterdam: John Benjamins Publishing Company. 295-325. Quirk, R., S. Greenbaum, G. Leech and J. Svartvik (1972), A Comprehensive Grammar of the English Language. London: Longman. Rizzi L. (1982), Issues in Italian Syntax. Dordrecht: Foris. Rizzi, L. (1997), ‘The fine structure of the left periphery’, in: L. Haegeman (ed.) Elements of Grammar. Handbook of Generative Syntax. Vol 1. Dordrecht: Kluwer Academic Publishers. 281-337. Rizzi, L. (2004), ‘Locality and left periphery’, in: A. Belletti (ed.) Structures and Beyond. The Cartography of Syntactic Structures. Vol 3. New York: Oxford University Press. 223-251. Rochemont, M.S. (1986), Focus in Generative Grammar. Amsterdam: John Benjamins Publishing Company. Rutherford, W. (1986), ‘Grammatical Theory and L2 Acquisition: a brief overview’, Second Language Research, 2: 1-15.

Postverbal subjects at the interfaces in Spanish and Italian learners

125

Rutherford, W. (1989), ‘Interlanguage and pragmatic word order’, in: S. Gass and J. Schachter (eds) Linguistic Perspectives on Second Language Acquisition. Cambridge: Cambridge University Press. 163-182. Safir, K. (1985), Syntactic Chains. Cambridge: Cambridge University Press. Scott, M. (2002), Oxford WordSmith Tools (version 4.0). Oxford: Oxford University Press. Sorace, A. (1993), ‘Incomplete vs. divergent representations of unaccusativity in nonnative grammars of Italian’, Second Language Research, 9: 22-47. Sorace, A. (1995), ‘Acquiring linking rules and argument structures in a second language’, in: L. Eubank, L. Selinker and M. Sharwood-Smith (eds) The current State of Interlanguage: Studies in Honor of William E. Rutherford. Amsterdam: John Benjamins Publishing Company. 153-175. Vallduví, E. (1990), The informational component. Unpublished PhD dissertation, University of Pennsylvania. Wasow, T. (1997), ‘End-weight from the speaker’s perspective’, Journal of Psycholinguistic Research, 26: 347-361. White, L. (1985), ‘The “pro-drop” parameter in adult second language acquisition’, Language Learning, 35: 47-62. White, L. (1986), ‘Implications of parametric variation for adult second language acquisition: an investigation of the Pro-drop parameter’, in: V.J. Cook (ed.) Experimental Approaches to Second Language Acquisition. Oxford: Pergamon Press. 55-72. Yusa, N. (2002), ‘The implications of unaccusative errors in L2 acquisition’, in: J. Costa and M.J. Freitas (eds) Proceedings of the GALA 2001 Conference on Language Acquisition. Lisbon: Associaçao Portuguesa de Linguistica. 289-296. Zagona, K. (2002), The Syntax of Spanish. Cambridge: Cambridge University Press. Zobl, H. (1989), ‘Canonical typological structures and ergativity in English L2 acquisition’, in: S. Gass and J. Schachter (eds) Linguistic Perspectives on Second Language Acquisition. Cambridge: Cambridge University Press. 203-221. Zubizarreta, M.L. (1998), Prosody, Focus and Word Order. Cambridge, MA: MIT Press. Zubizarreta, M.L. (1999), ‘Las funciones informativas: tema y foco’, in: I. Bosque and V. Demonte (eds) Gramática descriptiva de la lengua española (vol. 3). Madrid: Espasa. 4215-4244.

Adverb placement in post-intermediate learner English: a contrastive study of learner corpora John Osborne Université de Savoie Abstract This study compares the placement of adverbs in the written productions of postintermediate learners of English, from a variety of L1 backgrounds, and in those of native speakers. Particular interest is given to adverbs placed in Verb-Adverb-Object order, one which is normally not allowed by the grammar of Present-Day English, but which continues to appear in learner English at a post-intermediate stage, alongside other, more conventional, adverb positions. Learners whose L1 has obligatory verb-raising – Spanish, Italian and French – show the strongest tendency to use V-Adv-O order, but it is also observed in the productions of other learners, including those whose L1 does not normally admit adverbs in this position. Under certain conditions, notably where there is a heavy object NP and/or strong collocational tie between verb and adverb, V-Adv-O sequences also appear in nativespeaker English. It is suggested that learners over-extend these patterns to include “lighter” NPs and “weaker” V-Adv collocations, and that non-canonical adverb placement is motivated by semantic and phraseological factors.

1.

The problem And I just want to tell them also, because you want absolutely me to say a few words in English, I want to tell them also that France and I have always been friends of the United States. Jacques Chirac, interview with CBS & CNN, Paris, 16.3.2003

French-speaking learners of English have difficulty attaining target-like placement of adverbs. This difficulty is commonly attributed to a difference between English and French (amongst other languages) with respect to verbraising. In French and other verb-raising languages, verbs are obligatorily raised past negation and past the adverb, such that (1a) and (2a) below are well-formed in French, while (1b) and (2b), where the verb remains to the right of the negation or of the adverb, are ungrammatical. In Present-Day English, main verbs do not raise, giving the opposite situation from French: (3a) and (4a), with a raised verb, are ungrammatical, while (3b) and (4b) are well-formed. (1a) (2a)

Jean (n’)aime pas la glace Jean mange souvent de la glace

(1b) *Jean ne pas aime la glace (2b) *Jean souvent mange de la glace

128 (3a) (4a)

John Osborne *John likes not ice-cream *John eats often ice-cream

(3b) John doesn’t like ice-cream (4b) John often eats ice-cream

This question has attracted considerable attention from researchers in Second Language Acquisition, beginning with the studies reported in White (1991) and Trahey & White (1993). The aim of such studies has usually been to investigate hypotheses about the initial state and subsequent development of L2 grammars, principally from a Universal Grammar perspective. The main hypotheses, for cases where L1 is verb-raising and L2 non-raising, are summarised in Table 1 below. Table 1. Summary of SLA hypotheses Full Transfer / Full Access (Schwartz & Sprouse 1996)

The L1 grammar is the initial state, and is initially transferred to L2. This predicts raising for early learners; no raising for later learners, provided sufficient evidence is available for them to reset their grammar.

Minimal Trees (Vainikka & Young-Scholten 1996)

The initial state lacks functional categories altogether. This predicts no raising in the early stage of L2 acquisition, optional raising – by analogy with auxiliaries – for intermediate learners, and possibly subsequent retreat for later learners.

Valueless Features (Eubank 1996)

Initial state includes L1 categories, but certain features are unspecified – i.e. they have no “strength” value. This predicts optional raising for early learners; no raising for later learners once strength is fixed.

Local Impairment (Beck 1998)

Features that require or prohibit verb raising are impaired during maturation, resulting in durable optionality in verbraising.

Failed Functional Features (Hawkins & Chan 1997)

Only functional features instantiated in L1 are available beyond childhood to L2 learners. Learners will acquire categorical knowledge only of areas of L2 which share functional features with L1; all other knowledge will be probabilistic. Persistent errors in verb-raising should be linked to specific L1/L2 pairings.

Some of these hypotheses thus predict persistent optionality in L2 grammars (Beck as a general characteristic; Hawkins & Chan for specific language pairings). This prediction is of particular interest to the present study, since non target-like placement of adverbs continues to appear at a post-intermediate stage of learning, even in highly monitored production. Examples (5)-(11) below, all

Adverb placement in learner English

129

containing the same verb+adverb combination, see+clearly, illustrate the range of adverb positions attested in the learner corpora used for this study (details of the corpora are given in Section 2 below): (5) (6) (7) (8) (9) (10) (11)

Clearly it can be seen that money turns out to be the root of evil. (ICLEIT) [...] they clearly see the reasons for the men’s behaviour. (ICLE-GE) Here we see clearly the refusal of communication. (ICLE-FR) So we see clearly through these novels the break with the Victorians Novels. (ICLE-FR) In this example we see clearly the contrast between life and absence of life. (ICLE-FR) We see clearly in the novel that the hero’s mind works by means of associations (ICLE-FR) [...] so that the grown-ups could see it clearly. (ICLE-CZ)

Some of these placements are perfectly target-like; others might strike a native-speaker as strange. Given that variable placement of adverbs is also a feature of NS grammars (see below, Table 3 and the examples in Section 4), the question is in what way the optionality observed in learner grammars is different. Sorace (2003) suggests three main characteristics of optionality in L2: L2 grammars exhibit a greater tolerance for optionality than native grammars, L2 optionality tends to persist at advanced competence levels, and residual optionality is found at ultimate L2 attainment. This suggests that learner grammars should display a diminishing zone of optionality, which nevertheless remains broader than that of native speakers: In the typical L2 end state characterized by optionality, optional variants are not in free variation: a steady state is reached in which the target option is strongly but not categorically preferred and the nontarget option surfaces in some circumstances. (Sorace 2003: 139-140) The concern of the present study is to investigate what these circumstances might be, in the case of adverb placement, and to attempt to answer the following questions: x Does non native-like adverb placement continue to appear in the free production of post-intermediate learners? x Are results consistent (for the same L1) across different corpora? x Does adverb placement in L2 English vary according to L1? x If there is persistent difficulty in achieving native-like adverb placement, is this linked to specific L1-L2 pairings? x Do adverb-placement errors surface at random, or are there specifiable circumstances?

John Osborne

130

x Do the same factors trigger non native-like adverb placement irrespective of L1? 2.

Corpora and contrastive analysis

To answer these questions requires at least a two-way comparison, contrasting the productions of native-speakers of English with those of non-natives, and contrasting the productions of non-natives with different L1s with each other. In a developmental perspective, a third axis of comparison could be added, comparing productions of learners with the same L1, but at different levels of proficiency. This dimension has not been included here, and so the corpora used for the study include productions from users of approximately the same age, education and level of proficiency. Altogether, four corpora were used: two learner corpora and two comparable native-speaker corpora (see Table 2). Table 2. The corpora Corpus

Native/nonnative Chambéry Corpus Non-native ICLE Non-native Essay Bank Native LOCNESS Native

L1

N° of words

French 11 different L1s English (GB) English (GB + US)

1.1m 2.8m 165,000 95,000 + 168,000

The first learner corpus, the Chambéry Corpus, consists of approximately 300 argumentative and analytical essays from 2nd and 3rd year undergraduates in a French university, on subjects such as: “Smoking in America”, “ICT in education”, “Juvenile boot camps”, “Aboriginal Health and Social Problems”, “The Britons’ attitude towards foreign languages”. The essays were written with no constraints on time, and with free use of dictionaries or other reference books. They thus represent examples of highly monitored written production, where the learners have made every effort to eliminate or avoid possible errors. The essays have an average length of about 3,000 words, giving a total corpus of just over 1m words. The second learner corpus is the International Corpus of Learner English (Granger et al. 2002), containing 11 subcorpora, each 200-270,000 words, from speakers of 11 different L1s: French, Spanish, Italian, Russian, Polish, Czech, Bulgarian, Finnish, German, Dutch and Swedish. The learners represented in ICLE are also undergraduates, and their productions, although on average somewhat shorter than those in the Chambéry Corpus, were collected under similar conditions: argumentative essays, the majority of which were produced with no time limits and with access to reference books. The learner corpora thus have in common the same L2, English, but represent different L1s, of which three are verb-raising Romance languages (French, Spanish, Italian), four are non-raising Slavic languages (Russian, Polish, Czech, Bulgarian), three are V2 Germanic languages (German, Dutch, Swedish),

Adverb placement in learner English

131

and one (Finnish) is non Indo-European. All the learners represented in these corpora are at roughly the same stage in their academic study of English, but inevitably show some variation in their overall proficiency. Informal reading of their written productions suggests that most would fall within the Common European Framework definition of the “Proficient User” (Council of Europe 2001). For this reason, they are collectively described here, somewhat loosely, as post-intermediate learners, although the actual range between the weakest and the strongest is probably from B2 to C2. These learner corpora were compared with written productions from native speakers of English of a similar age and educational level, taken from the LOCNESS corpus (Louvain Corpus of Native English Essays), containing samples from both British and American students, and from an Essay Bank corpus compiled at the University of Savoie, containing 1st and 2nd year British undergraduate essays in three subject areas, literature, politics and sociology, and ranging in length from 2,000 to 3,000 words. All the corpora were POS tagged, using a Brill tagger, to enable recovery of adverbs. The batch search facility in MonoConc (Barlow 1999) was then used to extract occurrences of adverbs in sentence initial, sentence final, pre-verbal and V-Adv-O positions. In the case of V-Adv-O position, this necessitates searching for a number of different possible strings, since the tagging does not identify NPs as such, but only their separate components, and the word immediately following the adverb in a V-Adv-O sequence may be a noun, determiner, adjective or pronoun. It was also necessary to edit the results by hand in order to remove irrelevant occurrences, of which there were two particularly frequent types: cases where the adverb is more closely attached to the following NP (speak exactly the same language), and those where the element tagged as an adverb is in fact part of a multi-word verb (he put forward the idea). 3.

Patterns of adverb placement

This section is devoted essentially to describing patterns of adverb placement in learner English, and in particular the choice of V-Adv-O order. But before looking at adverb placement in more detail, there is another potential manifestation of verb-raising that should be mentioned: cases where the verb precedes a negation, giving a V-Neg-O sequence. 3.1

This needs not any proof: V-Neg-O

V-Neg-O sequences are not frequent in the learner corpora, except where the main verb is have, as in (12)-(14): (12)

Not only are these people desperate, but also some ordinary people have not the living wage. (ICLE-CZ)

132 (13)

(14)

John Osborne But I am afraid that now it is impossible to abrogate all clauses of Russian Criminal Code providing for the capital punishment because now we have not the stable economy. (ICLE-RU) However, there were some people aboard that had not the impression of spending memorable moments, quite the opposite. (Chambéry Corpus)

Have-Neg-O order is certainly possible in Present-Day English, but most uses tend to be of a formulaic nature (I haven’t the time/patience/nerve; I haven’t the faintest idea). Only two occurrences (equivalent to 0.85 per 1m words) were found in the NS corpora examined here, both of which are clearly formulaic (haven’t a care in the world, hadn’t the shadow of doubt). In learner English, have-Neg-O sequences are rather more frequent, possibly because of the dual status of have as an auxiliary and as a lexical verb. There are 114 occurrences in ICLE (i.e. approximately 40 per 1m words), and in both ICLE and the Chambéry Corpus such sequences represent just over 20% of negated have+Object phrases. With all other verbs, V-Neg-O order is marginal. Not surprisingly, there are no examples in the NS corpora, with the exception of a direct quote from an archaic religious text, and only six in the learner corpora (examples [15]-[20] below), some of which may in fact result from confusion with the negative determiner no. (15) (16) (17)

(18)

(19) (20)

3.2

Among many institutions of every state the prison system as an organ of punishment takes not the last place. (ICLE-RU) Money is the cornerstone of the modern society. This is a simple evidence that needs not any proof. (ICLE-SP) For it is true that here in Italy there exist not training programs where students may learn what are effectively the real needs of the labour market. (ICLE-IT) And by the way, she says contemptously and accusingly “Do you know that your daughter came not home until four o ’clock in the morning? (ICLE-GE) Alice Walker leaves not doubt that black people’s traditions are different form that of their fellow white citizens. (ICLE-GE) To conclude, crime seems to be state supported and justice takes not its course any longer. (ICLE-GE) Adverb placement

A necessary preliminary question for any corpus-based contrastive study is whether two corpora which are believed to be similar do in fact yield similar results for the phenomenon in question. Contrasting learner corpora (or components of a corpus) with a view to detecting possible L1 influences relies on the assumption that significant differences will not appear between comparable corpora of learners with the same L1. By the same token, comparison of two or more native speaker corpora would also be expected to yield similar results; if it

Adverb placement in learner English

133

does not, then the phenomenon may be subject to a variability that has nothing to do with the nature of learner interlanguage. Table 3 shows the preferred positions for adverb placement in the native speaker corpora studied here. Table 3. Adverb placement: L1 English (occurrences /100,000 words) LOCNESS US LOCNESS GB Essay Bank All NS

initial 200.3 139.0 117.0 154.6

pre-verbal 865.5 855.8 849.0 857.0

final 95.7 67.9 50.9 72.3

V-Adv-O 5.35 5.22 7.27 6.10

Total 1166.85 1067.92 1024.17 1090

The overall frequency of adverbs is similar in the three NS corpora, as is the number of adverbs used in pre-verbal position. The main divergences between the corpora are in the use of adverbs in initial and in final position, particularly between the American component of the LOCNESS corpus and the two British samples. These differences appear to be the result of stylistic or rhetorical preferences, possibly induced by instruction. The discrepancy in the number of initial adverbials, for example, is almost entirely due to the two connectors Also and Now, which together account for 65.5 occurrences per 100,000 words in the American sample, and only 12.6 and 11.5 occurrences respectively in the two British samples. For the adverb position that most closely interests us here, VAdv-O, the number of occurrences is consistently small, in all cases representing less than 1% of adverb uses. Table 4. Adverb placement: L1 French (occurrences /100,000 words) initial pre-verbal final Chambéry (French 1) 205.6 720.5 50.7 ICLE-FR (French 2) 169.4 864.3 70.9

V-Adv-O Total 25.4 1002.2 22.5 1127.1

The second set of comparable corpora is composed of the Chambéry Corpus and the French component of ICLE. Adverb positions for these corpora are shown in Table 4. The learners whose productions make up the two corpora share the same L1, French, but do not share the same educational and linguistic background, since the first have received schooling in France, and the second in Belgium. It is possible, therefore, that there will be stylistic and training-induced differences between these two groups, as between the native speakers, and that this may account for the slight variation in initial and final position. For the adverb placement that most closely concerns us here, choice of V-Adv-O rather than pre-verbal position, as a possible consequence of verb-raising, the differences between the Chambéry Corpus and ICLE-FR are not significant (Ȥ2 = 1.072, df = 1, p”1). Likewise, there is no significant difference for choice of VAdv-O vs. pre-verbal position among the three NS groups observed in Table 3 (Ȥ2 = 0.047, df = 2, p”1). Conversely, choice of these adverb placements does vary

John Osborne

134

significantly between the NS writers and both of the French-speaking groups: respectively, Ȥ2 = 15.18, df = 1, p<0.001 for NS students vs. Chambéry, and Ȥ2 = 9.116, df = 1, p<0.01 for NS students vs. ICLE-FR. When these comparisons are extended across all of the corpora, for both native and non-native speakers (Table 5), then two general tendencies appear. Firstly, the hierarchy of preferred positions is the same for all groups, native and non-native, with pre-verbal position being the most preferred, followed by initial, final and then V-Adv-O position, but the relative strength of preference varies from one group to another. Polish-speaking users of English, for example, appear more reluctant than others to place adverbs in initial or final position. This is in line with what would be expected in a case of optionality – that postintermediate learners show a pattern of usage which tends overall towards that of native speakers, but which diverges in varying degrees, for reasons which remain to be examined. Table 5. Adverb placement (occurrences /100,000 words) NS French 1 French 2 Italian Spanish Bulgarian Czech Polish Russian Dutch German Swedish Finnish

initial 154.6 205.6 169.4 199.9 172.2 239.2 214.2 74.9 233.9 218.8 202.8 223.6 229.8

pre-verbal 857.0 720.5 864.3 607.7 443.7 728.4 572.4 808.1 574.9 810.9 682.8 724.6 760.5

final 72.3 50.7 70.9 52.4 68.9 84.2 147.7 24.9 118.6 116.1 100.3 87.8 124.1

V-Adv-O 6.10 25.4 22.50 35.23 32.44 7.97 13.69 15.24 9.78 2.96 3.83 0.96 8.63

Total 1090 1002.2 1127.1 895.23 717.24 1059.77 947.99 923.14 937.18 1,148.76 989.73 1036.96 1123.03

Secondly, as regards the main focus of this study, V-Adv-O order, it is possible to distinguish three groups: i.

ii. iii.

a group where the occurrence of adverbs in V-Adv-O position is markedly more frequent than in the NS corpus (French, Italian, Spanish); a group where the occurrence of adverbs in this position is less frequent than in the NS corpus (Dutch, German, Swedish); an intermediate group where the frequency of V-Adv-O is comparable to, or slightly higher than that in the NS corpus (Finnish, Bulgarian, Russian, Czech and Polish).

Adverb placement in learner English

135

The higher frequency of V-Adv-O usage among speakers of verb-raising languages may well indicate that syntactic transfer is taking place. But the fact that speakers of non-raising languages also produce non target-like V-Adv-O sequences suggests that other influences could be at work too. And why do speakers of V2 languages tend to avoid such sequences, when they are allowed in their L1? The answer may lie partly in the conditions which trigger V-Adv-O placement, both as an infrequently used option in native-speaker English and in learner productions. 4.

V-Adv-O sequences in L1 English and in learner English

Both V-Adv-O and V-Neg-O sequences were common in English up to the 16th century, when with the development of do-support they began to decline. It is interesting to note that V-Adv-O sequences appearing in English texts between the early 16th and late 17th century often contain the same verb-adverb associations as those found in learner productions (verbs such as know, see, understand co-occurring with well, perfectly or clearly), thus displaying similarities between a transitional period in the evolution of the language and a transitional stage in interlanguage. A few examples are given in (21) to (25) below, all taken from the Helsinki corpus (Rissanen 1999). (21)

(22)

(23)

(24) (25)

bicause that without perspectiue knowledge, it is not easy to iudge truly the formes of them in flatte protacture. (Robert Record, The Path-Way To Knowledg, Containing The First Principles Of Geometrie, 1551) by reason wherof they saued all that longe tyme whiche at this dayes is spente in understandyng perfectly the greke or latyne. (Thomas Elyot, The Boke Named The Gouernour, 1531) and certeyne soldiers entered the ditch to veiwe exactly the state of the walles. (Sir John Hayward, Annals Of The First Four Years Of The Reign Of Queen Elizabeth, composed 1612) and I doe see plainely the exceeding benefits, that must needs come hereby (John Brinsley, Ludus Literarius Or The Grammar Schoole, 1627). He understood well the size of their understanding and their tempers. (Gilbert Burnet, History Of My Own Time, first published posthumously, 1723-4)

In Present-Day English, use of V-Adv-O placement is probably subject to a degree of idiolectal variation, but typically appears under two conditions, both of which will be discussed in this section. 4.1

NP shift

The first condition is illustrated, almost to extreme, in examples (26) and (27), taken from the NS corpora:

136 (26)

(27)

John Osborne After considering these articles that cover both sides of the issue, it is obvious that prayer does not belong in the public school classroom, as the articles that oppose prayer in public schools refute and weaken considerably the arguments for the reintroduction of prayer in public schools as a way to cure modern social ills. (LOCNESS US) The fluid construction of these genres between positive and negative, light and dark, the alluring and the repellent, the tantalising and the terrible, Eros and Thanatos, mirrors significantly the Romantic construction of homosexuality which is constituted interdependantly (sic) by the sensuality of homoeroticism and the terror of homophobia. (Essay Bank)

Each of these examples shows a postverbal sequence consisting of an adverb and a “heavy” (18 or 19-word) NP. Such heavy NP shift is frequently observed not only in postverbal sequences containing an adverbial, but also in those containing a prepositional phrase. The precise conditions of heavy NP shift are not well understood. It is not clear, for example, whether it is the length or the complexity of an NP that causes it to be shifted, or whether it is the relative length of the NP compared with other postverbal elements that is decisive (see for example Hawkins 1994), nor what role is played by other factors, such as new vs. old information. In some cases, NPs may even get shifted to the right of a component that is objectively longer than the NP itself. But there is no doubt that the weight of NPs does have a facilitating effect on the intrusion of another element – adverbial or prepositional phrase – between a verb and its object in Present-Day English. From a contrastive perspective, the question is whether this facilitating effect is comparable in the different groups. Table 6 shows the “heaviness” of object noun phrases occurring in V-AdvO sequences, according to L1. Heaviness is calculated here simply by the number of words; “light” NPs are those of 1-3 words (N, Det+N or Det+Adj+N), “medium” NPs are those of 4-6 words, mostly N of N constructions, and “heavy” NPs are those with 7 words or more, typically containing that complements, relative clauses or other post-modifiers. This is a somewhat arbitrary way of evaluating the heaviness of an NP, but Wasow (1997a) notes that different ways of characterising weight give very similar results. In the NS corpora, shifted NPs have a clear tendency towards heaviness; nearly half contain 7 words or more, and only 20% are light NPs of 3 words or less. This is in line with the observation by Stallings et al. (1998: 392) that “[w]hen the direct NP is short […] shifted structures […] are typically judged to be very awkward or ungrammatical”. In the learner corpora, however, the tendency is much less clear; in some cases (Polish, Bulgarian, Czech) there are roughly equal proportions of light, medium and heavy NPs; in others (Italian and Spanish, notably) the majority of shifted NPs are light. What is striking is that even though it is speakers of verb-raising languages who continue to be most different from speakers of English, the other groups also diverge from nativespeaker usage with respect to the types of NPs that they shift. In other words, even when, quantitatively, learners do not produce more V-Adv-O sequences than

Adverb placement in learner English

137

Table 6. Object NPs in V-Adv-O sequences NS English French 1 French 2 Italian Spanish Bulgarian Czech Polish Russian Dutch German Swedish Finnish

“light” NPs 20% 43% 39% 60% 55% 38% 32% 34% 43% 0% 22% 0% 29%

“medium” NPs 32% 35% 37% 19% 30% 31% 25% 32% 39% 62% 56% 0% 42%

“heavy” NPs 48% 22% 23% 21% 16% 31% 43% 34% 17% 38% 22% 100% 29%

natives, those that they do produce are likely to appear strange because they have non target-like characteristics. This can be seen in examples (28)-(29), taken from the German component of ICLE: (28)

(29)

[...] from one moment to another the silent crowd is turning to an exploding and errupting vulcano applauding frenetically the runner. (ICLE-GE) Friedrich Schiller, one of the greatest German writers who ever lived, gave once a statement which I can follow without any hesitation [...] (ICLEGE)

As already seen in Table 5, the German speaking learners in ICLE make very little use of V-Adv-O placement, with 3.83 occurrences per 100,000 words, compared with 6.1 occurrences in the NS corpora. But when they do place adverbs in this position, it is not necessarily for the same reasons as native speakers. There is no doubt a risk that non-canonical structures that would pass unnoticed in NS production will be judged erroneous in non-native production, just because they are known to be non-native, but examples (28) and (29) appear to diverge from NS usage in two different respects. In (28) the adverb frenetically forms an appropriate pair with the verb applaud (although collocation is more common with the corresponding noun – frenetic applause; see below, Section 4.2, on collocational effects in adverb placement), but it is extremely unusual to shift a 2-word noun NP such as the runner, which in this example has fewer syllables than the adverb itself. And even when the NP is heavy, that may not be a sufficient condition for it to be shifted; the relative clause in (29) certainly makes the NP heavy, but the closed-class adverb once typically occupies median (pre-verbal) rather than final position, thus making NP shift unnecessary.

138 4.2

John Osborne Verb-adverb collocations

The second of the conditions under which V-Adv-O order often appears in Present-Day English is shown in (30), which is taken from the Chambéry Corpus but contains a direct citation from an NS source: (30)

Alec Broers, the Vice-Chancellor of Cambridge declared that “at a time of diminishing support from public funds, this generosity is crucial and the University takes very seriously its responsibility to justify these contributions through continued excellence in research and teaching.” (Chambéry Corpus)

This example has not only a heavy object NP, but also a strong semantic association between the adverb and its verb. Omitting the adverb (??and the University takes its responsibility to justify...) would make the sentence meaningless, since take seriously functions here like a multi-word verb, both parts of which make a necessary contribution to the overall meaning. It should be noted that strong collocational links of this kind may not, by themselves, be sufficient to justify V-Adv-O order; of the two examples (31) and (32), both from ICLE, probably only the first would be judged appropriate in the target language. (31) (32)

Anyway, many people do not take seriously the importance of the committment they are going through with a baby. (ICLE-IT) Nowadays the people is beginning to be tired of this, so the Government has begun to take seriously the subject. (ICLE-SP)

Nevertheless, combinations such as take seriously, handle carefully, judge fairly, put clearly, etc differ from other verb-adverb associations in that the adverb cannot normally be placed to the left of the verb (??and the University seriously takes its responsibility...). This situation is puzzling in a language that normally requires non-final adverbs to occupy this position, unless one considers that these verb-adverb pairs are bound to each other in some way. In a study of verb associations with prepositional phrases, Wasow (1997a, b) classified verbpreposition combinations into three types: semantically opaque collocations (take ... into consideration), transparent collocations (bring ... to an end) and noncollocations (bring … to class). Analysis of his examples, extracted from the Canadian Hansard corpus, showed that heavy NP shift (i.e. giving the order take into consideration [NP] rather than take [NP] into consideration) occurred with 60% of the opaque collocations, 47% of the transparent collocations and only 15% of the non-collocational combinations, showing a clear link between collocation and NP shift. Classifying the nature of verb-adverb combinations in a similar way is more problematic, and I know of no study that has attempted to do so, but less systematic evidence suggests that semantic bonding between verb and adverb does play a role in NP shift. With take(s)/took [...] seriously [...], for example,

Adverb placement in learner English

139

analysis of 1,053 occurrences in the British National Corpus indicates that NP shift tends to take place once the weight of the NP exceeds four words (thus: takes the issue of communication seriously..., but could not take seriously his defence of Spanish independence). These are tendencies rather than systematic characteristics; there are a few cases of heavier NPs that are not shifted (take the question of high-rise housing seriously) and of lighter NPs that are shifted (take seriously the problems of smoking). But for other verbs commonly co-occurring with seriously – such as affect, damage, injure – there are no examples of NP shift, even with very heavy object NPs, no doubt because in such cases the adverb is regularly placed to the left: ...would seriously damage the independent role of National Rivers Authority and its ability to clamp down on pollution. To what extent do collocation and semantic association influence learners’ placement of adverbs? Analysis of 517 V-Adv-O sequences extracted from ICLE and the Chambéry Corpus reveals a considerable range of lexical items – 144 different adverbs, from accurately to zealously, and a similar diversity of verbs – but one major type of verb-adverb association does emerge from this diversity. About 20% of all the V-Adv-O sequences contain a verb of perception/cognition associated with an “enhancing” adverb. Some examples of these are shown in (33)-(46): understand + adverb (33) [...] we may therefore understand quite properly the sentence [...] (ICLECZ) (34) [...] the writer might think that the reader will understand better the text if he/she understands the allusion. (ICLE-SP) (35) I understand very well this point of view: grey factories are [...] (ICLEFR) (36) In the spoken language we probably understand more the *logical sense* [...] (ICLE-IT) (37) [...] they felt and treated nature sensitively, understood intuitively nature’s intercommunication with people. (ICLE-RU) (38) Wilde understand perfectly well the role of women in society. (ICLE-SP) (39) Moreover the reading of books not just helps students to understan and know better an author but to improve our openmindedness. (ICLE-IT) know + adverb (40) Such people don’t know the world around them but they know perfectly the TV programmes. (ICLE-CZ) (41) When I come home late on a Saturday’s night from a friend’s birthday party they know exactly the time of my arrival. (ICLE-GE) (42) But to know really well a book is not so simply. (ICLE-IT) verb + clearly (43) In this exemple we see clearly the contrast between life and absence of life... (ICLE-FR)

140 (44) (45) (46)

John Osborne why not trying to speak with each other and to explain clearly each other’s opinions [...] (ICLE-FR) Perhaps, this definition seems superficial, but it summarize clearly the point I am analized. (ICLE-SP) We have to be able to distinguish clearly each thing and make our own conclusions. (ICLE-SP)

A related, but more episodic phenomenon can be observed in other verbadverb combinations occurring in V-Adv-O sequences, where the adverb serves to intensify or to complement a semantically incomplete verb. Examples of these are shown in Table 7. Some of these combinations (follow blindly, refuse categorically) are genuine collocations, in the sense that they not only display a semantic association, but also co-occur in native usage. Others, like differentiate distinctively, are more idiosyncratic. Although they are logically more numerous in the groups where V-Adv-O order is more frequent, examples occur in nearly all groups, irrespective of L1. The one exception is the Swedish component of ICLE, where V-Adv-O sequences are particularly rare, and the only adverb appearing in this position is also. It may be that these verb-adverb associations are in fact a sign of developing proficiency, since one of the tasks of language learners in becoming more native-like is precisely to build up an increasing stock of appropriate collocations. As associated lexical items “become paired” (Wray 2002: 211) and are retrieved together from memory, there is an attendant risk that they may be deployed as a continuous unit, in defiance of syntactic rules. Table 7. One-off combinations accept passively accomplish brilliantly analyse objectively applaud frenetically change permanently choose freely choose unbiasedly denounce publicly 4.3

describe meticulously differentiate distinctively elect freely eliminate completely enforce strictly enrich culturally facilitate enormously

follow blindly grasp desperately influence positively reduce drastically refuse categorically scrutinize thoroughly solve jointly

Lexical deficiencies?

Associations of verb+adverb may also be a way of compensating for lexical difficulties or of avoiding risks, by using a relatively safe general verb and completing its meaning with an adverb. The joint use of verb+adverb as a semilexicalised unit then results in a V-Adv-O sequence, which could have been avoided by synthesising the overall meaning into a single less general verb, or by using a nominal expression of the type have a [x] effect on.

Adverb placement in learner English

141

Affect negatively (mar, hinder, impede, impair, etc): (47) (48)

Undeniably, the language affects badly (hinders) the learning process (Chambéry Corpus) Universities have been overloaded with an excessive number of students. This has affected enormously (damaged, reduced) the quality of the education. (ICLE-SP)

Affect positively (improve, benefit, help, etc): (49) (50)

(51)

People who dream lead happy lives and their inner happiness affects positively (benefits) those who suffer (ICLE-PO) So despite Mrs Ramsay’s death she is still influencing positively (having a good influence on) the people: Lily Briscoe but also Mr Ramsay who decides to go to the lighthouse with James [...] (ICLE-FR) [...] the way journalists select their words, or order them in a sentence, affects deeply (alters, colours) the meaning of the text (ICLE-FR)

Other cases where verb+adverb serve to replace a lexical item are more aspectual in nature, for example where renewal of a process is expressed through verb+again, rather than by lexical derivation (re-verb), or where iteration would have been better expressed through a verb or adjective of repetition: (52) (53) (54) 4.4

Consequently I classified again (reorganised) my ideas and I imagined a new outline. (Chambéry Corpus) At the end of my arguments I would like to write again (restate) my main thought (ICLE-CZ) He gives constantly hints (keeps hinting, gives repeated hints) that […] (ICLE-FI) Scope of adverbs

A final pattern which is present in NS production but overused in learner English is V-Adv-O sequences containing a conjunct adverb, notably additives such as also. Altogether, these account for another 20% of the V-Adv-O sequences in the learner corpora. There are, for example, 54 occurrences of V-also-O in ICLE, distributed throughout the corpus, with the exception of the German, Russian and Bulgarian components. Such sequences also appear, although less frequently, in the NS corpora, but they differ from many of the learner examples in that the scope of also tends to be limited to the following NP, as in (55), and so the position of also in the sentence is motivated, whereas in the learner corpora, also frequently appears to have scope over the entire clause, as in (56).

142 (55)

John Osborne

(56)

[...] they suggest a journey into the unknown, but in the context of this poem, they imply also a sexual exploration. (Essay Bank) Solidarity plays also an important part. (ICLE-FR)

5.

Conclusions: adverb placement and corpus studies

At this point, I should like to return to the questions listed in Section 1, summarising the answers which emerge from comparison of the corpora, and then draw some more general conclusions about the nature of adverb placement in post-intermediate learner English and about the kind of data that are desirable for making useful comparisons between learner corpora. Does non native-like adverb placement continue to appear in the free production of post-intermediate learners? Yes, even in carefully monitored written production, up to 4.5% of adverbs appear in non-canonical position, between the verb and its direct object, compared with only 0.6% in NS productions. However, this is not true of all learner groups, since some actually produce fewer V-Adv-O sequences than do native speakers. Are results consistent (for the same L1) across different corpora? For the only L1 for which two comparable corpora were analysed (French), there are some slight variations in the relative frequency of initial and final adverb placement, but very little difference in V-Adv-O position. Does adverb placement in L2 English vary according to L1? Assuming that the learners represented in ICLE are at a roughly comparable level, yes. Speakers of verb-raising Romance languages (Spanish, Italian, French) show a much greater frequency of V-Adv-O placement than speakers of V2 Germanic languages (Swedish, Dutch, German), with speakers of non-raising Slavic languages or of Finnish lying in between. If there is persistent difficulty in achieving native-like adverb placement, is this linked to specific L1-L2 pairings? The answer to the previous question suggests that it is, at least to a certain extent. The higher incidence of V-Adv-O sequences in the Spanish, Italian and French speaking groups could be the result of syntactic patterns being transferred from the L1. However, this does not explain why speakers of Swedish, Dutch and German, whose languages also permit V-Adv-O order, produce very few such sequences. Is this simply because of greater overall proficiency and familiarity with the target language, resulting in patterns of usage which are generally more target-like? Is it because the greater mobility of syntactic units in their L1 makes it easier for them to master what is, after all, something of an anomaly in English – a degree of freedom in the ordering of elements in postverbal position? Do adverb-placement errors surface at random, or are there specifiable circumstances? There is always the possibility, as Sidney Lamb (2000: 92) observes, “that some patterns [...] can be explained by the ingenuity of the linguist in finding patterns even in chaos”. Nevertheless, there do seem to be effects which, although far from systematic, are clearly recurrent. Firstly,

Adverb placement in learner English

143

conditions which favour non-canonical adverb placement in L1 English – heavy NPs and collocational links between verb and adverb – also play a role in learner English, but in slightly different ways. Whereas in NS productions both conditions are often met when there is V-Adv-O order – there is both a heavy NP and a strong link between verb and adverb – in learner English one of these conditions by itself may be enough. In addition, and variably according to L1, the thresholds for these conditions to trigger V-Adv-O tend to be lower than in NS English; shifted NPs are lighter on average, and the collocational links between verb and adverb are weaker or more idiosyncratic. Do the same factors trigger non native-like adverb placement irrespective of L1? The answer to this seems to be yes, but in different degrees. L1 transfer predisposes speakers of Spanish, Italian and French to place adverbs in noncanonical position more frequently than do learners with other L1s, but the specific contextual factors that favour V-Adv-O order (verb-adverb collocation, presence of a relatively heavy NP, use of discourse adverbs) can also be responsible for V-Adv-O sequences appearing in the productions of other learners. Accounts of learners’ placement of adverbs typically adopt a syntactic approach to the problem. While this is important for examining hypotheses about the initial state of L2 grammars and their subsequent development, it ignores many of the factors influencing the actual production of adverbs, particularly at later stages of learning. The general conclusion that can be drawn from comparison of these learner corpora is that adverb placement in L2 English, at post-intermediate to advanced level, is persistently probabilistic, even for learners whose L1, like English, does not normally allow V-Adv-O order. This means that any syntactic explanation has to allow for the possibility that learners will produce sequences which are ungrammatical both in L1 and in L2. As for the conditions which determine whether V-Adv-O order will be chosen in any given case, these are not exclusively syntactic; the placement of adverbs is also influenced by semantic and phraseological associations among components of the VP. If these associations act in opposition to learners’ knowledge of syntactic requirements, then the adverb and NP will have competing claims on adjacency to the verb. In resolving these competing demands, very proficient learners may end up making adverb placement choices which are indistinguishable from those of a native speaker. But given the number of factors involved and the indeterminate nature of evidence available in the input, it is not surprising that many learners continue to make occasional choices which would strike a native speaker as awkward. Learner corpora are invaluable for tracking apparently unsystematic aspects of L2 usage such as these. In cases where the phenomena under investigation are diffuse, subject to optionality, and occur only sporadically, access to a sufficiently large corpus is often the only way to discern any underlying patterns. Comparing learner corpora therefore has considerable potential for discovering whether these patterns are the same or different from one L1 to another, or from one L2 to another, and thus for understanding which

144

John Osborne

interlanguage phenomena are linked to specific L1-L2 groupings, and which are not. However, the factors involved can be numerous and difficult to quantify, particularly in the case of residual errors at post-intermediate levels of language learning, whose occurrence may also be sensitive to task conditions and the type of production. Corpora used in contrastive interlanguage studies thus need to meet the somewhat conflicting needs of variety and comparability. Three kinds of development are particularly useful in meeting these needs and for enriching the potential for contrastive studies based on learner corpora. (i) More varied learner corpora. Many existing learner corpora are of a similar type to those examined here – written productions, often argumentative or descriptive essays that are a common task in educational settings. But given that many language phenomena are sensitive to the task being carried out, type of discourse, speaker-hearer interaction, etc, there is a need for corpora collected in other conditions of production, particularly oral production, both in monologue and in interaction, covering a range of tasks and including, in some cases, videotaped data. The ESF database,1 now more than ten years old, is a good example of an oral second language database covering a variety of spoken activities in different L1-L2 pairings. More recent or on-going projects for collecting spoken L2 data are LINDSEI (Louvain International Database of Spoken English Interlanguage),2 and the PAROLE corpus under construction at the University of Savoie.3 (ii) More learner corpora for languages other than English. It is not surprising, given the domination of English as an L2 and the availability of English language learners all over the world, that far more learner data are available for English than for any other language. But it is useful to be able to compare not only productions in the same L2 from speakers of different L1s, but also productions of a similar nature across several L2s. So on-going work in compiling corpora in Spanish, French, Italian and other languages is a welcome addition. The French Learner Language Oral Corpora project (FLLOC), for instance, now has a sister project for Spanish, SPLLOC.4 Parallel corpora, of learners carrying out comparable tasks in different L2s, can provide particularly useful data, and this is also one of the aims of the PAROLE project, mentioned above. (iii) More supplementary information about the learners involved. The more one diversifies the types of production and the L1s and L2s of the learners involved, the more important it is to know whether the data being compared really are comparable. Corpora are generally accompanied by learner profiles providing information about the length of time spent learning the language, periods of stay in the country where the language is spoken, exposure to the language, knowledge of other languages, etc. But given the range of educational practices in different countries, the value placed on language learning, individual motivation, opportunities for using the language outside the classroom, and so on, the fact that two learners have apparently similar profiles does not guarantee that they have reached comparable levels of competence. It is helpful, therefore, to

Adverb placement in learner English

145

have as much information as possible about their linguistic proficiency, if possible from independent standardised tests. Notes 1

The transcribed data from the ESF (European Science Foundation) project can be browsed from the Max Planck Institute website: http://www.mpi.nl.

2

LINDSEI contains oral interviews with L2 English speakers from eleven different L1 backgrounds. It is one of the projects of the Louvain Centre for English Corpus Linguistics: http://cecl.fltr.ucl.ac.be/CeclProjects/Lindsei/ lindsei.htm.

3

The PAROLE corpus (Parallèle, Oral, en Langue Etrangère) contains parallel oral productions (i.e. obtained from identical tasks) from speakers of L2 English, French and Italian. Further information about this project, and about the Chambéry written corpus, is available on request from the author: [email protected].

4

The FLLOC data are available as part of the CHILDES Talkbank, and on the FLLOC website: http://www.flloc.soton.ac.uk. For information on the parallel SPLLOC project, see http://www.splloc.soton.ac.uk.

References Barlow, M. (1999), MonoConc Pro. Houston: Athelstan. Beck, M.-L. (1998), ‘L2 Acquisition and obligatory head movement: Englishspeaking learners of German and the Local Impairment Hypothesis’, Studies in Second Language Acquisition, 20(4): 311-348. BNC Consortium (2001), The British National Corpus, version 2. Distributed by Oxford University Computing Services. Council of Europe (2001), Common European Framework of Reference for Languages. Cambridge: Cambridge University Press. Electronic version available at http://www.coe.int/t/dg4/linguistic/Source/Framework_EN.pdf (last accessed on December 15, 2006). Eubank, L. (1996), ‘Negation in early German-English Interlanguage: more Valueless Features in the L2 initial state’, Second Language Research, 12(1): 73-106. Granger, S., E. Dagneaux and F. Meunier (eds) (2002), The International Corpus of Learner English. Handbook and CD-ROM. Louvain-la-Neuve: Presses Universitaires de Louvain. Hawkins, J. (1994), A Performance Theory of Order and Constituency. Cambridge: Cambridge University Press.

146

John Osborne

Hawkins, R. and C. Chan (1997), ‘The partial availability of UG in SLA: the failed functional features hypothesis’, Second Language Research, 13(3): 187-226. Lamb, S. (2000), ‘Bidirectional processing in language and related cognitive systems’, in: M. Barlow and S. Kemmer (eds) Usage Based Models of Language. Stanford: CSLI Publications. 87-119. Rissanen, M. (1999), The Helsinki Corpus of English Texts: Diachronic and Dialectal, in: K. Hofland, A. Lindebjerg and J. Thunestvedt (eds) ICAME Collection of English Language Corpora. Second edition. Bergen: University of Bergen HIT Centre. Schwartz, B. and R. Sprouse (1996), ‘L2 cognitive states and the Full Transfer/Full Access model’, Second Language Research, 12(1): 40-72. Sorace, A. (2003), ‘Near-Nativeness’, in: C. Doughty and M. Long (eds) The Handbook of Second Language Acquisition. Oxford: Blackwell. 130-151. Stallings, L., M. MacDonald and P. O’Seaghdha (1998), ‘Phrasal ordering constraints in sentence production: Phrase length and verb disposition in heavy-NP shift’, Journal of Memory and Language, 39: 392-417. Trahey, M. and L. White (1993), ‘Positive evidence and preemption in the second language classroom’, Studies in Second Language Acquisition, 15: 181204. Vainikka, A. and M. Young-Scholten (1996), ‘Gradual development of L2 phrase structure’, Second Language Research, 12(1): 7-39. Wasow, T. (1997a), ‘Remarks on grammatical weight’, Language Variation and Change, 9: 81-105. Wasow, T. (1997b), ‘End-weight from the speaker’s perspective’, Journal of Psycholinguistic Research, 26: 347-361. White, L. (1991), ‘Adverb placement in second language acquisition: some effects of positive and negative evidence in the classroom’, Second Language Research, 7: 133-161. Wray, A. (2002) Formulaic Language and the Lexicon. Cambridge: Cambridge University Press.

The use of the English article system by Chinese and Spanish learners María Belén Díez-Bedmar Universidad de Jaén Szilvia Papp University of Portsmouth Abstract In this paper we investigate the use of the English article system in two comparable learner corpora, Chinese-English and Spanish-English. Such investigation is significant as article use is at the interface of syntax, semantics and pragmatics. In addition, the differences between these particular learner groups are interesting in that the article system does not exist in Chinese and the Spanish article system is slightly different from the English one (Lyons 1999). The theoretical framework used for the analysis of the data is that of Bickerton’s (1981) semantic wheel based on a taxonomy of features - [± specific reference] and [± hearer knowledge] - and the theory of definiteness within pragmatics (e.g. Hawkins 1978, 1991). We highlight and explain similarities as well as differences between the article use of these two groups, using the methods of explanation of contrastive analysis (CA) and contrastive interlanguage analysis (CIA) in an integrated contrastive model (ICM) (Granger 1996, Gilquin 2000/2001). Given the assumptions that article underuse is a grammatical deficit and article overuse and misuse is a pragmatic deficit (cf. Thomas 1989), we test the hypothesis that Chinese learners have both a grammatical and pragmatic deficit, whereas Spanish learners may only have a pragmatic problem, especially in generic contexts.

1.

Introduction

Learners of English need to have a good grasp of the article system if they are to be able to use discourse reference appropriately in communication. Knowledge of the article system is a very important aspect of both grammatical and communicative competence since the article system is employed for the expression of definiteness and specificity and is linked to such pragmatic notions as shared assumptions between discourse participants about their knowledge of and familiarity with a referent or the inferrability, identifiability, salience and relevance of a referent to them as discourse participants. Some teachers of English may disagree with this and claim that non-native-like use of the article system is an aspect that usually does not hinder communication (see e.g. Seidlhofer 2004: 220). They would argue that in case of uncertainty, the intended meaning can easily be inferred by a process of negotiation between participants. Consider the hypothetical exchange in the following example, where B’s response is a request for clarification on whether a specific or a non-specific indefinite meaning was intended by A:

148 (1)

María Belén Díez-Bedmar and Szilvia Papp A: I want to marry a Norwegian girl. B: Do you know any?

This instant negotiation process to clarify the intended meaning is frequent and very effective in spoken discourse. In written discourse, however, non-native use of the article system may contribute to the intended message being obscured or incorrectly conveyed, since this negotiating process is not available (cf. Tomiyana 1980). Consider, for instance, the following excerpt from a learner’s essay comparing university degrees in China and the UK. We have highlighted the use of articles in front of the noun “university/ies”. It is clear that, although seemingly somewhat systematic, this learner’s article use is non-native-like: (2)

Compare and contrast the university in the UK and China This essay is going to talk about several different and similar aspects of the university in the UK and China. There are four main themes: the university contribution to the development of the local community, the growing number of international students, the campus life and studying resource sharing. The following paragraphs will discuss more details about the topics above. There are two interesting parallels between Ø university in the UK and China, which are the contribution to the development of the local community and the growing number of international students. […] (CNL122)1

In order to describe the use of the article system by Chinese and Spanish learners of English, we first review the theoretical framework for article use. Then, in line with the theme of this volume, we conduct a contrastive analysis (CA), where we briefly outline the expression of [± specific reference] and [± hearer knowledge] in English, Spanish and Chinese within a theoretical system of definiteness based on Hawkins’ (1978, 1991) and Bickerton’s (1981) work. In the contrastive interlanguage analysis (CIA) section, we report on findings from two comparable learner corpora (Chinese-English and Spanish-English learner data) and compare them with three native corpora (native English, Spanish and Chinese data). Thus, we contribute to the existing body of research with new learner and comparable native data analysed within Huebner’s (1983) analytical framework based on Bickerton’s (1981) taxonomy of [± specific reference] and [± hearer knowledge]. In fact, it was our wish to analyse article use as part of a pragmatic system within language use, which led us to choose Bickerton’s semantic/pragmatic taxonomy, since use of articles is based on contextual factors within discourse, mainly speakers’ assumptions about their hearers’ knowledge state (Hawkins 1978, 1991). If a language does not grammatically mark definiteness and specific/generic reference, then speakers of this language will have to learn a new grammatical subsystem in their L2. This is the case with Chinese learners of English, whose L1 does not grammaticalise nominal reference in the way English

The use of the English article system by Chinese and Spanish learners

149

does in its article system. Our prediction is that, since the Chinese language lacks a grammaticalised article system, Chinese learners of English face a grammatical learning task which will manifest itself in their underuse of the English articles due to the omission of the and a/an and a higher rate of use (overuse) of the zero (Ø) article. On the other hand, given the assumption that article misuse and overuse is a pragmatic deficit that also characterises L1 English child speech (cf. Thomas 1989), we expect that both Chinese and Spanish learners will exhibit a pragmatic problem by misusing and overusing articles. While for the Spanish speakers this prediction means that we expect them to misuse and overuse English articles in their interlanguage, for the Chinese these seemingly contradictory predictions mean that we would expect them not only to underuse, but also to misuse and overuse articles until they learn to restrict their use to the specific semantic contexts and pragmatic functions in which they are employed in English. 2.

Theoretical framework

The English article system is at the interface of syntax, semantics and pragmatics. Therefore, it can be described and explained either from a formalistic or functionalist perspective. In traditional grammars, the article system is typically described as part of the morpho-syntax of English, since the categories of definiteness, countability and reference are grammaticalised in a complex article system which is part of an even more complex determiner system premodifying the noun head within the English noun phrase (e.g. Quirk et al. 1985, Biber et al. 2002, Downing and Locke 2002, Huddleston and Pullum 2002, 2005, Carter and McCarthy 2006). In the generative tradition, the article system is described within the Determiner Phrase (DP), the determiner being one of a series of functional categories that head a maximal projection with abstract features such as [± definite], [± specific], [± count] (e.g. Szabolcsi 1987; for an overview, see Zamparelli 2005). On the other hand, when it comes to accounts of the article system within discourse, it is usually described as part of a range of semantic and discourse-pragmatic devices used by the English language for informationstructuring and topic continuity purposes within interaction (e.g. Hawkins 1978, 1991, Lambrecht 1994).2 2.1

Functions and meanings of English articles

2.1.1 Definite article the The English definite article the is the most frequent word in English.3 The sheer frequency of its occurrence is well over 7% of all words (Leech et al. 2001). Hawkins (1991) summarises the overall function of the definite article as:

150

María Belén Díez-Bedmar and Szilvia Papp The conventionally implicates that there is some subset of entities, {P}, in the universe of discourse which is usually manifest to S[peaker] & H[earer] on-line and within which definite referents exist and are unique. (Hawkins 1991: 414)

The “subset of entities in the universe of discourse” that Hawkins’ definition refers to is pragmatically delimited by the context of utterance, i.e. the mutual knowledge or shared beliefs between participants (see Hawkins 1991: 408): x the immediate situation set (elements of the physical context) and the larger situation of utterance (based on participants’ general background knowledge relative to the physical location of the speakers); x the previous discourse set; x general knowledge sets; and x a line of associative sets stemming from these. Hawkins’ notion of “mutual on-line manifestation” means that the reference of the NP introduced by the is clear to the hearer as well as the speaker, and it is familiar and/or identifiable by both. In other words, the use of the definite article directs the hearer to the referent of the NP by signalling that s/he is in a position to identify it. It invites the hearer to exploit clues in the linguistic and extralinguistic context to establish the identity of the referent, which they can see or have heard about or whose existence they can infer from something else they have heard, i.e. possible or probable rather than known existence (cf. Trenkiü 2002a). Another important notion in Hawkins’ definition is that of “uniqueness”: the definite article may signal that there is just one entity satisfying the description used, relative to a particular context. However, it has been pointed out that definiteness with plurals and mass nouns involves not uniqueness but inclusiveness or even identifiability (Lyons 1999: 278). 2.1.2 Indefinite article a/an According to Brown (1986: 96, cited in Trenkiü 2002a), a speaker uses the indefinite article when he “believes that his hearer doesn’t share or wonders whether his hearer shares his information”.4 Trenkiü (2002b) points out that if we take the “sharing of information” as having a mutually manifest set (i.e. a {P}-set in Hawkins’ definition), in which the referent would exist and be unique, then we can paraphrase the definition as follows: the speaker uses an indefinite article a/an/some “when he believes that his hearer does not have a P-set in which the referent would exist and be unique, or if the speaker wonders if his hearer would have a P-set in which the referent would exist and be unique” (Trenkiü 2002a: 65).

The use of the English article system by Chinese and Spanish learners

151

Downing & Locke (2002: 429-430) point out that with singular count nouns the article a can refer to both specific and non-specific entities, “the different interpretations being deduced from the different predications”, as can be seen in their examples I’ve bought a new car compared with I need a new car. 2.1.3 Zero/null article Ø or bare NPs Palmer (1939) distinguishes two types of null article preceding a noun head:5 1) the zero article for mass/non-count and plural nouns (such as milk and eggs); 2) the null article with bare count nouns and proper names (lunch, London).6 According to Chesterman (1991), an NP marked by the zero or null article is grammatically not delimited, actualised or distinguished, but is just an idea which potentially has scope over the whole class of entities represented by it.7 Downing & Locke (2002: 433) define generic nominal groups [NGs] in the following way: “Generic NGs refer to entities as representatives of their whole class, in abstract statements about their typical characteristics or habitual activities”. 2.2

Bickerton’s semantic wheel

After considering the functions and syntactic uses of English articles, namely expressing generic and specific reference, in Figure 1 we outline Bickerton’s (1981) semantic wheel which we use for the analysis of our data. Bickerton (1981) based his taxonomy on two binary semantic and discourse-pragmatic features: [± specific reference] or [± SR] and [± hearer knowledge] or [± HK]. Specificity involves the speaker’s intention to refer, whereas hearer knowledge refers to the speaker’s assumption about the hearer’s familiarity with or capacity to infer the referent (Hawkins 1978, 1991).

Figure 1. Bickerton’s (1981) semantic wheel for NP reference (from Huebner 1983)

María Belén Díez-Bedmar and Szilvia Papp

152

Huebner (1983) set up the following taxonomy based on Bickerton’s (1981) semantic wheel: 1. [-SR, +HK] – the, a, Ø (generics) 2. [+SR, +HK] – the (referential definites) a. unique or conventionally assumed unique referent b. referent physically present c. referent previously mentioned in discourse d. specific referent otherwise assumed common knowledge 3. [+SR, -HK] – a, Ø (referential indefinites, first mentions) 4. [-SR, -HK] – a, Ø (non-referentials – attributive indefinites, non-specific indefinites …) Table 1, based on Ekiert (2004) and adapted from Huebner (1985), Thomas (1989) and Goto Butler (2002), shows examples for each category: Table 1. Taxonomy of the English article system with examples Features Type 1 [-SR, +HK]

Type 2 [+SR, +HK]

Type 3 [+SR, -HK]

Environment Generic nouns

Referential definites previous mentions specified by entailment specified by definition unique in all contexts unique in a given context Referential indefinites first-mention nouns

Articles a, the, Ø

the

a, Ø

Type 4 [-SR, -HK]

Non-referential nouns attributive indefinites non-specific indefinites

a, Ø

Type 5

Idioms Other conventional uses

a, the, Ø

Examples

Ø Fruit flourishes in the valley. Ø Elephants have trunks. The Grenomian is an excitable person. They say the elephant never forgets. A paper clip comes in handy. An elephant never forgets. Pass me the pen. The idea of coming to the UK was… I found a book. The book was… The first person to walk on the moon… Chris approached me carrying a dog. I’ve bought a new car. A man phoned. I keep sending Ø messages to him. I’ve got Ø friends in the UK. I’ve managed to find Ø work. Alice is an accountant. I need a new car. I guess I should buy a new car. A man is in the ladies, but I haven’t seen him. Ø Foreigners would come up with a better solution. All of a sudden, he woke up. In the 1950s, there weren’t many cars. His family is now living Ø hand to Ø mouth.

The use of the English article system by Chinese and Spanish learners

153

3.

Contrastive analysis of definiteness and specificity in Chinese and Spanish

3.1

Definiteness and specificity in Chinese

The expression of definiteness in Mandarin Chinese has been investigated by several authors. Robertson (2000) builds on Li & Thompson’s (1981) work by providing an excellent summary of definiteness of Mandarin NPs. Since Chinese lacks an article system and it is a topic prominent language, definiteness or generic reference is expressed by word order. That is, Mandarin Chinese speakers place an NP in the preverbal topic position to mark it as definite. The topic position is occupied by an element which is, by definition, assumed to be known information by both speaker and hearer. In addition to word order, Mandarin Chinese also uses determiners for the expression of definiteness. In particular, the demonstratives zhèi (‘this’) and nèi (‘that’) are beginning to take on some functions in modern spoken Mandarin that the definite article the has in English (Huang 1999). Note that in standard Mandarin the equivalents of these demonstratives are 扨 zhè (‘this’) and ㇊ nà (‘that’). On the other hand, the numeral ৻ yƯ (‘one’) plus a classifier are taking on some of the functions that the indefinite article a(n) plays in English (Chen 2004). However, Sio (2006) points out that in Chinese it is always the structure to the left of the noun phrase that is responsible for the encoding of specificity, an observation that is obscured in unmodified noun phrases. The following sentences exemplify the four different semantic/discourse-pragmatic contexts in Chinese: (3)

Cars are useful.



懵 ᤚ ᦭↪⊛‫ޕ‬ cars are useful8

(4)

The car I bought is red.



ᚒ ⃿⊛ ㇊戕 懵 ᤚ 儱⦡⊛‫ޕ‬ I bought the car is red

(5)

I bought

a

(generic)

red car.

(definite)

(specific indefinite)

ᚒ ⃿ੌ 㧔৻㧕 戕 儱 懵‫ޕ‬ I bought a red car (6) 

I need a car. ᚒ 㔛ⷐ 㧔৻㧕戕 懵‫ޕ‬ I need a car

(non-specific indefinite)

154

María Belén Díez-Bedmar and Szilvia Papp

Note that in examples (3)-(6) the character ㇊ is used for ‘the’ and the partially optional character㧔৻㧕is used for ‘a’. However, articles are very rarely used in Mandarin Chinese. In fact, there is no use of ৻ yƯ (‘one’), 扨zhè (‘this’) or ㇊ nà (‘that’) in (7), the opening sentence of a native Chinese essay on capital punishment: (7)

₏ 㙟Ⓙ 㸊⒠ ᧨氥 ⏗ 常ⅉ when/once (time) mention death penalty, first make person/people 勣㎂Ⓙ 䤓⻀ associate/remember of just/exactly (emphasis) 㢾 棃㭽㭽䤓 嫛⒠ ♿ᇭ is ghastly execute punishment platform. (‘Whenever we mention the death penalty, what people think of is a ghastly execution on a platform.’) (CNN001)

On the basis of the description of the lack of an article system in Chinese, we hypothesise that Chinese speakers will underuse the definite and indefinite articles in English, overuse the zero article, and even misuse them until they manage to learn to restrict the use of them to the specific semantic contexts and pragmatic functions in which they are employed in English. 3.2

Definiteness and specificity in Spanish

The definite article, one of the elements which characterise Romance languages (Leonetti 1999: 789), is used to restrict and define the reference of noun phrases. Similarly to English, the definite article in Spanish, el, la, los and las,9 and the indefinite article, un, una, unos and unas,10 can be used to convey generic meaning, as can be seen in El guepardo es fácil de domesticar (‘The cheetah is easy to tame’) or Un guepardo es fácil de domesticar (‘A cheetah is easy to tame’),11 in which we refer to any representative member of a class instead of the whole class, which would be expressed with the definite article in Spanish (El guepardo es fácil de domesticar). However, generic reference in Spanish requires the presence of any article, be it definite or indefinite (Laca 1999: 896), since the absence of the article denotes, in Christophersen’s (1939: 33-35) terms, a “parti-generic meaning”, i.e. not referring to all the members of a class (Laca 1999: 903, 904, 908). Therefore, we cannot have a bare noun phrase in a [-SR, +HK] context, as can be seen in *Ø Guepardo es fácil de domesticar (‘Cheetahs are easy to tame’). Nevertheless, there is an exception to the generalisation that bare noun phrases cannot have a generic reference. Those are the cases in which we find modifications such as así

The use of the English article system by Chinese and Spanish learners

155

(‘in this way’), de este tipo (‘of this kind’) and como ese (‘like this’), which restrict the type of the elements referred to in the bare noun phrase to the ones of a particular type and, as a result, refer to the totality of elements, i.e. toto-generic reference. See, for instance, Ø Tipos como ese no suelen tener paciencia (‘Men of that kind are not usually patient’) (Laca 1999: 908). As for [+SR, +HK] contexts, the definite article in Spanish, el, la, los and las, conveys the same meaning as the in English in such contexts, for example A. Bryce llegó ayer a Santander. El conocido escritor… (‘A. Bryce arrived in Santander yesterday. The well-known writer…’), Recuerdo las primeras imágenes de la llegada a la luna (‘I remember the first images of the moon landing’) (cf. Leonetti 1999: 796-800).12 The indefinite article and bare noun phrases are also found in Type 3 contexts, that is, [+SR, -HK]. On the one hand, the indefinite article may be used to refer for the first time to a person or thing, e.g. Parece que lo ha descubierto un periodista (‘It seems that a journalist found it out’), or to denote a quantified expression, as in Amanda vino con unas amigas, which would correspond to the English use of some (‘Amanda came with some friends’). On the other hand, bare noun phrases can be used in non-quantified expressions such as Traigo Ø botellas de vino (‘I bring Ø bottles of wine’), as opposed to the quantified Traigo unas botellas de vino (‘I bring some bottles of wine’). Finally, we may also find the definite article and bare noun phrases to express [-SR, -HK]. Indefinite and bare noun phrases can be used in evaluative or classifying interpretations like Pilar es (una) persona responsable13 (‘Pilar is a responsible person’). Besides, it is important to note that the position of the noun phrase is an aspect to take into account, since the indefinite determiner and bare noun phrases may alternate in noun phrases which are in predicative position (Leonetti 1999: 851).14 Regarding bare noun phrases in non-specific reference contexts, they can be seen in Por precaución, hizo hervir Ø agua dos veces (‘To be on the safe side, he ordered boiling water twice’) (Laca 1999: 899) or when defining somebody’s job, occupation or role, as in Es catedrática en Valencia (‘She is a Professor in Valencia’). After this brief review of the use of the definite, indefinite and zero articles in Spanish, we can conclude that there is one major difference in the use of articles in English and Spanish: the zero article cannot be used to express generic reference (in toto-generic interpretations) in Spanish. Thus, *Ø Guepardo es fácil de domesticar or * Ø Guepardos son fáciles de domesticar would be incorrect in Spanish, whereas Ø Cheetahs are easy to tame is correct in English. Apart from this, some minor differences can also be noticed in contexts where Spanish does not require any article and English requires an indefinite one. Consider, for instance, María tiene Ø coche and *Mary has Ø car. Therefore, we hypothesise that Spanish learners of English will mainly present problems when using the zero article to convey generic meanings in English.

156 4.

María Belén Díez-Bedmar and Szilvia Papp Overview of L1 and L2 studies on article acquisition

Before reviewing the research on the use of the article system by Chinese and Spanish learners of English, we will outline the acquisition of articles by children and present some findings on article use by other L2 learners of English. Regarding the use of articles by children, they have been found to omit articles between the ages of 18-36 months. However, they seldom use determiners incorrectly. They acquire the indefinite article before the definite article and the pragmatic concept of non-shared assumptions is present from age 3. Since after this age they have a developed notion of specificity and discourse referentiality and finally (crucially) of familiarity of DP referent, the majority of children after that age exhibit a near-mastery level of use of the definite article (for a review, see Thomas 1989). Several studies have investigated the use of the English article system by learners with an L1 with no articles (see for example Young 1996 for Czech and Slovak; Geranpayeh 2000 for Persian; Robertson 2000, Goto Butler 2002 and Lardiere 2004 for Chinese; Trenkiü 2002b for Serbian; White 2002 and 2003 for Turkish; Cowan et al. 2003 for Korean; Ionin 2004 for Korean and Russian; Žegarac 2004 for Croatian). Trenkiü (2002a) looked at omission and substitution errors among Serbian learners of English. In opposition to other studies analysing and explaining the data with Bickerton’s semantic wheel (Huebner 1983, Tarone & Parrish 1989, Thomas 1989, Master 1990), she argues that substitution errors occur because learners create non-target-like interpretations for articles, that is, their formfunction mappings are learner-specific. According to her account, Serbian learners believe that the indefinite article is used to mark “individuation” and the definite article is used to mark discrete entities that already come individuated. This account is similar to those of Goto Butler (2002) and Yoon (1993) who found that Japanese learners of English have differential concepts of the semantic and pragmatic terms of “hearer’s knowledge”, “specificity” and “countability”, or Žegarac (2004), who claims that Croatian learners come up with and test idiosyncratic hypotheses about the English article system. Ekiert (2004) investigated 20 Polish ESL and EFL learners of English (another Slavic language without an article system) in a cloze-type elicitation study. She found that the most difficult categories for both groups at all proficiency levels were generics and idioms. Cowan et al. (2003: 456) report on Korean learners of English who have problems with articles, among other aspects of the language, which are clearly induced by their L1 syntax. Regarding French learners, Osborne (2004) reports on the inconsistencies found between learners’ explicit knowledge and their actual use. Two dissertations have compared the use of the article system by Spanish learners and learners with an L1 with no article system (Simons 2001 for Spanish and Korean learners, and Trademan 2002 for Spanish and Japanese learners). Both of these dissertations have found that learners with an L1 with no article

The use of the English article system by Chinese and Spanish learners

157

system (i.e. Korean and Japanese learners of English) performed markedly worse than learners with Spanish as L1. The earlier studies looked at the acquisition sequence of articles and found that learners whose L1 lacks an article system first acquire the zero article, then the definite article, and lastly the indefinite article (e.g. Parrish 1987, Master 1988a, b, 1997, Thomas 1989, Yoon 1993, Simons 2001). The developmental pattern for the definite article usually shows a U-shaped curve, that is, an initial worsening indicated by increasing overuse of the definite article which decreases with proficiency (Liu & Gleason 2002). Thomas (1989) showed evidence that learners (both L1 and L2 learners) overuse the definite article as they associate it with referentiality. Child L1 acquirers overuse it in specific indefinite contexts, typically where a new noun phrase is introduced into the discourse with a definite article as opposed to an indefinite article. This has been termed “ego-centric speech” (cf. Thomas 1989). It has also been found that L2 learners use the definite article at a significantly higher rate in specific indefinite than nonspecific indefinite contexts. 4.1

Chinese

Robertson’s (2000) study is devoted to article use by Chinese learners of English. Robertson looked at article omission and found that most of his 18 Chinese learners dropped the article in 30% of obligatory contexts, and that accuracy was higher in definite contexts (79.7%) than in indefinite contexts (72.1%), with a statistically significant difference between the two. This result corroborates other researchers’ findings (Platt 1977, Huebner 1985, Parrish 1987). Thus, Robertson’s study showed evidence of Chinese learners’ problems with indefinite contexts.15 However, Robertson (2000) did not look at article overuse and misuse. He claims that learners have difficulty forming correct form-function mappings between the surface forms and abstract features of the DP (i.e. [± number] and [± definite]). As a result, in the 22% of obligatory contexts where they omit the definite or indefinite article they use different strategies, such as the “determiner drop” principle due to the extended determiner scope of a previous NP, the “recoverability” principle when a referent is inferable from the context, and a “lexical transfer” principle from their L1, Chinese. Fen-Chuan (2001) also investigated article use by Chinese learners of English. She found that Chinese learners have difficulty distinguishing [± hearer knowledge] (e.g. misuse of the for a or Ø) and [± countability] (e.g. misuse of a for Ø, or Ø for a). This result is corroborated by Butler’s (2000) finding on Japanese learners, another L1 with no article system, that, as the proficiency of learners increases, their problems with detection of hearer knowledge and countability still remain. Lardiere (2004) analysed one female native Chinese speaker’s English and noted that she supplied the definite and indefinite article in obligatory contexts in 84% and 76% of the cases respectively. Her results lend support to the finding that definite articles are acquired earlier than indefinite articles. However, 16%

158

María Belén Díez-Bedmar and Szilvia Papp

and 24% error rates with both articles show that her female subject is still far from being in control of the definiteness feature of English. In this section we have reviewed evidence that learners whose L1 does not have an article system may create idiosyncratic representations of the English article system, and that Chinese speakers tend to have a considerable problem using the English article system. 4.2

Spanish

The use of the article system by Spanish learners has also been covered in the literature. First, it is important to note that articles are relatively unproblematic for advanced learners of English whose L1 is Spanish. In fact, the results of the two studies which analyse the written production of first-year students at university level, Valero Garcés (1997) and Díez-Bedmar (2004, 2005), show similar findings regarding the percentage of errors in the learners’ production of the article system.16 First, Valero Garcés (1997: 75) reports that omission, the only category in which problems with articles are included, accounts for 5.2% of the errors in her corpus. Within this category, it is the omission of the definite article the which represents most of the errors. Then, the research by Díez-Bedmar (2004, 2005) reveals that problems related to the article system, including over-, under- and misuse of articles, represent only 5.3% of the total amount of errors made by first-year university students. The role that transfer plays in the incorrect use of the article system by Spanish university students has been the focus of some studies. García Gómez & Bou Franch (1992) analysed the L1-induced errors made by second-year university students. Their data only show four instances in which the article system is incorrectly used, all of them due to overgeneralisation of the definite article, which represents 7.4% of the total number of errors in their corpus. Partly following Wyatt’s (1973) error taxonomy, González Cruz (1996) divided the errors which she found in her corpus into transfer and no-transfer errors. Her results show that transfer errors are twice as frequent as no-transfer errors, the problems with the articles accounting for 16.71% of all the L1-transfer errors and 0.42% of the no-transfer errors. Similarly, Cebreiros Álvarez (2004), in his study on transfer and empty categories in the oral interlanguage of Spanish learners and immigrants, also found that the problems that the article system poses to Spanish university students were motivated by transfer in a high percentage of cases, which was reduced slightly as the learners’ level increased from beginner to intermediate and advanced (46%, 41% and 38%, respectively). Other research has cast light on the problems that articles pose to Spanish learners at secondary level. As was the case with university students, some investigations, such as Bazo Martínez (2001) and Wood Wood (2004), have highlighted the role that the L1 plays in secondary school learners’ written production. A cross-sectional analysis of the written production of 199 learners at secondary level when performing different tasks was also conducted by Bueno González (1992).17 His results reveal that the problems with the article system,

The use of the English article system by Chinese and Spanish learners

159

included in the word-class “determiners”, only account for 9.62% of the errors in his corpus. Following the error taxonomy by Dulay, Burt & Krashen (1982), Jiménez Catalán (1996) reveals that the substitution of articles accounts for 3.18% of the errors in her corpus, which represents the eighth most common error (the errors of omission and addition of articles are not included in the top ten errors, and their percentages are not provided). The production by learners at official language schools in Spain has also been considered. For example, Moreno Ibáñez & Ruiz Gracia (1985), using an explanatory error taxonomy (following Selinker 1972) consisting of overgeneralisation, transfer, performance, teacher induced errors and avoidance errors, consider that the errors in the use of articles are located in the category of overgeneralisation, the most frequent category at the three levels under study. However, the percentages of errors for each aspect in this category are not included, so it is impossible to know whether the use of articles accounts for most of the cases in this category or not. Having reviewed the literature on the studies where the use of the article system is mentioned when analysing the interlanguage of Spanish learners of English, five important issues stand out: i) it seems more or less clear that the English article system does not pose many problems to Spanish learners at secondary or university level, even though the low percentages in the review are not comparable due to the various methodologies used by the researchers; ii) the papers published so far use different taxonomies, which prevents the exact comparability of results across studies; iii) the analyses have been mainly quantitative, even though percentages are not always provided ; iv) there are no publications which exclusively focus on the use of the article system by Spanish learners of English at university level;18 and v) it seems that the L1 plays an important role in the incorrect use of the English article system by Spanish learners of English, which may decrease as the learners’ proficiency level increases. 5.

Data analyses: corpora, methodology and findings

In order to carry out the CA and CIA analyses (see Gilquin, this volume), three native corpora and two learner corpora were collected and analysed. Table 2 shows a breakdown of the corpora used in this study. Table 2. Breakdown of the corpora used in the present study L1 English Chinese Spanish Chinese Spanish

Type NS NS NS NNS NNS

No of words or characters 22,504 words 954 characters 39,006 words 39,663 words 39,881 words

No of essays 44 1 119 74 101

160

María Belén Díez-Bedmar and Szilvia Papp

As a first step, for the CA, we need to describe the native corpora and the analysis we carried out on them. The native English corpus used as our control corpus is composed of 44 essays written by first-year students at the University of Portsmouth, UK, on the same topics that were given to the Chinese learners of English. The English L1 corpus consists of 22,504 words, each essay containing 510 words on average. As far as the Chinese L1 data is concerned, we only collected one essay from a first-year Chinese-speaking student at the University of Portsmouth performing the same task. The Spanish L1 corpus was collected at the Universidad de Jaén in Spain. It is composed of 119 essays, totalling 39,006 words, each essay containing 330 words on average, written by first-year Spanish university students doing a Spanish Studies degree. The topics on which these learners wrote were exactly the same topics as the ones provided to the Spanish learners of English for the learner corpus. These three L1 corpora (English, Chinese and Spanish) were tagged using the coding system specified in Table 3. Thus, it was possible to compare the use of the article system in the learners’ L1 and in the English native corpus in a corpus-based CA approach. Consequently, we could predict the problems that the learners of English were prone to have on the basis of the differences in the way of expressing genericity, specificity and definiteness in their L1 and their foreign language, English. Table 3. Tagset for the article system (based on Bickerton 1981)

DA (definite article) IA (indefinite article) ZA (zero article)

Type 1 generic

Type 2 definite

Type 3 specific indefinite

Type 4 non-specific indefinite

1DA

2DA

1IA

3IA

4IA

1ZA

3ZA

4ZA

As a second step, for the CIA, two learner corpora were compiled and analysed. First, we analysed a subsection of the Portsmouth Chinese-English learner corpus (Papp forthcoming), which is part of the second version of the International Corpus of Learner English (ICLE) project (Granger et al. forthcoming). The corpus for this study consists of 74 essays written by Chinese learners of English, comprising a total of 39,663 words, each essay containing 540 words on average. The learners had studied English for an average of 8 years at school and at least one year at university. The data used for the analysis of the production by Spanish learners of English is a subsection of the longitudinal learner corpus compiled and errortagged at the Universidad de Jaén. This subsection is composed of 101 essays, amounting to 39,881 words, each essay consisting of 390 words on average,

The use of the English article system by Chinese and Spanish learners

161

written under the same external conditions, i.e. there was a time limit imposed on essay writing and no access to reference material. The learners had studied English for an average of 7 years at school (primary and secondary) and one year at university. Therefore, they can be compared with the Chinese learners in our study and with the learners who wrote under the same conditions in the ICLE project, even though their proficiency levels may differ.19 Once the learner corpora were collected, transcribed and corrected by native speakers, we tagged all incorrect uses of the definite and indefinite articles in the data using the general error tag “GA”, as stated in the ICLE error tagging manual (Dagneaux et al. 1996: 10). Then, we devised a set of tags which would allow us to conduct a more fine-grained analysis of the correct and incorrect uses of the English articles by our learners. We divided the instances of correct and incorrect article use into four types, so that we indicated which semantic context the noun phrase belonged to (Type 1: generic, Type 2: definite, Type 3: specific indefinite, Type 4: non-specific indefinite). In the cases of correct use, we tagged the article which the learner had appropriately used by means of the coding system outlined in Table 3. In the case of the incorrect instances (over-, under- and misuse), instead of stating the article incorrectly used by the learner (which is easily retrievable in a concordance), we indicated the article that would have been used by a native speaker in that specific context.20 Thus, we find the acronym for the target article after the type of context (1, 2, 3 or 4) and the general error tag (GA) to indicate an error in the use of the article system. As a result, the sixteen error tags in Table 4 were used, corresponding to the types of errors found in our learner corpora. Table 4. Tags for incorrect use of English articles in each semantic context Article used by the learner Target article

Type 1 generic

DA (definite article) IA (indefinite article) ZA (zero article)

1GAIA 1GAZA 1GADA 1GAZA 1GADA 1GAIA

Type 2 definite

Type 3 specific indefinite

Type 4 non-specific indefinite

2GADA

3GADA 3GAZA 3GADA 3GAIA

4GADA 4GAZA 4GADA 4GAIA

2GADA

Consider, for example, the following sentences by a Chinese and a Spanish learner, respectively:

María Belén Díez-Bedmar and Szilvia Papp

162 (8) (9)

Of course not, (4GAZA) the life is (3GAIA) the basic human right, […] (CNL22) […] and staying with your family during (2GADA) summer. (SPL-2-A-6)

Having tagged the correct uses and error-tagged the two learner corpora, we then proceeded to retrieve the instances of the correct and incorrect uses of the articles in the corpora. To retrieve the instances and count the obligatory contexts we used WordSmith Tools Version 3 (Scott 1999). In Table 5, we give a summary of all the corpora we used in our study with a count of all obligatory contexts. Table 5. Summary of obligatory contexts for articles in all the corpora used L1

Type

No of words or characters

English Chinese Spanish Chinese Spanish

NS NS NS NNS NNS

22,504 words 954 characters 39,006 words 39,663 words 39,881 words

No of obligatory contexts 3,266 133 4,774 5,774 3,952

Percentage

14.51% 13.94% 12.24% 14.65% 9.91%

There is a significant difference between the ratios of obligatory contexts and total word count in each of the five corpora (chi-square = 375.536, df = 4, p<0.001). This significant difference is due to the Spanish learners underproducing obligatory contexts for articles in their English essays compared to the Chinese learners and the native English speakers. 6.

Discussion

6.1

Contrastive Analysis

Table 6 displays the results obtained from the contrastive analysis of English, Spanish and Chinese article use. In other words, we can see the type of articles and their frequency in each semantic context that native speakers of English, Spanish and Chinese use in their L1 as evidenced in our three native corpora. From these data, we expect that Spanish learners of English will have problems with the use of the indefinite and zero articles in Type 1, as the indefinite article is not used by Spanish native speakers in the corpus, although it would be possible (cf. Section 3.2), and the zero article is not used in the generic context. On the other hand, Chinese learners of English are expected to overuse the zero article or, in other words, fail to supply the definite or indefinite articles. Also, they may find it difficult to restrict the use of the definite and indefinite articles to their correct semantic contexts, i.e. use them with their English

The use of the English article system by Chinese and Spanish learners

163

distribution. Therefore, while Spanish learners are hypothesised to have a pragmatic problem, Chinese learners are hypothesised to have both a grammatical and a pragmatic problem (cf. Thomas 1989). Table 6. Article use in the NS corpora by semantic context English L1 1 (generic) 2 (definite) 3 (specific indefinite) 4 (non-specific indefinite) Total English L1 Spanish L1 1 (generic) 2 (definite) 3 (specific indefinite) 4 (non-specific indefinite) Total Spanish L1 Chinese L1 Total Chinese L1 6.2

DA

IA

ZA

Total

64 (22.22%) 1265 (100%) 0

21 (7.29%) 0 419 (71.99%)

203 (70.48%) 0 163 (28%)

288 1265 582

0

126 (11.14%)

1005 (88.85%)

1131

1329 (40.69%)

566 (17.33%)

1371 (41.97%)

3266

30 (100%) 426 (100%) 0

0 0 74 (86.04%)

0 0 12 (13.95%)

30 426 86

0

41 (29.28%)

99 (70.71%)

140

456 (66.86%)

115 (16.86%)

111 (16.27%)

682

4 (3%)

17 (12.78%)

112 (84.21%)

133

Contrastive Interlanguage Analysis

In order to come up with a measure of accuracy in the use of each article (the, a/an and Ø), we counted the number of correct uses in obligatory contexts relative to the sum of all obligatory contexts and the contexts in which the article was supplied inappropriately (following the approach proposed by Pica 1984). This method of analysis, as we will see, results in a different hierarchical order for the accuracy of use of articles in general (i.e. not by types) than what is usually found by other researchers. The results for Chinese learners are displayed in Table 7, where the appropriate uses are in bold. All three differences are significant as measured by the z-test (see Butler 1995: 92): between the and a (z = 3.23, p<0.001), between the and Ø (z = 7.10, p<0.001), and between a and Ø (z = 1.73, p<0.05). The hierarchy of accuracy for Chinese learners is the following, with the being the least accurately used (due to the high overuse of the definite article the in the wrong contexts): Ø > a > the

164

María Belén Díez-Bedmar and Szilvia Papp

Table 7. Percentage of the, a and Ø in appropriate and inappropriate contexts for the Chinese group the contexts a contexts Ø contexts

the 67.40 7.46 15.79

a 0.85 73.18 0.74

Ø 9.26 14.81 75.94

The percentages of correct use for Spanish learners are shown in Table 8. The difference between the and a is significant (z = 7.46, p<0.001), and so is the difference between Ø and a (z = 9.138, p<0.001). However, the difference between the and Ø is not significant (z = 2.72, p = 0.9967). Therefore, the hierarchy of accuracy is the following for the Spanish learners: a > the / Ø Table 8. Percentage of the, a and Ø in appropriate and inappropriate contexts for the Spanish group the contexts a contexts Ø contexts

the 89.07 0.46 6.41

a 0.12 96.39 0.58

Ø 5.58 2.05 85.94

As seen from these results, the indefinite article is used with a higher degree of accuracy than the definite article in both groups, even though Spanish learners are better overall in their use of those articles. Additionally, Chinese learners seem to use the zero article most correctly, as has already been stated (Master 1997: 216), but it is difficult to ascertain whether this is a correct use of the zero/null article or non-suppliance, i.e. underuse, of any articles in those contexts. When we compare the Chinese and Spanish groups on the total correct uses of the, correct uses of a and correct uses of Ø separately, the differences are all highly significant (z = all p<0.001). When calculating the percentage of correct use of each article in each semantic context, the results found are those in Table 9 and the corresponding Figure 2. The comparison of the data obtained from the two learner corpora casts light on various issues. Firstly, Spanish learners have a similar (in the use of the zero article in Type 1 contexts) or higher accuracy rate compared to Chinese learners when they use English articles in all four semantic contexts. Secondly, Chinese learners have more problems than Spanish learners with the use of articles in Type 3 (specific indefinite) contexts. Therefore, examples such as those in (10) and (11) are frequent in the Chinese learner corpus:

The use of the English article system by Chinese and Spanish learners (10) (11)

165

[…] and give them (3GAIA) chance to change. (CNL54) turn on the television and get some thing to eat just like (3GAIA) the couch potato. (CNL19)

Table 9. Chinese and Spanish learners’ accurate use of English articles by semantic context Type 1 generic Chinese 97.14% 42.86% 85.99%

DA IA ZA

Type 2 definite

Spanish 100% 83.33% 85.99%

Chinese 88.90%

Type 3 specific indefinite

Spanish 93.83%

Chinese

Spanish

Type 4 non-specific indefinite Chinese Spanish

74.10% 69.42%

96.37% 100%

82.68% 83.98%

99.20% 98.46%

100 90 80 70 60

Chinese

50

Spanish

40 30 20 10 0 DA

IA

ZA

Type 1

DA Type 2

IA

ZA Type 3

IA

ZA

Type 4

Figure 2. Overall percentage of accuracy in article use by Chinese and Spanish learners of English Thirdly, especially striking is the misuse of the indefinite article in Type 1 (generic) contexts among Chinese learners of English, cf. (12) and (13): (12) (13)

It’s a traditional view that only (1GAIA) man can carry the family. (CNL69) […] people want to choose (1GAIA) baby’s sex. (CNL08)

Finally, Spanish learners of English show high accuracy percentages for the use of the definite, indefinite and zero articles in the four contexts, as seen in Table 9. However, the use of the indefinite and zero articles in Type 1 contexts

María Belén Díez-Bedmar and Szilvia Papp

166

shows some difficulty. Therefore, it is interesting to compare the frequency of use of the three articles in this context in the native Spanish, native English and Spanish learner corpus. To begin with the definite article, Spanish native speakers prefer it in these contexts (rather than the use of the indefinite article, the zero article not being possible in Type 1 contexts), as can be seen in Figure 3. When using it in the foreign language, these learners do not struggle with them, with an accuracy rate of 100% (see Table 9), but show an underuse of this article if compared to native speakers of English (64 instances in the native English corpus and 16 in the Spanish learner corpus). Regarding the use of the indefinite article, Spanish native speakers tend not to use it, as can be seen in Figure 3, even though it is possible to use it in this context in their mother tongue (cf. Section 3.2). Such a tendency may influence the underuse of this article in Type 1 contexts when Spanish learners write in the foreign language (5 instances in the Spanish learner corpus and 21 in the native English corpus), but does not cause major problems in its use, with an accuracy rate of 83.33%, even though this accuracy rate is the lowest for the Spanish learner corpus. Lastly, and contrary to expectations (see Section 6.1), Spanish learners manage to use the zero article in these contexts correctly (85.99%), even though it cannot be used in Type 1 contexts in their mother tongue. However, learners use it much more frequently than native speakers of English (198 instances in the native English corpus vs. 540 in the Spanish learner corpus). 600 500 400 Native Spanish 300

Native English Spanish Learner Corpus

200 100 0 1DA

1IA

1ZA

Figure 3. Use of the definite, indefinite and zero articles in the native Spanish and English corpora (absolute frequencies) Two possible reasons may be put forward for this overuse. On the one hand, it may be the case that the overuse stems from a teaching-induced error, since the use of the zero article to express generic meanings is overstressed by teachers in secondary schools in Spain.21 On the other hand, the definite article in

The use of the English article system by Chinese and Spanish learners

167

Type 1 contexts may be accompanied by a noun in singular or in plural form in Spanish, with a different nuance of meaning, since the former refers to a homogeneous group and the latter to a group which may not be homogeneous: (14)

(15)

Otro sitio más de la Sierra de Cazorla transformado por (1DA) el turismo (SPN-32) ‘Another place in Sierra de Cazorla changed because of tourism’ Aquel verano aprendí que (1DA) los amigos pueden ser para toda la vida (SPN-24) ‘That summer I learnt that friends may last forever’

Furthermore, the definite article in Type 1 contexts in English is normally followed by a noun in the singular, and it is also found in plural noun phrases with nationality nouns, e.g. the Germans, or noun phrases with an adjective functioning as a head, e.g. the rich, other cases of the definite article plus a plural noun not being possible to express generic reference (cf. Quirk et al 1985: 283, Huddleston & Pullum 2002: 407). This may be the reason why Spanish learners of English tend to use the definite article with singular nouns in Type 1 contexts (81.25%), whereas they may be likely to use the plural nouns with the zero article in the cases in which they would use the definite article followed by a noun in the plural form in their mother tongue, which may lead to an underuse of the definite article in Type 1 contexts and an overuse of the zero article in these contexts as seen in Figure 3. Furthermore, when doing so, they would also follow the teacher’s misleading recommendation that the generic contexts in English are expressed by means of the zero article. 6.3

Final remarks

Our CIA results seem to bear out the CA predictions that Chinese learners exhibit more non-native features in their overall use of English articles than Spanish learners. This was expected as Chinese learners need to establish a grammaticalised system for the expression of specific/generic reference and to take into account the hearer’s knowledge state, whereas Spanish learners need to adjust an existing system to the foreign language. We analysed our data according to Pica (1984), who proposes that both learners’ correct and incorrect use of morphemes should be taken into account in calculations. This resulted in an unexpected finding that, in contradiction to previous studies, the indefinite article seems to be less difficult to acquire, and it is rather the definite article that presents a problem, even at an advanced stage of acquisition, to both Chinese and Spanish learners. This corroborates Thomas’ (1989) claim that, especially in specific indefinite contexts, but also in non-specific indefinite contexts, the overgeneration (and overgeneralisation) of the definite article is a common phenomenon in L2 learning, similar to the phenomenon in L1 acquisition. In relevance theoretic terms (Sperber & Wilson 1995) this means that learners need to consider and accommodate the hearer/reader’s knowledge state by the

168

María Belén Díez-Bedmar and Szilvia Papp

appropriate and accurate use of the definite article in English. Additionally, we found that generic contexts are difficult for Chinese learners, and to a lesser extent, for Spanish learners. Chinese learners were found to use the zero article more than Spanish learners and they were also found to overgeneralise the use of the definite article more in inappropriate contexts, while Spanish learners overuse the zero article in Type 1 contexts, to the detriment of the use of the definite article in this same context. All in all, we can therefore say that Chinese learners exhibit a grammatical and pragmatic problem, whereas Spanish learners exhibit a pragmatic problem, which is manifest in their different use of articles in generic contexts. Acknowledgements The authors would like to thank Gaëtanelle Gilquin, Jennifer Thewissen and Fiona Barker for their helpful comments on an earlier draft of this paper. Notes 1

The reference to the corpora used is given after the examples in brackets. The three letters refer to the corpus from which the sentence is taken. Thus, we will refer to the Chinese learner corpus with “CNL”, to the Chinese native corpus as “CNN”, to the Spanish learner corpus with “SPL” and to the Spanish native corpus with “SPN”.

2

The semantic factors include notions such as specificity, existence, uniqueness, inclusiveness, distinctness, discreteness, and the discoursepragmatic factors include notions such as salience, prominence, relevance, inferrability, discourse referentiality, familiarity of the DP referent, hearer and speaker knowledge, identifiability, accessibility, mutual manifestness on-line (see Sperber & Wilson 1995), (non)shared assumptions, new, continuous, reintroduced NP referents in discourse, as well as the Givenness hierarchy.

3

For an account of the uses of the definite article, please refer to Quirk et al. (1985: 265-272, 282-286) or Huddleston & Pullum (2002: 368-371, 407408), among others.

4

On the uses of the indefinite article a/an, see, among others, Quirk et al. (1985: 272-273) and Huddleston & Pullum (2002: 371-372, 406-407).

5

A note of caution is needed here regarding the terminology to be used in the rest of this paper. Zero, null article or bare NP have been used to designate NPs which are not preceded by any determiner (predeterminers, central determiners or postdeterminers in Quirk et al. 1985: 255-264). The grammar by Quirk et al. (1985) prefers the term “zero article”, while the

The use of the English article system by Chinese and Spanish learners

169

grammar by Huddleston & Pullum (2002) refers to “bare NPs”. We use both terminologies in this paper. 6

In our study we have decided to include null articles in Type 5 (idioms and set expressions, see Table 1), and thus have excluded them from our data analysis.

7

For a review of the uses of the zero article, please refer to Quirk et al. (1985: 265, 282, 274-281, 286-287) and Huddleston & Pullum (2002: 409410), among others.

8

All translations from Chinese into English, English into Chinese, English into Spanish, and Spanish into English are ours.

9

El is used for masculine singular, la for feminine singular, los for masculine plural and las for feminine plural. The invariable lo is not included in this piece of research, since it is used in specific emphatic constructions (Lo interesante que es ese libro, ‘How interesting that book is’) or may function as a pronoun (Ella también lo creía, ‘She also thought /believed it’).

10

Un is used for masculine singular, una for feminine singular, unos for masculine plural and unas for feminine plural. The existence of a plural form for the indefinite article, i.e. unos and unas, is exclusive to the paradigm of indefinite articles in Ibero-romanic languages (Leonetti 1999: 841).

11

Unless otherwise indicated, the examples used for Spanish come from Leonetti (1999).

12

Note that one of the main characteristics of the grammatical behaviour of the article in Spanish is the possibility to combine with different grammatical categories (adjective phrases, prepositional phrases and relative clauses). Therefore, we are not dealing with those cases in which the article is used before non-finite clauses and subordinate clauses with que, for example El haber trabajado aquí le resultará útil en el futuro (‘Having worked there will be useful in the future’) or No depende de mi el que la reunión se celebre o no (‘It’s not my decision if the meeting takes place or not’).

13

The languages which have a plural form for the indefinite article tend to use this form, unos, to express an evaluative metaphorical interpretation (Leonetti 1999: 852), as in Eran linces or Eran unos linces (‘They were lynx’ and ‘They were sharp-eyed’, respectively), where it can be seen that the use of the indefinite article in the noun phrase favours the metaphorical interpretation of the sentence.

170

María Belén Díez-Bedmar and Szilvia Papp

14

The reason for this alternation is the fact that this optionality represents a progressive extension of the indefinite article to meanings which were previously conveyed by bare noun phrases (Leonetti 1999: 851).

15

Perhaps related to this is the finding that Chinese learners of Hungarian are more likely to use the definite than the indefinite conjugation of the verb in Hungarian (Langman & Bayley 2002), despite the fact that definite verb forms express largely redundant functions and that indefinites are more frequent in Hungarian.

16

Note that these two studies are not fully comparable, since the error taxonomies differ. Valero Garcés (1997) divided the errors into the categories of spelling, morphological, syntactical, lexical, omission, addition and others, while Díez-Bedmar (2004, 2005) used her learner corpus, error-tagged with the UCLEE (Dagneaux et al. 1996), and therefore including the categories of form, grammar, lexico-grammar, lexis, word redundant, word missing, word order, register and style. Within this error taxonomy, the incorrect use of articles is in the category of grammar.

17

This study includes the four years which existed at secondary level before the implementation of the LOGSE educative system in Spain, that is, 1º BUP, 2º BUP, 3º BUP and COU.

18

To the best of our knowledge, the only piece of research which focuses exclusively on the use of the article use by Spanish students is the PhD thesis by Wood Wood (2002) with students at secondary level, which has not been published yet.

19

Since there is no independent external measure of proficiency in the ICLE project apart from learners’ information about previous learning experiences and exposure to the target language and their institutional status (Granger et al. 2002: 14), proficiency level is an issue when it comes to comparisons between different learner groups. The only external criterion we have for the Chinese students is their recorded IELTS score which ranged between 4.5-7 at the time the data was collected. As far as the Spanish students are concerned, they are all supposed to have, at least, an intermediate to post-intermediate level, as stated in the curriculum of English in secondary schools and the University entrance exam (Selectividad).

20

The data in the two learner corpora were checked by native speakers of English, to whom we are grateful for their help and comments.

21

This conclusion has been reached after questioning some teachers of English at secondary level. In fact, one of them stated that she explained the zero article in Type 1 contexts and opposed it to the use of the definite

The use of the English article system by Chinese and Spanish learners

171

article in Type 2 contexts, but did not explain the use of the definite article in Type 1 contexts. References Bazo Martínez, P. (2001), ‘A qualitative analysis of the interlanguage found in compulsory secondary education students in the Canary Islands’, Revista Canaria de Estudios Ingleses, 43: 209-218. Biber, D., S. Conrad and G. Leech (2002), Longman Student Grammar of Spoken and Written English. Harlow: Longman. Bickerton, D. (1981), Roots of Language. Ann Arbor, MI: Karoma Press. Bueno González, A. (1992), ‘Errores en la elección de palabras en inglés por alumnos de Bachillerato y C.O.U’, in: A. Bueno González, J.A. Carini Martínez and Á. Linde López (eds) Análisis de Errores en Inglés: Tres Casos Prácticos. Granada: Universidad de Granada. 39-105. Butler, C. (1995), Statistics in Linguistics. Oxford: Blackwell. Butler, Y.G. (2000), The role of metacognition in the development of the English article system among non-native speakers. PhD dissertation. Dissertation Abstracts International, A: The Humanities and Social Sciences, 60(8): 2837-A. Carter, R. and M. McCarthy (2006), Cambridge Grammar of English. Cambridge: Cambridge University Press. Cebreiros Álvarez, R. (2004), ‘The operation of transfer and interlanguage principles: the case of empty categories in the interlanguage of Spanish learners of English’, VIAL, 1: 33-54. Chen, P. (2004), ‘Identifiability and definiteness in Chinese’, Linguistics, 394: 1129-1184. Chesterman, A. (1991), On Definiteness: A Study with Special Reference to English and Finnish. Cambridge: Cambridge University Press. Christophersen, P. (1939), The Articles: A Study of their Theory and Use in English. Copenhagen: Einar Munksgaard. Cowan, R, H.E. Choi and D.H. Kim (2003), ‘Four questions for error diagnosis and corrections in CALL’, CALICO Journal, 20: 451-463. Dagneaux, E., S. Denness, S. Granger and F. Meunier (1996), Error Tagging Manual Version 1.1. Centre for English Corpus Linguistics. Louvain-laNeuve: Université catholique de Louvain. Díez-Bedmar, M.B. (2004), Approaches to the interlanguage evolution of firstyear students at the University of Jaén. Unpublished MA dissertation, The University of Jaén. Díez-Bedmar, M.B. (2005), ‘Struggling with English at university level: errorpatterns and problematic areas of first-year students’ interlanguage’, in: P. Danielsson and M. Wagenmakers (eds) The Corpus Linguistics Conference Series (available online at http://www.corpus.bham.ac.uk/ PCLC/, last accessed on January, 10 2007).

172

María Belén Díez-Bedmar and Szilvia Papp

Downing, E. and P. Locke (2002), A University Course in English. Second edition. London: Routledge. Dulay, H., M. Burt and S. Krashen (1982), Language Two. New York: Oxford University Press. Ekiert, M. (2004), ‘Acquisition of the English article system by speakers of Polish in ESL and EFL settings’, Teachers College, Columbia University Working Papers in TESOL & Applied Linguistics, 4(1): 1-23. Fen-Chuan, C. (2001), ‘The acquisition of English articles by Chinese learners’, Second Language Studies, 20(1): 43-78. García Gómez, E. and P. Bou Franch (1992), ‘Estudio experimental sobre interferencias lingüísticas’, in: J.R. Losada Durán and M. Mansilla García (eds) Actas de VIII Congreso Nacional de Lingüística Aplicada. Vigo 2, 3 y 4 de mayo de 1990. Vigo: Servicio de Publicaciones de la Universidad de Vigo. 279-291. Geranpayeh, A. (2000), ‘The acquisition of the English article system by Persian speakers’, Edinburgh Working Papers in Applied Linguistics, 10: 37-51. Gilquin, G. (2000/2001), ‘The Integrated Contrastive Model. Spicing up your data’, Languages in Contrast, 3(1): 95-123. González Cruz, M.I. (1996), ‘The role of Spanish transfer: the need to include contrastive analysis in EFL teaching’, in: S.G. Fernández-Corugedo (ed.) Some Sundry Wits Gathered Together. A Coruña: Sevicio de Publicacións Universisade da Coruña. 55-71. Goto Butler, Y. (2002), ‘Second language learners’ theories on the use of English articles: an analysis of the metalinguistic knowledge used by Japanese students in acquiring the English article system’, Studies in Second Language Acquisition, 24(3): 451-480. Granger, S. (1996), ‘From CA to CIA and back: an integrated approach to computerized bilingual and learner corpora’, in: K. Aijmer, B. Altenberg and M. Johansson (eds) Languages in Contrast. Papers from a Symposium on Text-based Cross-linguistic Studies. Lund 4-5 March 1994. Lund: Lund University Press. 37–51. Granger, S., E. Dagneaux and F. Meunier (eds) (2002), The International Corpus of Learner English. Handbook and CD-ROM. Louvain-la-Neuve: Presses Universitaires de Louvain. Granger S., E. Dagneaux, F. Meunier and M. Paquot (eds) (forthcoming), The International Corpus of Learner English. Handbook and CD-ROM. Version 2. Louvain-la-Neuve: Presses Universitaires de Louvain. Hawkins, J.A. (1978), Definiteness and Indefiniteness: A Study in Reference and Grammaticality Prediction. London: Croom Helm & San Diego, CA: Academic Press. Hawkins, J.A. (1991), ‘On (in)definite articles: implicatures and (un)grammaticality prediction’, Journal of Linguistics, 27: 405-442. Huang, S. (1999), ‘The emergence of a grammatical category definite article in spoken Chinese’, Journal of Pragmatics, 31: 77-94.

The use of the English article system by Chinese and Spanish learners

173

Huddleston, R.D. and G.K. Pullum (2002), The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Huddleston, R.D. and G.K. Pullum (2005), A Student’s Introduction to English Grammar. Cambridge: Cambridge University Press. Huebner, T. (1983), A Longitudinal Analysis of the Acquisition of English. Ann Arbor, MI: Karoma. Huebner, T. (1985), ‘System and variability in interlanguage syntax’, Language Learning, 25: 141-163. Ionin, T.R. (2004), Article semantics in second language acquisition. PhD dissertation. Dissertation Abstracts International, A: The Humanities and Social Sciences, 64(9): 3270-A. Jiménez Catalán, R.M. (1996), ‘Frequency and variability in errors in the use of English prepositions’, Miscelánea: A Journal of English and American Studies, 17: 171-187. Laca, B. (1999), ‘Presencia y ausencia de determinante’, in: I. Bosque and V. Demonte Barreto (eds) Gramática Descriptiva de la Lengua Española. Madrid: Editorial Espasa Calpe, S. A. 891-928. Lambrecht, K. (1994), Information Structure and Sentence Form. Cambridge: Cambridge University Press. Langman, J. and R. Bayley (2002), ‘The acquisition of verbal morphology by Chinese learners of Hungarian’, Language Variation and Change, 14(1): 55-77. Lardiere, D. (2004), ‘Knowledge of definiteness despite variable article omission in second language acquisition’, Proceedings of the Annual Boston University Conference on Language Development, 28(1): 328-339. Leech, G., P. Rayson, and A. Wilson (2001), Word Frequencies in Written and Spoken English: Based on the British National Corpus. Harlow: Longman. Leonetti, M. (1999), ‘El artículo’, in: I. Bosque and V. Demonte Barreto (eds) Gramática Descriptiva de la Lengua Española. Madrid: Editorial Espasa Calpe, S. A. 787-890. Li, C. and S. Thompson (1981), Mandarin Chinese: A Functional Reference Grammar. Berkeley: University of California Press. Liu, D. and J.L. Gleason (2002), ‘Acquisition of the article the by nonnative speakers of English: an analysis of four nongeneric uses’, Studies in Second Language Acquisition, 24(1): 1-26. Lyons, C. (1999), Definiteness. Cambridge: Cambridge University Press. Master, P. (1988a), ‘Teaching the English article system (Part 1)’, English Teaching Forum, 26: 2-7. Master, P. (1988b), ‘Teaching the English article system (Part 2)’, English Teaching Forum, 26: 18-25. Master, P. (1990), ‘Teaching the English articles as a binary system’, TESOL Quarterly, 24: 461-498. Master, P. (1997), ‘The English article system: Acquisition, function, and pedagogy’, System, 25: 215-232.

174

María Belén Díez-Bedmar and Szilvia Papp

Moreno Ibáñez, C. and C. Ruiz Gracia (1985), ‘Aplicación de las teorías del análisis de errores a un caso práctico’, in: F. Fernández (ed.) Pasado, Presente y Futuro de la Lingüística Aplicada en España. Valencia: Servicio de Publicaciones de la Universidad de Valencia. 371-377. Osborne, J. (2004). ‘Articles and non-count nouns in learner English: perception and production’, in: B. Lewandowska-Tomaszczyk (ed.) Practical Applications in Language and Computers. àódĨ: àódĨ University Press. 359-369. Palmer, H.E. (1939), A Grammar of Spoken English on a Strictly Phonetic Basis. Second edition. Cambridge: Heffer. Papp, Sz. (forthcoming), ‘Portsmouth Chinese-English learner corpus’, in: S. Granger, E. Dagneaux, F. Meunier and M. Paquot (eds) The International Corpus of Learner English. Handbook and CD-ROM. Version 2. Louvainla-Neuve: Presses Universitaires de Louvain. Parrish, B. (1987), ‘A new look at methodologies in the study of article acquisition for learners of ESL’, Language Learning, 37(3): 361-383. Pica, T. (1984), ‘Methods of morpheme quantifications: their effect on the interpretation of second language data’, Studies in Second Language Acquisition, 6: 69-78. Platt, J. (1977), ‘The ‘creoloid’ as a special type of interlanguage’, Interlanguage Studies Bulletin, 2: 22-38. Quirk, R., S. Greenbaum, G. Leech and J. Svartvik (eds) (1985), A Comprehensive Grammar of the English Language. London: Longman. Robertson, D. (2000), ‘Variability in the use of the English article system by Chinese learners of English’, Second Language Research, 16(2): 135-172. Scott, M. (1999), Wordsmith Tools. Version 3. Oxford: Oxford University Press. Seidlhofer, B. (2004), ‘Research perspectives on teaching English as a lingua franca’, Annual Review of Applied Linguistics, 24: 209–239. Selinker, L. (1972), ‘Interlanguage’, International Review of Applied Linguistics, 10: 209-231. Simons, G.V. (2001), The acquisition of the English determiner system: sequence, order and transfer. PhD dissertation. Dissertation Abstracts International, A: The Humanities and Social Sciences, 62(5): 1818-A. Sio, J. (2006), Modification and reference in the Chinese nominal. Unpublished PhD dissertation, Leiden University. Sperber, D. and D. Wilson (1995), Relevance. Communication and Cognition. Second edition. Oxford: Blackwell. Szabolcsi, A. (1987), ‘Functional categories in the noun phrase’, in: I. Kenesei (ed.) Approaches to Hungarian, Vol. 2. Szeged: JATE. 167-189. Tarone, E. and B. Parrish (1989), ‘Task-related variation in interlanguage: the case of articles’, Language Learning, 38: 21-44. Thomas, M. (1989), ‘The acquisition of English articles by first- and secondlanguage learners’, Applied Psycholinguistics, 10: 335-355. Tomiyana, M. (1980), ‘Grammatical errors and communication breakdown’, TESOL Quarterly, 14: 71-79.

The use of the English article system by Chinese and Spanish learners

175

Trademan, J.E. (2002), The acquisition of the English article system by native speakers of Spanish and Japanese: A cross-linguistic comparison. PhD dissertation. Dissertation Abstracts International, A: The Humanities and Social Sciences, 63(6): 2225-A. Trenkiü, D. (2002a), ‘Nominal definiteness in English. Hawkins’ (1991) account and some modifications’, Views and Voices, 1: 53-72. Rijeka: University of Rijeka Press. Trenkiü, D. (2002b), ‘Form-meaning connections in the acquisition of English articles’, EUROSLA Yearbook, 2: 115-133. Valero Garcés, C. (1997), ‘The interlanguage of Spanish students beginning English Philology’, GRETA, 5(2): 74-78. White, L. (2002), ‘Morphological variability in endstate L2 grammars: the question of L1 influence’, Proceedings of the Annual Boston University Conference on Language Development, 26(2): 758-768. White, L. (2003), ‘Fossilization in steady state L2 grammars: persistent problems with inflectional morphology’, Bilingualism: Language and Cognition, 6(2): 129-141. Wood Wood, M. (2002), La transferencia de la L1 a la L2 (Español-Inglés): el artículo en la lengua escrita de los alumnos en las pruebas de acceso a la universidad. Unpublished PhD dissertation, Las Palmas de Gran Canaria. Wood Wood, M. (2004), ‘Algunas reflexiones sobre la interlengua de los alumnos en la PAU’, in: L.M. Manrique de Lara, R. Sobaina Romero and M. Sánchez Artiles (eds) 1as Jornadas de Reflexion y Debate sobre las Pruebas de Acceso a la Universidad. Las Palmas de Gran Canaria: Servicio de Publicaciones de la Universidad de las Palmas de Gran Canaria. 83-96. Wyatt, V. (1973), ‘An analysis of errors in composition writing’, ELT, 27(2): 177-186. Yoon, K.K. (1993), ‘Challenging prototype descriptions: perception of noun countability and indefinite vs. zero article use’, IRAL, 31(4): 269-289. Young, R. (1996), ‘Form-function relations in articles in English interlanguage’, in: R. Bayley and D.R. Preston (eds) Second Language Acquisition and Linguistic Variation. Amsterdam: John Benjamins Publishing Company. 135-175. Zamparelli, R. (2005), ‘The structure of (in)definiteness’, Lingua, 115: 915-936. Žegarac, V. (2004), ‘Relevance theory and the in second language acquisition’, Second Language Research, 20(3): 193-211.

Participle clauses in learner English: the role of transfer Christelle Cosme Centre for English Corpus Linguistics F.N.R.S. – Université catholique de Louvain Abstract Participle –ing and –ed clauses – both in their adnominal and adverbial function – have been shown to be underused in the academic writing of advanced French- and Dutchspeaking EFL learners (Granger 1997). This feature contributes to the stylistic deficiency of learner essays. It is argued in this paper that one of the reasons for the underuse might be transfer-related. English is assumed to have a predilection for participle structures (Kortmann 1995). French and Dutch, by comparison, are less keen on using such constructions and usually favour other alternatives. The claim that English makes more intensive use of participle constructions than French and Dutch is submitted to close examination of authentic multilingual data. It is not argued, however, that transfer is the only reason for the underuse. The influence of other factors (notably developmental and teaching-induced) should also be investigated.

1.

Introduction

The focus of the present article is on participle clauses, both present and past. Functionally speaking, the emphasis is placed on participle clauses that function adverbially and adnominally (i.e. as noun modifications). Concretely, four clause types are investigated, viz. adverbial present participle clauses (1), adverbial past participle clauses (2), adnominal present participle clauses (3), and adnominal past participle clauses (4).1 (1) (2)

(3) (4)

Most Iraqi policemen are on the streets, directing traffic and guarding buildings. (T) When pressed for evidence, the administration shelters behind the familiar and convenient mantra of how sensitive intelligence sources must be protected. (T) But in a world looking to science for answers, the bond of trust between practitioners and the public needs strengthening. (T) On another map, he points out the series of canals built with astonishing speed in the early 1990s. (T)

The aim of the research is twofold. To start with, the paper examines the use of participle clauses in learner English, relying heavily on previous research undertaken among others by Granger (1997). At a subsequent stage, the learner corpus study is supplemented with a corpus-based contrastive study of participle

178

Christelle Cosme

clauses in the three languages under investigation, namely English, French and Dutch. The objective of this second phase is to establish whether the use of participle clauses by EFL (English as a Foreign Language) learners could be somehow related to cross-linguistic differences between the learners’ L1 and their L2. Learner corpora and multilingual corpora are thus viewed in this paper as being closely interrelated types of data, following Granger’s (1996) Integrated Contrastive Model (ICM). It is believed that the investigation of multilingual corpora can not only predict some learners’ errors, but also help explain some interlanguage features. 2.

Participle clauses in learner English

A number of studies have shown that non-finite clauses – of which participle clauses form an integral part – are significantly underused in learner English, i.e. in texts produced by EFL learners from a variety of mother tongue backgrounds.2 Parrott (2000: 371) observes in this regard that “[i]f we compare a few pages of writing by learners and by native speakers, or a few minutes’ worth of their speech, we usually find that learners use non-finite clauses far less than native speakers”. In a similar trend, Hinkel (2002: 75) points to a clear underuse in advanced non-native speakers’ essays (produced by speakers of six languages, viz. Chinese, Japanese, Korean, Vietnamese, Indonesian and Arabic) of what she calls “reduced adjective clauses” and “reduced adverb clauses”, thereby referring to non-finite adnominal and adverbial clauses. The real problem, Parrott (2000: 371) argues, is that the avoidance of appropriate structures such as non-finite clauses often escapes the attention of teachers, who are generally more concerned with lexical, spelling, word order or tense errors than with style. In her investigation of the use of participle clauses in native vs. learner English, Granger (1997) observes a significant underuse of participle clauses in the academic writing of advanced French-, Dutch- and Swedish-speaking EFL learners.3 Careful comparison between native and learner data shows that there are twice as many participle constructions in the native corpus as in the learner corpus. Moreover, this underuse seems to be characteristic of the three categories of learners under investigation in Granger’s (1997) study. Closer inspection of the corpus data shows that the underuse concerns both –ing clauses and –ed clauses, whether in their adverbial or in their adnominal functions. In her corpus study, Granger makes a distinction between three types of participle clauses, viz. adnominal participle clauses, both –ed (5) and –ing (6), adverbial participle clauses, which can be either conjunction-headed (7) or preposition-headed, the most frequent preposition being ‘by’ (8), and supplementive clauses (i.e. adverbial clauses that are not introduced by any linkword), both –ed (9) and –ing (10).

Participle clauses in learner English: the role of transfer (5) (6) (7) (8) (9)

(10)

179

Home-start is the second of the two projects supported by the Times Christmas Appeal. (T) They might not issue press releases outlining their preferences but it does not take much imagination to assess their intentions. (T) This is the moment that Tony Blair has feared ever since making his ever-so-slightly humiliating round-the-world trip […]. (T) The braver, effective course is to deny them cause and succour by defusing wider grievances and thereby disarming them. (T) Pressed about claims by Robin Cook, the former foreign secretary, that Tony Blair had known that there was no threat from weapons of mass destruction before the outbreak of war, Mr Straw said that his predecessor “was simply and completely wrong about that”. (T) Countries as respectable as Canada have failed to submit names of operators and groups to the UN suspect list, inhibiting global freezing of their assets. (T)

That the underuse concerns all types of participle clauses clearly emerges from Table 1, which is a summary of Granger’s corpus findings. However, past participle supplementive clauses are an exception to the rule, as they turn out to be equally frequent in the learner corpus (NNS) and in the native corpus (NS). Note that the two corpora (NS vs. NNS) have a similar size, which makes it possible to compare the raw frequencies of each type of participle clause. Table 1. Frequency of participle clauses in the native corpus (NS) vs. in the learner corpus (NNS) (Granger 1997) Adnominal participle clauses –ed –ing Adverbial participle clauses conjunction-headed preposition-headed Supplementive clauses –ed –ing Total

NS 226 133 93 182 39 143 67 15 52 475

NNS 127 79 48 75 18 57 41 14 27 243

As indicated in Table 1, adnominal participle clauses represent the largest category of participle constructions. Such clauses appear to be significantly underused by advanced EFL learners, with about twice as many occurrences in the native corpus (NS) as in the learner corpus (NNS).4 In addition, Granger (1997) observes that while adnominal –ing clauses are consistently underused by the three categories of learners, the underuse of adnominal –ed clauses is largely due to a very low frequency of such clauses in the Dutch learner corpus. As

180

Christelle Cosme

regards adverbial participle clauses, Granger’s findings suggest that they represent the most significantly underused category of participle constructions, with about 2.5 times as many occurrences in the native corpus as in the learner corpus. As shown in Table 1, the underuse is most marked in the case of preposition-headed adverbial participle clauses. The third category of participle structures, namely supplementive clauses, appears to be underused by the three types of learners. As stated above, the underuse is however only manifest in the case of –ing supplementive clauses. According to Granger (1997: 194), the latter present “a very flexible way of structuring information and learners would benefit from increased mastery of this technique”. Granger’s (1997) study has obvious implications for EFL instruction and teaching materials. According to Granger (ibid. 195), the underuse of participle clauses – and non-finite clauses in general – undoubtedly contributes to the stylistic deficiency of learner essays. The lack of integrated style characteristic of learner essays makes the message fairly diluted and not 100% effective. There are, in Granger’s (ibid. 185-186) view, two reasons why non-finite clauses in general – and participle clauses in particular – should figure more prominently in EFL grammars, especially those targeted at advanced learners, and why they should be given more weight in teaching. First, as pointed out by Beaman (1984) and by Chafe & Danielewicz (1987), such clauses are particularly frequent in academic writing as discourse structuring strategies. Second, they constitute, as argued by Greenbaum (1988), a major means of syntactic compression, thereby allowing learners to reach a more compact, integrated style. Thus, finite clauses can often easily be abbreviated to non-finite or even verbless clauses. Granger (1997: 195) concludes that “EFL/ESL grammars and writing textbooks should highlight the role played by participle clauses both as a means of syntactic compression and as a discourse structuring device”. Granger (1997: 188ff.) mentions three possible reasons for the underuse. To start with, the insufficient treatment of participle clauses in EFL grammars may be responsible for the avoidance of such clauses in EFL essay writing. According to Granger (ibid. 185), participle clauses are insufficiently described, scattered across several sections of the grammar and often presented as free variants of finite clauses, with no mention of any factors favouring one type over the other.5 A second reason for the underuse might be seen in “the spectre of the dangling participle” (ibid. 188), i.e. the fear learners might experience of using stylistically awkward constructions of the type exemplified in (11), where the subject of the participle clause (in this case, the animate being “she”) does not correspond to that of the matrix clause (in this case, “the dog”). (11)

? Being blind, the dog guided her across the street. (Granger 1997: 187)

Lastly – and this is directly relevant to the issue discussed in this study – Granger (ibid. 189) mentions yet another reason for the underuse, namely transfer from the mother tongue. Kortmann (1995: 229) claims that, compared to other languages, English has an obvious predilection for participle constructions, especially

Participle clauses in learner English: the role of transfer

181

adverbial ones. Translations, Granger (1997: 189) argues, are the very embodiment of this cross-linguistic contrast, as “when one attempts to translate the English participle clause types into the other three languages under investigation, i.e. French, Swedish and Dutch, several of the English structures turn out to be either inexistent or much less frequently used”. There is therefore ample evidence to suggest that “learners underuse some of the participle structures because they have a more restrictive use in their mother tongue” (Granger 1997: 189).6 However, a word of caution is needed since, as rightly warned by Granger (1997: 189) herself, “in the absence of reliable text-based crosslinguistic descriptions of these structures, one should be careful not to draw hasty interlingual conclusions”. The aim of the study presented in this paper will therefore be to provide these text-based cross-linguistic descriptions. In an attempt to back Granger’s (1997) intuition, the remainder of this article will thus be devoted to a full-fledged contrastive analysis of the use of participle clauses in three languages, namely English, French and Dutch. 3.

Participle clauses in a corpus-based contrastive perspective

3.1

Data and methodology

The corpus analysis involves two different types of data. To start with, a number of observations are made on the basis of comparable data, i.e. original texts in two or more languages that are matched by criteria such as time of composition and register (see e.g. Johansson & Hasselgård 1999 and Granger 2003b for a detailed discussion of comparable corpora). The comparable data come from the TRILLED corpus (TRILingual Louvain EDitorials corpus), currently being compiled by researchers at the Centre for English Corpus Linguistics (University of Louvain, Belgium), and made up of original newspaper editorials in English, French and Dutch.7 At its present state of development, it contains c. 1 million words of English editorials, c. 750,000 words of French editorials and c. 500,000 words of Dutch editorials. For the purposes of the research, we extracted a random sample of 300 sentences in each language.8 It is believed that a careful analysis of these 300 sentences per language (i.e. all in all, 900 sentences) should provide useful insights into the frequency of participle clauses in the three languages under investigation. To complement the comparable corpus findings, use was also made of a trilingual translation corpus called PLECI (Poitiers-Louvain Échange de Corpus Informatisés) with a view to investigating how English participle clauses are translated into French and Dutch. The corpus, which is currently being compiled by the Centre for English Corpus Linguistics (University of Louvain, Belgium) and the University of Poitiers (France), is made up of fiction and newspaper articles. At its present state, it comprises c. 600,000 words of English texts translated into French, c. 700,000 words of French texts translated into English

182

Christelle Cosme

and c. 200,000 words of English texts and their respective Dutch translations. Note that all the translations included in PLECI have been produced by professional translators translating into their mother tongue. Because of the difficulties connected with the automatic retrieval of the structures under investigation, we decided to use only a sample of PLECI, namely c. 25,000 words from the English Ÿ French fiction component and c. 25,000 words from the English Ÿ Dutch fiction component.9 Detailed investigation of the translation corpus data should ultimately help us to gain a greater understanding of the different alternatives offered by French and Dutch to render English participle clauses. Broadly speaking, the translation corpus analysis is an attempt to provide more qualitative insights into the way each of the three languages typically packages information. The study thus combines the respective strengths of comparable and translation data. Johansson & Hasselgård (1999: 146) are among the first to have emphasised the benefit of a conjoined use of comparable and translation corpora in cross-linguistic research, thus taking advantage of the specific merits of both types. This principle is closely followed in this article. The basic assumption is that comparable and translation data are not mutually exclusive. Rather the contrary: they can supplement one another. The two types of data are therefore felt to be necessary to gain a deeper understanding of the use of participle clauses in the three languages under investigation. While lexis lends itself perfectly to corpus investigations, the same does not always hold for syntactic phenomena, the chief reason being that syntactic structures are usually more difficult to extract from a corpus than simple words (Gilquin 2002: 183). Because of the difficulty connected with the automatic retrieval of macro-linguistic patterns, most contrastive studies dealing with syntax or discourse are still to a large extent dependent on intuition, or resort to corpus examples only to illustrate or support the claims being made. It must, however, be acknowledged that, though it is true that lexis has the lion’s share in present-day contrastive linguistics, much effort has been made in recent years to fill the gap between lexical contrastive studies and syntactic or discursive ones.10 Though it must be admitted that the use of fully automatic methods is generally impossible in syntactic or discursive contrastive research, the view held in this article is that it is possible to automate the process to some extent. Computerised methods are thus adopted whenever technically feasible. Thus, in order to ease the retrieval of the structures under study, it was decided to provide both the comparable and the translation corpus data with linguistic annotations using an automated method. Concretely, sentence-level and clause-level tags were inserted at the appropriate places in the comparable corpora with the help of a purpose-built XML-compliant toolbox called (Watrin 2002). As for the translation corpus data, they were enriched with annotations regarding the types of shifts they involve (e.g. a shift from an –ing clause in English to a finite relative clause in French or Dutch). The inserted tags could subsequently be retrieved easily with the help of concordancers – monolingual in the case of comparable data and bilingual in the case of translation data. Providing greater detail on the annotation process falls

Participle clauses in learner English: the role of transfer

183

beyond the scope of the present article. For more information on the annotation of comparable and translation data, the reader is referred to, respectively, Cosme (2003) and Cosme (2006). 3.2

Corpus findings

3.2.1 Preliminary observation Close examination of the overall frequency of participle clauses in the comparable data yields somewhat unexpected results. With a total of 103 occurrences, the French subcorpus appears to contain a considerably larger number of participle clauses than the English subcorpus (which contains 77 occurrences). In the Dutch subcorpus, the frequency of participle clauses drops sharply to a mere 15 occurrences. The strikingly low number of participle clauses in Dutch is accurately reflected in the translation corpus data. As shown in Figure 1, English participle clauses are fairly seldom rendered by participial constructions into Dutch (only 113 cases out of 425, i.e. 27% of the cases). Put differently, they are most of the time (i.e. in the remaining 73% of the cases) translated by means of something other than a participle clause in Dutch (i.e. a non-participial construction), e.g. a finite relative clause or a coordinate construction.

350

312

300 250 200

211 188

150

participial 113

non-participial

100 50 0 EN => FR

EN => DU

Figure 1. Frequency of participial vs. non-participial constructions used to translate English participle clauses into French and Dutch (PLECI) The fairly large number of participle clauses in French as compared to their frequency in English would lead us to expect a high percentage of participial constructions in the French translation corpus. This, however, is not the case. As indicated in Figure 1, English participle clauses are not so frequently rendered by

184

Christelle Cosme

participle clauses into French (only 188 cases out of 399, i.e. 47% of the cases). Though this percentage is higher than in Dutch translations (where it amounts to a mere 27%), it is admittedly lower than would have been expected on the basis of the comparable corpus findings. This somewhat surprising finding is, however, best explained by the very nature of the corpus data under investigation. It should indeed be borne in mind that although they present the major advantage of being “linguistic contrast come alive” (Mauranen 1999: 181), translation corpus data are usually affected by a series of effects related to the very activity of translating. Hence, great caution is needed in interpreting the results of a translation corpus study. In his/her reflection, the translator is for instance faced with a series of alternatives in the target language to render a specific construction present in the source text. S/he has to select, among these potential alternatives, the one that best suits the context. In some cases, s/he might thus prefer a non-participial construction over a participial construction, even though a participial construction would have been grammatically and stylistically acceptable. This might explain the fact that the proportion of participial constructions in French translations is slightly lower than would have been expected. Furthermore, the comparable corpus finding that participle clauses are more frequent in French than in English should be somewhat qualified. Closer inspection of the corpus data indeed shows that this does not apply to all types of participle clauses. In other words, considering all types of participle clauses as a whole without making finer distinctions might somehow skew the results. More fine-grained analysis is therefore needed in order to gain a deeper understanding of the use of participle clauses in the three languages. 3.2.2 The four types of participle clauses in English, French and Dutch As stated above, we distinguish between four types of participle clauses, viz. (a) adnominal present participle clauses, (b) adverbial present participle clauses, (c) adnominal past participle clauses, and (d) adverbial past participle clauses. As is the case in most grammars (e.g. Quirk et al. 1985: 1123ff.), Granger’s (1997) category of supplementive clauses (see above) is included in the adverbial category. Careful analysis of the frequency of the four distinct types of participle clauses in the three languages reveals the figures given in Table 2 – and represented graphically in Figure 2. A first empirical observation that can be made on the basis of Table 2 and Figure 2 is the low frequency of adverbial past participle clauses in the three languages. Thus, examples of the type illustrated in (12), (13) and (14) are hardly ever found in the corpus data. In view of their scarcity, adverbial past participle clauses will be excluded from the following analysis. (12)

Pressed about claims by Robin Cook, the former foreign secretary, that Tony Blair had known that there was no threat from weapons of mass destruction before the outbreak of war, Mr Straw said that his predecessor “was simply and completely wrong about that”. (T)

Participle clauses in learner English: the role of transfer

185

Table 2. Frequency and distribution of the four types of participle clauses in English, French and Dutch (TRILLED)

Present adnominal Present adverbial Total present Past adnominal Past adverbial Total past

ENGLISH # % 21 42.9 28 57.1 49 100.0 24 85.7 4 14.3 28 100.0

FRENCH # % 10 33.3 20 66.7 30 100.0 71 97.3 2 2.7 73 100.0

DUTCH # 2 1 3 11 1 12

% 66.7 33.3 100.0 91.7 8.3 100.0

80 70 60 50

English

40

French

30

Dutch

20 10 0 Adn. pres. Adv. pres. Adn. past Adv. past

Figure 2. Frequency of the four types of participle clauses in English, French and Dutch (TRILLED) (13)

(14)

Muni d’un micro relié à un haut-parleur, un homme vêtu d’une djellaba brune et la tête ceinte d’un turban immaculé appelait fermement la foule à se disperser. (T) (‘Equipped with a microphone connected to a loudspeaker, a man wearing a brown jellaba and with an immaculate turban round his head urged the crowd to scatter’) Vergeleken bij de serie bomaanslagen waarmee de ramadan maandag begon, is het op een paar explosies van aanvallen op Amerikaanse troepen na rustig in de hoofdstad. (T) (‘Compared with the series of bomb attacks with which the Ramadan began on Monday, the capital is peaceful with

186

Christelle Cosme the exception of a few explosions resulting from attacks on American troops’)

It also emerges from Table 2 and Figure 2 that present participle clauses – whether adnominal or adverbial – occur with an amazingly low frequency in the Dutch subcorpus (3 occurrences only). Thus, compared to English and French, constructions of the type exemplified in (15) and (16) are very rare in the Dutch subcorpus. As for English and French, they appear to exhibit striking differences in their use of present participle clauses, the former making more intensive use of such constructions than the latter (49 vs. 30 occurrences). Moreover, it turns out that the observed contrast is true for both adnominal present participle clauses and adverbial ones.11 Thus, English participial constructions of the type illustrated in (17) and (18) are employed more often than their French counterparts like those exemplified in (19) and (20). (15)

(16)

(17)

(18) (19)

(20)

Het vorige kabinet heeft zich in principe akkoord verklaard met vergaande voorstellen om politie en justitie recht te geven op persoonsinformatie die berust bij een breed scala aan bronnen, variërend van bank tot sportvereniging. (T) (‘The former cabinet has agreed on extreme proposals to grant the police and the judicial system access to personal information that is the property of a wide variety of sources, ranging from banks to sports organisations’) “Ik woonde negen jaar in een huurwoning in Chamchamal”, vertelt de vrouw in het stadion, doelend op een plaats net voorbij de voormalige grens tussen Saddams Irak en het vrije noorden. (T) (‘“I’ve been living in a rented flat in Chamchamal for nine years,” says the woman in the stadium, pointing to a place a bit further than the former border between Saddam’s Iraq and the free north’) At the root of the current instability are the very people most Iraqis reject the remnants of Saddam’s Baath party, and extremists flooding in from neighbouring countries in hope of establishing religious rule. (T) Al-Qaida’s cells are taking measures against increasing electronic surveillance, operating semi-autonomously […]. (T) A New York, hier, l’assemblée générale des Nations unies discutait en séance extraordinaire d’un projet de résolution demandant à la Cour internationale de justice de se prononcer sur la légalité de la construction par Israël d'un “mur” en Cisjordanie. (T) (‘Yesterday in New York, the UN General Assembly met in an extraordinary session to discuss a resolution project asking the International Court of Justice to reach a verdict on the legality of a “wall” built by Israel in the West Bank’) Retenant ses larmes, son visage buriné marqué par la souffrance, Mohammad Tabaza regarde quatre jeunes sympathisants masqués du Hamas écrivant à la bombe de peinture sur le mur le nom de son fils de 23 ans, tué dans le raid. (T) (‘Holding back his tears, his craggy face a

Participle clauses in learner English: the role of transfer

187

testimony of his suffering, Mohammad Tabaza is watching four young masked Hamas activists who are paint spraying the wall to write the name of his 23-year-old son killed in the raid’) The observed cross-linguistic differences in terms of frequency of present participle clauses – both adnominal and adverbial – appear to be directly reflected in the translation corpus data. As such constructions have been shown to occur with a higher frequency in English than in French and especially Dutch, we may expect that English present participle clauses (i.e. so-called –ing clauses) will not likely give rise to word-for-word translations into French, let alone into Dutch. In other words, we may expect the translation corpus findings to show a relatively small proportion of cases where English present participle clauses are simply preserved in French or Dutch. These expectations are confirmed by the translation corpus analysis. Indeed, Figure 3 shows that English –ing clauses are translated by non-participial constructions into French in the majority of the cases (in exactly 163 cases out of 273, i.e. in 60% of the cases). As expected as well, the proportion of non-participial constructions is considerably larger (84%, i.e. 250 cases out of 299) when it comes to translating English –ing clauses into Dutch.

300 250 250 200 150

163

participial non-participial

110

100 49 50 0 EN => FR

EN => DU

Figure 3. Frequency of participial vs. non-participial constructions used to translate English –ing clauses into French and Dutch (PLECI) In other words, English –ing clauses are – in the case of translation into French – not very often (i.e. in 40% of the cases) and – in the case of translation into Dutch – hardly ever (i.e. in a mere 16% of the cases) rendered by means of similar participial constructions. This finding suggests that French and Dutch must have alternative ways of conveying the meaning expressed by English present participle clauses. A look at the translation corpus data reveals recurrent patterns of correspondence between English adnominal –ing clauses and finite relative

188

Christelle Cosme

clauses in French and Dutch (21). As for adverbial –ing clauses, they turn out to be very often translated by means of prepositional phrases in French and coordinate constructions in Dutch (22). (21)

(22)

[…] and at the earliest possible moment join that ship providentially awaiting him at Vera Cruz. Ÿ […] et le plus tôt possible rejoignit ce navire qui providentiellement l’attendait à Vera Cruz. (P) (‘… that ship that was providentially awaiting him at Vera Cruz’) Ÿ […] om zo snel mogelijk op het schip te zien te komen dat, als door de Voorzienigheid gestuurd, in Vera Cruz op hem wachtte. (P) (‘… that ship that, as if sent by Providence, was awaiting him at Vera Cruz’) Ralph spoke to himself, sounding the bass strings of delight. Ÿ Ralph poussa une exclamation de plaisir sur un ton de basse. (P) (‘Ralph gave an exclamation of delight in a bass voice’) Ÿ Ralph praatte in zichzelf en bracht de bassnaar van verrukking in trilling. (P) (‘Ralph spoke to himself and his bass voice quavered with delight’)

Another interesting finding emerging from the comparable corpus data (see Table 2 and Figure 2 above) is the incredibly high frequency of adnominal past participle clauses in French (71 occurrences), as compared to English and Dutch (displaying 24 and 11 occurrences, respectively). Of the four distinct types of participle clauses under examination, adnominal past participle clauses are the only ones to be more frequent in French than in English, the other three types being more frequent in English than in French. Thus, cases like (23) appear to occur with a markedly higher frequency than their English (24) and Dutch (25) counterparts. (23)

(24) (25)

Un chantier estimé par les Nations unies et la Banque mondiale à 56 milliards de dollars pour 2004-2007, dont 17,5 milliards pour la seule année prochaine. (T) (‘A building site estimated by the United Nations and the World Bank at 56 billion dollars for 2004-2007, of which 17.5 billion dollars account for next year only’) Such an outcome, they assert, echoing the theme initiated by Ernst and Young at the weekend, would seriously destabilise public finances. (T) Ook in Irak zelf kampt de raad van 24 door de VS benoemde leden met een geloofwaardigheidsprobleem. (T) (‘In Iraq too, the board of 24 members appointed by the United States is having to struggle against a credibility problem’)

It thus follows from the analysis of the four types of participle clauses that the somewhat unexpected higher frequency of participle clauses in French compared with English is largely attributable to the large number of adnominal past participle clauses in French. As was the case with present participle clauses, the

Participle clauses in learner English: the role of transfer

189

observed cross-linguistic differences in terms of frequency of adnominal past participle clauses appear to be manifest in the translation corpus data as well. As such syntactic structures have been shown to be more frequent in French than in English and yet more frequent in English than in Dutch, we may postulate that English adnominal past participle clauses (i.e. so-called adnominal –ed clauses) will more likely be rendered by word-for-word translations (i.e. translations by means of adnominal past participle clauses) in French than in Dutch. Close inspection of the translation corpus data shows that this is the case. As indicated in Figure 4, English adnominal –ed clauses are quite often rendered by formally equivalent constructions into French (53 cases out of 86, i.e. 62% of the cases). It appears to be less frequently the case when it comes to translating English adnominal –ed clauses into Dutch, formally equivalent constructions being resorted to in only about half of the cases (49% exactly).

60

53

50 40

45

46

33

participial

30

non-participial

20 10 0 EN => FR

EN => DU

Figure 4. Frequency of participial vs. non-participial constructions used to translate English adnominal –ed clauses into French and Dutch (PLECI) Adnominal past participle clauses being about three times as frequent in French as in English (see Table 2 above), one would however have expected a larger percentage of participial constructions in French translations. In other words, 62% seems surprisingly low in view of the comparable corpus findings. This unexpectedly low percentage again highlights the great care which should be taken in interpreting translation corpus findings, translated texts being influenced by a series of complex concurrent factors. What matters here is that, as shown in Figure 4, the proportion of formally equivalent constructions is higher when translating into French than when translating into Dutch. Careful examination of the translation corpus data reveals that, as is the case in example (26), Dutch very often favours finite alternatives over adnominal past participle clauses.

190

Christelle Cosme

(26)

Down at street level another poster, torn at one corner, flapped fitfully in the wind. Ÿ Beneden ter hoogte van de straat flapte een ander biljet, dat aan één van de hoeken was ingescheurd, stuipachtig in de wind. (P) (‘… another poster, which was torn at one corner, …’)

4.

Linking up learner and contrastive research: the role of transfer

The cross-linguistic descriptions provided above should ultimately help us to test Granger’s (1997) transfer-hypothesis. The present section should thus be viewed as an attempt to link the learner corpus study carried out by Granger to the contrastive corpus study conducted in this paper.12 Note that, unlike in Granger’s study, no mention will be made of the underuse of participle clauses by Swedishspeaking EFL learners, as Swedish was not included in the cross-linguistic study conducted in this paper. The research question addressed in this section thus reads as follows: can the underuse of participle clauses by French- and Dutch-speaking EFL learners observed by Granger (1997) be accounted for by cross-linguistic influence? In other words, is it likely that French- and Dutch-speaking learners avoid using participle clauses on the grounds that such constructions occur with a lower frequency in their mother tongue (i.e. either French or Dutch) than in English? In order to assess the role of transfer in the underuse, it was necessary to determine the exact nature of the differences between English, French and Dutch. This was achieved by conducting a full-fledged contrastive study of participle clauses in the three languages. A first fairly general finding of the corpus-based contrastive study is that there is a larger quantitative discrepancy between English and Dutch than between English and French. This may have an implication for learner language. L1-L2 discrepancy being a potential cause of avoidance, we may expect the underuse to be more marked in the case of Dutch-speaking EFL learners than in the case of French-speaking EFL learners. Close examination of Granger’s (1997: 188) corpus findings confirms the prediction, as the underuse is slightly more marked in the Dutch learner corpus than in the French learner corpus. A more in-depth study of the use of the different types of participle clauses in the three languages enables us to make more accurate predictions regarding the use of participle clauses by French- and Dutch-speaking learners. Table 3 makes the link between the cross-linguistic findings and expected learner performance. A first major finding concerns the use of present participle clauses – both in their adnominal and adverbial functions – in the three languages. As appears from Table 3, such constructions have been shown to occur with a considerably lower frequency in Dutch (3 occurrences) than in French (30 occurrences) and especially English (49 occurrences). As a result, we may expect French- and Dutch-speaking EFL learners to underuse such structures, which is confirmed by Granger’s (1997) findings. In addition, present participle constructions being particularly scarce in Dutch, the underuse may be expected to be more marked in

Participle clauses in learner English: the role of transfer

191

the Dutch learner corpus. This, however, does not emerge from Granger’s study (ibid. 188), which shows that –ing clauses are equally underused by both categories of learners. Thus, the expected difference in terms of degree of underuse of present participle clauses by French- and Dutch-speaking learners is not verified. But the actual underuse of present participle clauses by both categories of learners is testified by Granger (1997). Hence, it may be argued that transfer is likely to play a role in the underuse of present participle clauses by both Dutch- and French-speaking EFL learners. The use of such constructions being apparently more restricted in French and Dutch than in English, the temptation may be strong for Dutch- and French-speaking learners to avoid using them in their essay writing. The translation corpus analysis has revealed a number of frequent alternatives to English present participle clauses. It is thus highly probable that French- and Dutch-speaking learners will often resort to these alternatives. Instead of using adnominal –ing clauses (27), for instance, learners – both French- and Dutch-speaking ones – may be tempted to use finite relative clauses (28). Likewise, we may expect Dutch-speaking learners to use, in lieu of adverbial present participle clauses (29), coordinate structures of the type illustrated in (30). (27) (28) (29) (30)

[…] and at the earliest possible moment join that ship providentially awaiting him at Vera Cruz. (P) […] and at the earliest possible moment join that ship that providentially awaits him at Vera Cruz. Ralph spoke to himself, sounding the bass strings of delight. (P) Ralph spoke to himself and sounded the bass strings of delight.

Table 3. Summary of cross-linguistic findings and expected learner performance Cross-linguistic findings Present participle clauses (adnominal and adverbial) Adnominal past participle clauses > >>

EN > FR >> DU

Expectations for learner language French-speaking Dutch-speaking EFL learners EFL learners Underuse Great underuse

FR >> EN > DU

No underuse

Underuse

= uses a larger number of x than = uses a much larger number of x than

The second major finding of the corpus-based contrastive study relates to the use of adnominal past participle clauses. As indicated in Table 3, such constructions have been shown to be markedly more frequent in French (71 occurrences) than in English (24 occurrences) and especially Dutch (11 occurrences). In view of these results, we may predict a marked underuse of

192

Christelle Cosme

adnominal past participle clauses by Dutch-speaking EFL learners. The prediction is confirmed by Granger (1997: 188), who observes that the underuse of adnominal –ed clauses by EFL learners is especially noticeable in the Dutch learner corpus. Thus, transfer most likely plays a role in the underuse of adnominal past participle clauses by Dutch-speaking EFL learners. The use of such clauses being apparently fairly restricted in Dutch, Dutch-speaking learners may avoid using them (31) and resort more often to, for instance, finite relative clauses (32) – which have been shown to be frequent alternatives in translating English participle clauses into Dutch. (31) (32)

Down at street level another poster, torn at one corner, flapped fitfully in the wind. (P) Down at street level another poster, which was torn at one corner, flapped fitfully in the wind.

In the case of French-speaking EFL learners, the situation is more complex. The corpus-based contrastive study shows that French makes more intensive use of such constructions than English and that these constructions are fairly often maintained in French translations. As a result, if transfer is to play a role in the use of adnominal past participle clauses by French-speaking EFL learners, it is more likely to be made manifest in a significant overuse than in a significant underuse such as that observed by Granger (1997). Hence, while transfer may well play a part in the underuse of adnominal –ed clauses by Dutch-speaking learners, it is a most unlikely explanation for the underuse of such constructions by French-speaking learners. To sum up, it is argued in this paper that transfer may be one key factor in the underuse of all types of participle clauses by Dutch-speaking EFL learners. The question of whether transfer can play a part in the underuse of participle clauses by French-speaking EFL learners is, however, more complex to answer. In the case of French-speaking learners, it appears that a distinction should be made between present participle clauses – both adnominal and adverbial – and adnominal past participle clauses. While transfer may contribute to the underuse of the former, it is less likely to play a role in the underuse of the latter. We do not argue, however, that transfer (and, more particularly, the phenomenon of avoidance due to L1-L2 discrepancy) is the one and only explanation for the underuse. Other factors may also underlie the underuse (see e.g. Jarvis 2000, Odlin 2003). Some factors may even be more dominant than cross-linguistic influence. One key factor is the learners’ degree of maturity (cf. Sampson 2003). Participle clauses being features of integrated discourse (see Chafe’s [1982] distinction between integrated and fragmented style) and of syntactic maturity, they are not abundant in novice writers’ texts. Task settings may also have a profound influence. Ädel (this volume), for instance, shows that learners’ performance varies greatly according to whether the tasks are to be completed under timed or untimed conditions. Participle clauses being indices of integrated discourse and of great syntactic maturity, producing them inevitably

Participle clauses in learner English: the role of transfer

193

takes time. Hence, learners’ performance in actually producing such clauses may be expected to be better in untimed than in timed conditions.13 5.

Conclusion

Through the emphasis it places on the role of multilingual data in learner corpus research, this paper aims to contribute to one of the most important methodological aspects of learner corpus research. In line with Granger’s (1996) Integrated Contrastive Model, we argue that the role of multilingual data in learner corpus research is twofold. Not only can they help predict interlanguage features, but they are also of great help in actually explaining learners’ performance. The central focus of the paper has been largely on the second, i.e. explanatory, function (cf. Granger’s [1996] diagnostic approach). In other words, the starting point was Granger’s (1997) observed underuse of participle clauses by both French- and Dutch-speaking EFL learners, and the aim of using multilingual data – both comparable and translation data – was to provide a plausible explanation for the underuse. The fundamental assumption was that the underuse might be due to cross-linguistic influence (or transfer): learners may underuse participial constructions simply because such structures are used less frequently in their mother tongue. In order to test this hypothesis further, it was necessary to carry out a contrastive study of the use of participle clauses in English, French and Dutch. The corpus-based cross-linguistic analysis not only enabled us to establish transfer as one potential reason for the underuse of some participial constructions by EFL learners, but it also highlighted cases where transfer obviously cannot play a role (e.g. the underuse of adnominal past participle clauses by French-speaking learners). However, transfer is by no means the one and only explanation for the underuse. Too many studies have drawn hasty conclusions from observed L1-L2 discrepancies, assuming that crosslinguistic differences necessarily lead to “deviant” behaviour in learners’ performance. This study, however, takes a different stance. There seems to be a complex interplay of factors underlying the underuse, including syntactic maturity, task settings, and even teaching-induced factors. This interplay of factors may be the reason why, to use Jarvis’ (2000: 246) words, “no area of second language research has received as much attention and remained as elusive as the influence of the first language (L1)”. We argue that a carefully conducted corpus-based contrastive study can help, if not overcome, at least ease the difficulty involved in establishing transfer. Acknowledgements This study has been made possible thanks to the support of the Belgian National Fund for Scientific Research (Research Fellowship), which I gratefully acknowledge. I am also grateful to the three volume editors for their insightful comments on the article, and in particular to Gaëtanelle Gilquin for her precious

194

Christelle Cosme

help throughout the revision process. Special thanks also go to Jennifer Thewissen for helping me with the translations of the French and Dutch examples, as well as to Sylviane Granger for her valuable comments on earlier drafts of the article. Notes 1

Unless otherwise stated, the examples given in this paper are taken from two multilingual corpora, viz. a comparable corpus called TRILLED and a translation corpus called PLECI (see Section 3.1 for additional information on the two types of corpus data). Throughout the paper, the examples taken from TRILLED are signalled by means of (T), and those coming from PLECI by means of (P).

2

Note that besides being generally underused, non-finite clauses are also frequently misused. Parrott (2000: 372) observes, for instance, the use of –ing constructions instead of full infinitives (e.g. *We went to Woolworth’s for buying our spring bulbs), the use of the present participle instead of the past participle (e.g. *The train robbery, thinking to be the greatest of the century, took place in 1963), or even the use of the full infinitive instead of the present participle (e.g. *I watched them to dance).

3

The native control corpus used by Granger (1997) is the Louvain Corpus of Native English Essays (LOCNESS). As for the learner data, they consist of argumentative essays from the International Corpus of Learner English (ICLE). The reader is referred to the following URL for additional information on the two corpora: http://cecl.fltr.ucl.ac.be/CeclProjects/Icle/icle.htm (last accessed on December 11, 2006). See also Granger (2003a) for greater detail.

4

Granger (1997) also observes a high degree of misuse of adnominal participle clauses by EFL learners. A number of adnominal participle clauses were found infelicitous by a native speaker (i), who replaced them with finite relative clauses (ii). (i) ? Not to mention these ethnical minorities not feeling at home. (Granger 1997: 190) (ii) Not to mention these ethnical minorities who don’t feel at home. (Granger 1997: 190)

5

Granger (1997) argues, however, that Johansson & Lysvåg’s (1986) grammar is a major exception to the rule, as it presents a very detailed description of participle structures in two chapters, one on non-finite expansion of the sentence and the other on –ing constructions. That participle clauses are very often presented as free variants of finite clauses may be best exemplified by Downing & Locke’s (1992: 286) claim that

Participle clauses in learner English: the role of transfer

195

“[t]he non-finite verb forms –ing, to-infinitive and –ed participle are used non-restrictively to express the same meanings as the finite forms”. 6

Laufer & Eliasson (1993: 37) see L1-L2 discrepancy as one potential cause of avoidance. In their study of the use of phrasal verbs by Hebrewand Swedish-speaking EFL learners, they observe a significant underuse of such verbs by Israeli students. Swedish students, by contrast, do not seem to underuse such constructions. The authors argue that this finding may be linked to the inexistence of phrasal verbs in Hebrew and, in the case of Swedish EFL learners, to the abundance of phrasal verbs in Swedish. The absence of phrasal verbs in the learners’ L1 may thus lead Hebrew-speaking EFL learners to avoid such constructions in their L2. In the words of Kamimoto et al. (1992: 256), it appears that “while linguistic difference [can] still lead to difficulty (whatever that mean[s]), that difficulty might resolve itself in avoidance rather than in knee-jerk errormaking”.

7

Comparability in the TRILLED corpus is ensured by selecting editorials from quality newspapers in the three languages. English editorials are taken from The Guardian, The Independent, The Times, The Observer, The Sunday Telegraph, The Daily Telegraph and The Economist; French editorials from Le Monde, Le Figaro and Libération; and Dutch editorials from NRC Handelsblad, Trouw, Het Parool, Utrechts Nieuwsblad and Haagsche Courant.

8

After numbering each sentence in the three subcorpora (with the help of the numbering function available in Microsoft Word), we used a fairly simple random sampling Visual Basic program to extract 300 numbers from each subcorpus. Each number corresponding to a sentence, we thus ended up with a random sample of 300 sentences in each subcorpus.

9

The selection of the 25,000-word samples was made simply by taking the first few words of each novel contained in PLECI. The choice for fiction was primarily motivated by the higher frequency of participle clauses – especially present participle ones – in fiction than in journalistic texts (see e.g. Biber 1988, Kortmann 1991).

10

In the Scandinavian countries, for instance, a number of contrastive studies based on the ENPC (English-Norwegian Parallel Corpus) or on the ESPC (English-Swedish Parallel Corpus) have dealt extensively with macro-linguistic phenomena. The SPRIK project (Språk i kontrast / Languages in Contrast), for instance, has given rise to a series of syntactic and discursive contrastive studies (e.g. Fabricius-Hansen 1998). See http://www.hf.uio.no/forskningsprosjekter/sprik/english/ (last accessed on December 11, 2006) for additional information on the project.

11

Incidentally, English and French also appear to exhibit some differences as to the types of adverbial present participle clauses they favour. About two

196

Christelle Cosme thirds (14 occurrences out of 20, i.e. 70%) of the French adverbial present participle clauses are introduced by en, forming so-called “gérondifs” (i). By comparison, among the 28 adverbial present participle clauses found in the English subcorpus, only 8 are introduced by a linkword, be it a preposition (ii) or a conjunction (iii), the remaining being supplementive (iv). (i) Et l’attaché de presse de l’ambassade d’Espagne a été abattu en sortant de chez lui. (T) (‘And the press attaché of the Spanish Embassy was killed leaving his home’) (ii) David Blunkett, the Home Secretary, said it was vital that the need for security for staff overseas was balanced by giving them the scope to do their jobs properly. (T) (iii) The French and German leaders made their decision while attending a European Union summit in Brussels. (T) (iv) Al-Qaida’s cells are taking measures against increasing electronic surveillance, operating semi-autonomously […]. (T)

12

A word of caution is needed here, as the two studies (i.e. Granger’s [1997] study and our own cross-linguistic research) involve different types of data. Whilst Granger’s study relies on argumentative essays (from ICLE), our own research is based on editorials (for the comparable corpus study) and novels (for the translation corpus study). However, it is believed that this genre discrepancy does not hamper the comparison. So (2005: 74ff.), for instance, clearly demonstrates that there is a large degree of overlap between argumentative essays and editorials. The similarity between argumentative essays and novels is more difficult to establish. The choice for novels in the translation corpus has admittedly been heavily influenced by the availability of texts, translation corpora consisting of scientific articles being on the whole much scarcer than those consisting of fictional texts. Moreover, as already discussed above, fiction appears to be a suitable text type for the investigation of participle clauses. It is therefore argued that, despite the genre discrepancy, a translation corpus of novels provides useful information for the analysis, as it gives access to translation equivalents.

13

As pointed out by Ädel (this volume), most of the ICLE essays written by French- and Dutch-speaking learners were produced under untimed conditions. So, presumably, the underuse of participle clauses would have been even greater if more essays had been written under timed conditions.

References Beaman, K. (1984), ‘Coordination and subordination revisited. Syntactic complexity in spoken and written narrative discourse’, in: D. Tannen (ed.)

Participle clauses in learner English: the role of transfer

197

Coherence in Spoken and Written Discourse. Norwoord, N.J.: Ablex. 4580. Biber, D. (1988), Variation across Speech and Writing. Cambridge: Cambridge University Press. Chafe, W. (1982), ‘Integration and involvement in speaking, writing, and oral literature’, in: D. Tannen (ed.) Spoken and Written Language: Exploring Orality and Literacy. Norwood, NJ: Ablex. 35-53. Chafe, W. and J. Danielewicz (1987), ‘Properties of spoken and written language’, in: R. Horowitz and S.J. Samuels (eds) Comprehending Oral and Written Language. New York: Academic Press. 83-113. Cosme, C. (2003), Clause combining in English, French and Dutch. State of the art and pilot study: Towards annotating sentences and clauses. Unpublished MA dissertation, Université catholique de Louvain. Cosme, C. (2006), ‘“Spelling things out” in translations: A corpus-based study of explicitation’. Paper presented at the SLT-Congress (The Study of Language and Translation), Ghent (Belgium), 12-14 January 2006. Downing, A. and P. Locke (eds) (1992), A University Course in English Grammar. Hertfordshire: Prentice Hall. Fabricius-Hansen, C. (1998), ‘Informational density and translation, with special reference to German-Norwegian-English’, in: S. Johansson and S. Oksefjell (eds) Corpora and Cross-linguistic Research: Theory, Method, and Case Studies. Amsterdam & Atlanta: Rodopi. 197-234. Gilquin, G. (2002), ‘Automatic retrieval of syntactic structures. The quest for the Holy Grail’, International Journal of Corpus Linguistics, 7(2): 183-214. Granger, S. (1996), ‘From CA to CIA and back: an integrated approach to computerized bilingual and learner corpora’, in: K. Aijmer, B. Altenberg and M. Johansson (eds) Languages in Contrast. Papers from a Symposium on Text-based Cross-linguistic Studies. Lund 4-5 March 1994. Lund: Lund University Press. 37-51. Granger, S. (1997), ‘On identifying the syntactic and discourse features of participle clauses in academic English: native and non-native writers compared’, in: J. Aarts, I. de Mönnink and H. Wekker (eds) Studies in English Language and Teaching. In Honour of Flor Aarts. Amsterdam & Atlanta: Rodopi. 185-198. Granger, S. (2003a), ‘The International Corpus of Learner English: A new resource for foreign language learning and teaching and second language acquisition research’, TESOL Quarterly, 37(3): 538-546. Granger, S. (2003b), ‘The corpus approach: a common way forward for Contrastive Linguistics and Translation Studies’, in: S. Granger, J. Lerot and S. Petch-Tyson (eds) Corpus-based Approaches to Contrastive Linguistics and Translation Studies. Amsterdam & Atlanta: Rodopi. 1730. Greenbaum, S. (1988), ‘Syntactic devices for compression in English’, in: J. Klegraf and D. Nehls (eds) Essays on the English Language and Applied

198

Christelle Cosme

Linguistics on the Occasion of Gerhard Nickel’s 60th Birthday. Heidelberg: Julius Groos Verlag. 3-10. Hinkel, E. (ed.) (2002), Second Language Writers’ Text. London: Erlbaum. Jarvis, S. (2000), ‘Methodological rigor in the study of transfer: identifying L1 influence in the interlanguage lexicon’, Language Learning, 50(2): 245309. Johansson, S. and H. Hasselgård (1999), ‘Corpora and cross-linguistic research in the Nordic countries’, in: S. Granger, L. Beheydt and J.-P. Colson (eds) Contrastive Linguistics and Translation (special issue of Le Langage et l'Homme 34). Leuven: Peeters. 145-162. Johansson, S. and P. Lysvåg (eds) (1986), Understanding English Grammar. Vol. 1. Oslo: Norwegian University Press (Universitetsforlaget AS). Kamimoto, T., A. Shimura and E. Kellerman (1992), ‘A second language classic reconsidered – the case of Schachter’s avoidance’, Second Language Research, 8: 251-277. Kortmann, B. (1991), Free Adjuncts and Absolutes in English. Problems of Control and Interpretation. London & New York: Routledge. Kortmann, B. (1995), ‘Adverbial participle clauses in English’, in: M. Haspelmath and E. König (eds) Converbs in Cross-linguistic Perspective. Berlin & New York: Mouton de Gruyter. 189-237. Laufer, B. and S. Eliasson (1993), ‘What causes avoidance in L2 learning: L1-L2 difference, L1-L2 similarity, or L2 complexity?’, Studies in Second Language Acquisition, 15: 35-48. Mauranen, A. (1999), ‘Will ‘translationese’ ruin a contrastive study?’, Languages in Contrast, 2(2): 161-185. Odlin, T. (2003), ‘Cross-linguistic influence’, in: W. Doughty and M. Long (eds) The Handbook of Second Language Acquisition. Oxford: Blackwell Publishing. 436-486. Parrott, M. (ed.) (2000), Grammar for English Language Teachers. With Exercises and a Key. Cambridge: Cambridge University Press. Quirk, R., S. Greenbaum, G. Leech and J. Svartvik (1985), A Comprehensive Grammar of the English Language. London: Longman. Sampson, G. (2003), ‘The structure of children’s writing: moving from spoken to adult written norms’, in: S. Granger and S. Petch-Tyson (eds) Extending the Scope of Corpus-based Research. New Applications, New Challenges. Amsterdam & New York: Rodopi. 177-193. So, B.C.P. (2005), ‘From analysis to pedagogic applications: Using newspaper genres to write school genres’, Journal of English for Academic Purposes, 4(1): 67-82. Watrin, P. (2002), Un outil d’analyse de corpus. Traineeship report submitted in partial fulfilment of the degree in Linguistic Engineering. Centre for English Corpus Linguistics. Louvain-la-Neuve: Université catholique de Louvain.

Easy to understand but difficult to use? Raising constructions and information packaging in the advanced learner variety1 Marcus Callies Philipps-Universität Marburg Abstract This paper examines the frequency of occurrence and contextual use of raising constructions in the written production of advanced German and Polish learners of English, based on material from the International Corpus of Learner English (ICLE) and comparable native speaker writing. Raising is an interesting phenomenon to study with respect to both argument realisation and information structure and has been shown to be problematic for child L1 and adult L2 learners of English. In view of the high degree of typological markedness of English raising constructions, raised structures are hypothesised as being underrepresented in the writing of advanced learners, most likely due to avoidance. A quantitative and qualitative textlinguistic analysis of the corpus data reveals that tough-movement in particular is significantly underrepresented in advanced learner writing, and that the learners have problems as to the appropriate use of all types of raising constructions in written discourse in terms of information structuring and thematic progression.

1.

Introduction

Recently, second language acquisition (SLA) research has seen an increasing interest in advanced stages of acquisition and questions of near-native competence, but there are still relatively few studies of advanced learners compared to learners at early and intermediate stages of the learning process. It has been a matter of controversy to what extent and under which circumstances adult speakers of a foreign/second language (L2) can achieve native-like proficiency.2 In many (European) countries the ultimate goal of foreign language teaching at the advanced level is for the students to achieve a near-native command of the L2. However, it is often left unspecified what native-like proficiency means exactly (de Haan 1997: 55). Despite the growing interest in what has also been called the advanced learner variety (ALV), the field is still struggling with both a definition and clarification of the concepts “advanced learner” and “nativelikeness”, as well as an in-depth description of the ALV, especially when it comes to learners’ acquisition of optional and highly L2 specific phenomena in all linguistic subsystems. Advanced learners have typically mastered the L2 rules of morphosyntax, and their written production is mainly free from serious grammatical errors. However, non-native speaker (NNS) writing often sounds unidiomatic and shows subtle differences to texts produced by native speakers

202

Marcus Callies

(NSs). The exact reasons for this “non-nativeness” or “foreign-soundingness” are difficult to pin down and are frequently explained by using vague cover terms such as “unidiomaticity” or “style”. In the last decade, learner corpus research has yielded substantial empirical evidence that texts produced by (advanced) learners and NSs differ in terms of frequencies of certain words, phrases and syntactic structures.3 In a recent overview of the field, Granger (2004: 135) defines advanced interlanguage as “the result of a very complex interplay of factors: developmental, teachinginduced and transfer-related, some shared by several learner populations, others more specific”. According to her, typical features of the ALV are overuse of high frequency vocabulary, overuse of a limited number of prefabs and a much higher degree of personal involvement, as well as stylistic deficiencies, often characterised by an overly spoken style or a somewhat puzzling mixture of formal and informal markers. In addition, there is evidence that another factor that distinguishes advanced learners from NSs is the way they use linguistic structures to organise information in discourse (Carroll et al. 2000). Information structure (IS) management turns out to be problematic even for advanced L2 learners as they experience problems with information sequencing and the end-weight principle in several syntactic patterns (e.g. unusually heavy focus constituents in it-clefts or non-extraposed, thus heavy clausal subjects, see Callies 2006a). The present paper provides a corpus-based examination of raising constructions in the written production of advanced German and Polish learners of English based on material from the International Corpus of Learner English (ICLE) and comparable native speaker writing. In raising constructions,4 a subject/object of a subordinate clause is moved to the subject/object position of a “higher” clause (thus “raised”), resulting in a comparatively large distance between syntactic form and semantic meaning, potential ambiguity and vagueness of surface forms. Raising constructions are not only interesting in terms of argument realisation, but also from the point of view of information structure, and have been shown to fulfil important discourse functions (Mair 1987, 1990, Noël 1997, 1998, Givón 2001). The aim of this study is to examine the frequency of occurrence of raising constructions to find out what factors may influence their use, and to what extent discourse-functional principles – which have largely been neglected in the sparse research on advanced learners – play a role in the ALV. 2.

Raising constructions in English

English allows a number of clause-internal and clause-external syntactic operations which are either impossible or limited in other languages. One of the clause-internal effects is the relatively large degree of freedom in selecting the basic syntactic functions of subject and object which results in a great number of alternations, that is occurrences of a verb with a range of combinations of arguments and adjuncts in various syntactic contexts such as transitivity alternations or the middle construction. Among the clause-external effects are for example raising constructions, syntactic operations that move arguments across

Raising constructions and information packaging in the advanced learner variety

203

clause boundaries. Three types of raising are recognised in the literature and are exemplified below (raised arguments are given in bold type): (1)

a. It seems that Sue is tired. b. Sue seems to be tired.

(subject-to-subject raising, SSR)

(2)

a. We believe that they retire next week. b. We believe them to retire next week. (subject-to-object raising, SOR)

(3)

a. It is difficult to argue with him. b. He is difficult to argue with.

(object-to-subject raising / tough-movement, TM)

In (1) and (2) above, the subjects of the subordinate clauses, Sue and they respectively, are moved to the subject/object position of the “higher” clauses. In (3), it is the object of the subordinate clause which is realised as subject of the matrix clause. The most typical verbs and adjectives that control or trigger raising constructions have been identified through corpus-based descriptive studies of contemporary English. These findings will be reported in the following sections. In the case of subject-to-subject raising (SSR), there are two possible structural variants with complement clauses that are controlled by a number of verbs and adjectival predicates:5 that-extraposition and SSR as illustrated in (1) above. Biber et al.’s (1999: 732) findings show that in all registers SSR is used for the great majority of complement clauses that are controlled by seem and appear, be likely, be unlikely, be certain and be sure. Thus, SSR is the unmarked variant with these verbs and adjectival predicates. In the rest of this paper, I will concentrate on SOR and TM, and exclude SSR. I will, however, consider a variant of SSR that occurs with passive forms of believe-type verbs (see Section 2.1 below). 2.1

Subject-to-object-raising

SOR typically occurs with cognition verbs such as assume, believe or consider (also termed believe-type verbs), verbs of intention, desire or decision such as expect, need, like, prefer or want, and verbs of discovery, e.g. find (Biber et al. 1999: 696). Givón (2001: 273) observes that SOR “most commonly involves verbs of intention, perception or cognition whose subject is typically a dativeexperiencer”. These predicates may appear in two structural variants: a finite thatcomplement clause and SOR involving a non-finite to-infinitive clause: (4)

a. We believe that they retire next week. b. We believe them to retire next week.

(that-complement clause) (SOR)

In addition, most of these verbs also allow a corresponding passive construction in which the subject of the that-clause becomes the subject of the main clause. Thus, the result is a form of SSR (Biber et al. 1999: 697ff.):

204 (4)

Marcus Callies c. They are believed to retire next week.

(passive, SSR)

Biber et al. (1999: 710, 713; similarly Mair 1990: 111) list expect, allow, find, require, enable, ask, prove and consider as the most common verbs that occur in the patterns exemplified in (4b+c) above. Depending on register, with some verbs, notably allow, expect, find and consider (Mair 1990: 113, Biber et al. 1999: 711), the passive construction is even more frequent than the active counterpart.6 2.2

Tough-movement

Mair (1987: 61) claims that it is only a small semantic group of adjectives that frequently occurs with TM, namely “adjectives referring to degrees of difficulty and a number of adjectival and nominal predicates expressing value judgements”, see Table 1 below, which lists the most frequent adjectives triggering TM. Table 1. Most frequent raising adjectives triggering TM in the Survey of English Usage Corpus (875,000 words; Mair 1987) RAISING ADJECTIVE

OBSERVED COUNT

difficult easy hard impossible interesting others (one each) total

36 15 7 5 2 13 78

Similar accounts can also be found in the standard reference grammars of English. Quirk et al. (1985: 1229) note that it is “adjectives referring to degrees of ease and comfort” which trigger TM. Biber et al. (1999: passim) also identify “adjectives of ease or difficulty” as the semantic group that controls TM,7 and find that “easy, difficult and hard are all notably frequent” (ibid. 719, 728). There are three main structural alternatives with infinitive clauses that complement the adjectival predicates listed above. (5)

a. To argue with him is difficult. b. It is difficult to argue with him. c. He is difficult to argue with.

(non-extraposed to-clause) (extraposed to-clause) (post-predicate to-clause with TM)

While non-extraposed infinitive clauses are rare, Biber et al. (1999: 728ff.) find that overall, TM is slightly more common than extraposition for to-clauses controlled by adjectival predicates of ease or difficulty. To-clauses with these adjectives are most common in academic prose. This holds for both extraposed to-clauses and raised objects. Constructions with difficult and easy are considerably more common than constructions with other ease or difficulty

Raising constructions and information packaging in the advanced learner variety

205

adjectives. In addition, there is no register difference: the proportion of extraposed to-clauses compared to raising constructions is about the same in both conversation and academic prose, as appears from Table 2 below. Thus, neither should be considered the unmarked variant. The fact that they occur in roughly the same proportions in different modes suggests that similar discourse factors are operative (see also the discussion in Section 3). Table 2. Frequency counts of TM constructions with the four most common adjectives of ease or difficulty (per million words) in the Longman Spoken and Written English Corpus (approx. 40 million words, Biber et al. 1999: 728) CONVERSATION

raising adjective difficult easy hard impossible

extraposed 6 6 9 2

raised (%) 6 (50) 10 (63) 14 (61) 1 (33)

ACADEMIC PROSE

extraposed 52 26 4 24

raised (%) 79 (60) 36 (58) 4 (50) 4 (17)

Mair (1987, 1990) did not find any significant difference between the occurrence of TM across speech and writing. He concluded that TM constructions were not typical of informal usage, and considered them to be stylistically neutral (1987: 69). However, in view of the more recent and possibly more representative findings by Biber et al. (1999), TM appears to be more typical of academic prose with three out of four of the most frequent adjectives shown in Table 2. 3.

Textual and discourse functions of raising constructions

Raising constructions have been shown to fulfil important discourse functions in terms of information packaging and thematic progression. Several studies have examined believe-type verbs, seeking to explain the choice between a thatcomplement clause (believe that), SOR with an infinitival clause (believe X to be) and the passive counterpart involving SSR (X is believed to be) by drawing on textlinguistic and information-structural factors. Noël (1997) investigated the reasons for the choice between a that-clause and SOR followed by a to-clause with believe-type verbs. He found that in the large majority (90%) of instances of SOR (believe X to be), the raised object (i.e. the actual subject of the infinitive clause) takes up a referent from a preceding clause in the same sentence or from a previous sentence (1997: 277). The following examples, taken from the NS corpus used in the present study (see Section 5 for further information), illustrate this (raised structures are given in bold, underlined constituents are coreferential). (6)

However, if much older women have children, she is less likely to be able to care for them when they are older. It would be unfair to have a child at

206

(7)

Marcus Callies say 60 knowing that you may be suffering from ill health in the near future and expect the child to support you. (LOCNESS) “Crime does not pay”. I have heard this statement a number of times throughout my life. I am at that stage in my life where I believe it to be a question and not a statement. (LOCNESS)

By contrast, the subject of a that-clause takes up a referent from the previous discourse much less often. In fact, the majority of such clauses introduce new referents or re-introduce those that have not been mentioned for some stretch of discourse. Noël (1997: 279) concludes that while contextual givenness seems almost to be a prerequisite for the choice of to-complementation, newness appears to require a that-clause, whose subjects in turn do not need to be new. With regard to the passive variants that represent a form of SSR, Mair (1990: 181) had earlier distinguished between active and passive uses of believetype verbs, stating that the passive version (X is believed to be) allows the promotion of the subject of the complement clause to the subject of the matrix clause, hence sentence-initial position, which is the unmarked position for contextually known/given information. In a subsequent paper, Noël (1998: 1061), however, extended the information-structural function of SOR with active matrix verbs to passive matrices: not just infinitives with passive matrices but also infinitives with active matrices do information-structurally different things from thatclauses. The subjects of that-clauses often introduce “new” referents, whereas the subjects of infinitival complements typically take up previously mentioned referents, irrespective of whether their matrices are active or passive. Noël also showed that in some cases of SOR, a competition of the active matrix clause subject and the given subject of the infinitive for topic position may lead to a referentially redundant situation which can be resolved by using SSR, i.e. a passive matrix verb, see the corpus example below and its suboptimal SORcounterpart in (8b): (8)

a. One in ten Americans now receives food stamps. The federal government considers a family of four to be in poverty if its total income is $14,800 or less. If that same family earns $27,380, it is considered low-income. (LOCNESS) b. If that same family earns $27,380, the federal government/they consider it low-income.

In addition, the passive option is often used to avoid mentioning the active subject (Noël 1997: 26). Only rarely is this variant used to emphasise a non-topical active subject using a prepositional phrase as in (9).

Raising constructions and information packaging in the advanced learner variety (9)

207

One case study involved an architect named Sam, who felt the drug made him feel “better than well”. He became more poised, thoughtful, and could concentrate much easier than before. Most importantly, his symptoms of depression disappeared. The drug has also been tested and found safe by all major pharmaceutical agencies. (LOCNESS)

The function of TM as a means of information packaging is discussed in detail by Mair (1987). He shows that this sentence type serves to appropriately distribute given and new information in a sentence and thus supports the thematic progression of discourse. Mair finds that the majority of raised NPs in TM constructions refer to given information which is often indicated by the use of pronominal forms and anaphoric devices such as it, these and this (Mair 1987: 63), or a simple (sometimes anaphoric) NP without modifiers (this/that + NP): (10)

(11)

Opposition to the whole idea of orphanages refute it mainly because of two reasons. They disagree with the break-up of families and feel the whole policy would be too costly. While these two ideas are difficult to content [sic] with, the truth is in the numbers. (LOCNESS) As the Lottery would sound more appealing due to the larger winnings, greater public interest and the fact that it is easier to fill in than a pools coupon, less people would do the pools coupon [...]. (LOCNESS)

Similarly, Biber et al. (1999: 1728ff.) identify a conspiracy of register, grammatical and information packaging factors, and personal style that influence the choice between TM and an extraposed to-clause. In particular, grammatical complexity and information packaging work together. By contrast, the extraposed variant is used when the implied object of the to-clause presents new information. Then, the object is usually a complex structure which contains another complement clause or a heavy, complex NP with a relative clause or complex modifiers. Thus, raising is dispreferred, since this would result in a heavy, complex NP being the subject. The following lengthy example may suffice to illustrate this interplay: (12)

If we consider that our society rewards valuable labor - valuable contributions - with first, economic measures, and, second, measures of status, then the child care worker may reasonably be judged as not providing a “valuable” contribution to society. It is difficult to imagine, however, a society where workers are encouraged to take employment seriously - to say nothing of finding satisfaction from employment and where the work force is filled with a diversity of willing workers, without considering the fundamental value of child care providers. If we recognize that a work force filled with such diversity of workers will inevitably include women - of all ages - as well as man with family responsibilities, then we see the undeniable economic connection between this “low value” contribution of child care providers, and the economic

208

Marcus Callies strength of the nation. It is not difficult to conclude that child care workers are underpaid, and the issue will be difficult to resolve without involving employers or the government in the child care industry - which may not be the ideal solution. (LOCNESS)

In a similar vein, Givón (2001: 272) discusses the function of raising as a means of foregrounding an important topic which is “converted from an argument of the subordinate clause to a grammatical argument – either subject or object – of the main clause”. He also notes the topicalising pragmatic effect of raising. 4.

Raising constructions in language acquisition

4.1

Previous research

Tough-movement was one of the first complex syntactic structures studied closely in L1 acquisition (C. Chomsky 1969, Cromer 1970, and most recently Anderson 2005). TM constructions were found to be particularly problematic for child learners of English and are consequently acquired rather late, i.e. the acquisition is not complete before the age of nine. Experimental studies (e.g. Cromer 1970, Anderson 2005) have identified three stages in children’s acquisition of these structures: (1) an initial stage in which children assign it non-adult-like readings, interpreting the subject NP as the subject rather than as the object of the infinitive verb, (2) an intermediate stage in which they vary in their responses, sometimes giving adult-like, sometimes non-adult-like readings, and (3) a final stage in which they show adult-like competence. While most of the syntactic phenomena that exhibit a large distance between grammatical function and semantic meaning can be assumed to represent problems for learners of English as a foreign language (EFL), only few of them have been investigated empirically (Kortmann 1998: 156; for semantically marked subjects see Callies 2006b). “Tight-fit”, that is more explicit and semantically transparent structures in the L2 have been shown to be acquired earlier, are preferred options and have a wider range of application also in the advanced stages of the learning process. By contrast, “loose-fit”, that is less explicit and semantically opaque variants are avoided even when they exist in the L1 (Kellerman 1979a: 42, Kortmann 1998: 160). Cook (1973) was the first to replicate Cromer’s (1970) experiment with adult L2 learners of English from several L1 backgrounds. Interestingly, on the basis of his results he was able to identify three similar developmental stages in L2 acquisition. Subsequent studies that also replicated Cromer’s study in like or methodologically slightly adapted fashion with EFL learners from different L1 backgrounds produced corresponding results: learners regularly misanalysed the subject of the TM construction as the logical subject, not the object, of the underlying clause (d’Anglejan & Tucker 1975 with native speakers of Canadian French, Cooper et al. 1979 with Arabic and Hebrew EFL learners, Bongaerts

Raising constructions and information packaging in the advanced learner variety

209

1983 with Dutch learners, Trosberg 1983 with Danish learners, Chiang & Costello 1983 and Yip 1995 with Chinese learners, and Yamaoka 1988 with Japanese learners). Bongaerts (1983) found that the structure was relatively easier for the Dutch than for the French, Arabic and Hebrew subjects in the other experiments, possibly due to a structural parallelism between Dutch and English. However, Jordens (1977, 1978) reports that with near-synonymous and structurally related sentences (raising, mediopassive constructions and marked subjects), advanced Dutch learners of German showed different acceptability rates. Dutch learners were more likely to accept sentences with “canonical” subjects and objects than sentences with marked (“raised”) subjects or objects. Yip (1995: 153ff.) observed that in production, Chinese learners frequently use what she calls “pseudo-tough-movement”, i.e. interlanguage structures that seem to be misuses of TM as in (13) and canonical TM structures exhibiting overpassivisation as in (14): (13) (14)

I am very easy to forget. (‘I forget [things] very easily.’) Foreigners are easy to be misunderstood. (‘Foreigners are easy to misunderstand.’)

Yip claims that instances like (13) involve more than a mere misapplication or overgeneralisation of TM to underlying subjects (rather than objects). In fact, canonical forms are often, for some learners even consistently, accompanied by passivisation of the verb as in (14), while target-like forms occur only rarely. Chinese learners also show considerable uncertainty with respect to the grammaticality of TM. In view of this evidence, Yip argues that what seems to be a non-target-like use of TM is in fact an innovative interlanguage form that results from a misanalysis of TM in English in combination with an across-theboard subject-raising strategy rooted in L1, L2 and universal preferences (Yip 1995: 156, 159). Thus, Chinese EFL learners interpret structures like (13) as instances of subject raising (‘It is very easy for me to forget [things]’), while a correct interpretation of object-raising (TM) is mapped onto an over-passivised form like I am very easy to be forgot (‘It is very easy to forget me’) (ibid. 168). In sum, previous research shows that L2 learners – similarly to children acquiring their mother tongue – have difficulties in correctly interpreting TM constructions, and exhibit similar developmental stages with more advanced learners showing higher levels of competence. However, most of the studies discussed above suffer from methodological shortcomings, since they have mostly tested meaning comprehension and correct identification of the actual (logical) subject of few items using experimental tasks only (role play, grammaticality judgements, translation). They did not examine written learner production, i.e. contextualised, longer stretches of written texts. The corpus-based approach adopted in the present paper provides such an opportunity, and it also

210

Marcus Callies

allows for an investigation of avoidance phenomena and the role of the principles of information structure in the ALV. 4.2

Why they are difficult to acquire: The markedness of raising constructions

Givón (1991) proposes a definition of markedness that includes the following correlates: structural complexity, frequency distribution and cognitive complexity. According to these criteria, marked elements are structurally more complex, less frequent and therefore cognitively more salient. They require more attention, mental effort, resulting in more processing time for the recipient. Givón’s understanding of markedness is extremely useful in the present context since it integrates the notion of cognitive complexity, and postulates a correlation between markedness and the cognitive-physiological complexity of linguistic units: marked structures require more cognitive effort to process them. In the following, I will show that this also applies to raising constructions. Raising is one of the phenomena studied by Hawkins (1986), who proposes that many contrasts between English and German can be explained in terms of a typological continuum whereby languages vary according to the degree to which morphological and syntactic surface form and semantic meaning correspond. He argues that there is a greater ambiguity and/or vagueness of surface forms in English. While German exhibits a tight-fit between surface form and semantic representation, English shows a loose-fit correspondence. This can also be observed in raising constructions. In SOR, the raised object is not an argument of the finite, but the non-finite verb: (15)

a. We believe that his teachers retire next week. b. We believe his teachers/them to retire next week.

(SOR)

In (15b) above, the NP his teachers functions grammatically as direct object of the verb (note that the respective pronoun appears in accusative case), with the infinitive clause as object complement. Such an interpretation, however, would in fact result in the incorrect inference ‘We believe his teachers’ (cf. Biber et al. 1999: 696). Thus, in terms of meaning, his teachers functions logically as the subject of the underlying clause (his teachers retire next week), and not as the direct object of the verb. Mair (1987: 64ff.) observes that tough-movement constructions are deceptively similar to simple sentences which consist of a subject, a copula and a predicative adjective governing an infinitive. Compare: (16) (17)

a. This problem is difficult to see. b. It is difficult to see this problem. a. John is reluctant to go. b. *It is reluctant to go John.

Raising constructions and information packaging in the advanced learner variety

211

He further notes that the relationship between raised constructions such as toughmovement and their underlying bi-clausal forms can be even more obscured by the existence of phrases in which the raised arguments are used as if they were ordinary attributive adjectives (Mair 1990: 63ff.): (18)

a. ... one of his tremendously hard to understand papers b. ... who speaks fluent but difficult to understand English c. ... in a clear and easy-to-follow manner

In sum, when compared with their underlying, bi-clausal variants, raising constructions exhibit a larger distance between syntactic form and semantic meaning, since the grammatical subject/object is not the logical or semantic subject/object of these sentences. This results in ambiguity and potential vagueness of surface forms. Consequently, they are functionally and semantically more complex, less transparent and less explicit, are cognitively more costly and require more processing time for the recipient in terms of the analysability and decoding of the form-function relation (Hawkins 1986, Legenhausen & Rohdenburg 1995: 138). 4.3

Typologically-oriented contrastive linguistics and implications for L2 acquisition

Eckman (1977a) proposed a universal implicational hierarchy of raising processes: (19)

Subject-to-object > subject-to-subject > object-to-subject

This hierarchy states that languages that have object-to-subject raising (or TM) also have the two raising types that are higher on the hierarchy. By contrast, there should be no language that has, e.g. SSR, but not SOR. Eckman mentions English, French and German as languages which have all three types of raising, Hungarian, Modern Greek and Polish as those that have SOR and SSR but lack TM (or object-to-subject), and Armenian, Hebrew, Turkish, Egyptian and Lebanese Arabic as languages that only have SOR. Similar to Eckman, Van Valin (2001: 53) notes that “cross-linguistically, matrix-coding-as-subject constructions [i.e. subject-raising, MC] are much more common than matrix-coding-as-object constructions [i.e. object-raising, MC]”. From a functional perspective, TM is a uniform cross-linguistic phenomenon in that it explicitly indicates topicalisation of the raised NP. However, the formal linguistic means to express this function are different from language to language, and correlate with other typological syntactic features such as basic word order and subject-prominence (Comrie & Matthews 1990: 55). Mair (1992) shows that English allows SOR after several causation and cognition verbs, whereas German does not. Generally speaking, although both English and German have all three types of raising, a contrastive analysis of raising in these two languages shows that in German, raising is more restricted

212

Marcus Callies

and only a subset of the English raising verbs and adjectives occur (König 1971: 85-89, Hawkins 1986: 75ff., König 1995: 155f., 157, Legenhausen & Rohdenburg 1995: 135). In all three types of raising, English is more productive than German because there are many more verbs and adjectives that trigger a certain raising construction (Hawkins 1986: 75ff.). Referring back to Eckman’s (1977a) hierarchy, Polish is even more restricted than German. As for SSR/SOR, Fisiak et al. (1978: 154) note that “few plausible cases of ‘subject raising’ are to be found in Polish”. In other words, English sentences in which the subject of an embedded clause is raised to the main clause subject position have no congruent equivalents in Polish, e.g. (20)

He is thought by her to be very rich. *Jest myĞlany przez nią byü bardzo bogaty.

An exception is the Polish equivalent to the English verb seem (zdawaü siĊ / wydawaü siĊ) which does allow SSR: (21)

He seems to be glad. On wydaje siĊ byü zadowolony.

With regard to SOR, Fisiak et al. (ibid. 151) point out that English allows a choice between two types of complementisers in a context where only one is possible in Polish, namely the that-clause: (22)

a. I expected John to come. *Spodziewaáem siĊ Janka przyjĞü. b. I expected that John would come. Spodziewaáem siĊ, Īe Janek przyjdzie.

Thus, equivalent Polish sentences to I expect/believe you to be honest are ungrammatical in Polish (ibid. 153). Again, one exception is the Polish equivalent to the English verb consider (uwaĪaü), where SOR is possible. However, while deletion of the infinitive copula is optional in English, it is obligatory in Polish (ibid. 154): (23)

a. I consider you a winner. UwaĪam ciĊ za zwyciĊzcĊ. b. I consider you (to be) funny. *UwaĪam ciĊ byü Ğmiesznym.

Finally, in TM constructions there is a systematic contrast between English infinitives and Polish gerundives. While the semantic constraints seem to be similar in both languages, there is a major difference in the type of complementisers selected for the embedded clause:

Raising constructions and information packaging in the advanced learner variety (24)

213

a. His name is easy to remember. *Jego imiĊ jest áatwe zapamiĊtaü. b. *His name is easy for remembering. Jego imiĊ jest áatwe do zapamiĊtania.

A direct translation from English into Polish without a change of complementiser is impossible. One of the very few papers on raising in Polish (Boniewicz 1982) reports that “Raising in Polish is not generally known” and concludes that – compared to English – “the scope of Raising in Polish is very narrow” and “Polish Constructions avoid infinitival complementation” since “non-verbal complements are preferable” (Boniewicz 1982: 98, 109). To conclude, as far as the types and the frequency of raising constructions are concerned, a contrastive analysis shows that English is not only highly marked when compared with Polish and German, but also in a wider crosslinguistic perspective. Givón (2001: 282) states that “English may be the most promiscuous language when it comes to raising. Other languages, even when typologically rather close to English, allow little or no raising”. Consequently, English clearly exhibits a higher degree of typological markedness. In this paper, I adopt a functional-typological approach to SLA (Eckman 1996, Giacalone Ramat 2003), which is based on the assumption that the findings and generalisations made by language typologists can also be applied to language acquisition, and argues for the significance of universal (implicational) hierarchies for the prediction and explanation of (non-)transfer in SLA. The basic assumption is that interlanguages (ILs) are natural languages. If typological language universals are universal to all natural human languages, then they should also hold for ILs, as captured by the Interlanguage Structural Conformity Hypothesis: “All universals that are true for primary languages are also true for interlanguages” (Eckman, Moravcsik & Wirth 1989: 195). Since many language universals can be expressed in terms of (implicational) hierarchical relations with respect to cross-linguistic/typological markedness, it is reasonable to assume that such hierarchical relations between linguistic phenomena should also be present in ILs. Consequently, it should be possible to predict the occurrence of selected linguistic features in ILs, depending on their position in the hierarchy and the relative degree of typological markedness. When used within the functional-typological approach, the notion of typological markedness is defined as an empirically motivated construct, “determined on the basis of crosslinguistic data” (Eckman 1996: 201f.). This concept has been shown to be highly significant and useful for the prediction of the (non-)transferability of linguistic structures and their learnability, interacting with language transfer. In view of the large amount of diverse empirical evidence, it is widely accepted that L2 acquisition is to some extent influenced by the learner’s native language, and that this influence may affect all linguistic subsystems (cf. Odlin 2003). Thus, the question is not whether or not transfer exists, but when, what, how much and why specific mother-tongue

214

Marcus Callies

phenomena are transferred while others are not. I will use the term language transfer in the sense of crosslinguistic influence, defined as “the influence resulting from similarities and differences between the target language and any other language that has been previously (and perhaps imperfectly) acquired” (Odlin 1989: 27), thereby incorporating positive transfer as well as interference, avoidance and overproduction. With regard to making predictions as to the learnability of linguistic structures and their (non-)transferability from a learner’s native language, Kellerman (1979b, 1983) initiated a re-evaluation of the notion of transfer. In his view, the learner is an active decision-maker on what linguistic structures may be transferable to the second language. Transfer is seen as a cognitive process subject to a psycholinguistic rather than linguistic understanding of markedness in terms of psychological and perceptual complexity, not structural complexity. Eckman (1977b) introduced the Markedness Differential Hypothesis (MDH) and argued that on the basis of a contrastive analysis of two languages (L1 and L2) and the inclusion of the concepts of typological markedness and cross-linguistic influence, it should be possible to predict areas of difficulty for an L2 learner. In a nutshell, the MDH claims that L1 structures that are different from L2 structures and typologically more marked will not be transferred, whereas L1 structures that are different from L2 structures and typologically less marked are more likely to be transferred. Additionally, predictions can be made as to both the order and difficulty of linguistic features in the acquisition process: less marked structures will be acquired first or without difficulty, more marked structures are expected to be acquired later or with greater difficulty. In sum, the MDH identifies potential difficulties in the L2 learning process not merely on the basis of similarities and differences derived from a contrastive analysis (CA) of two languages (as in traditional CA), but through a combination of the concepts of typological markedness and crosslinguistic influence. I argue that raising constructions, and TM in particular, can be considered to be problematic even for advanced learners of English for two reasons: 1) they are marked because they exhibit a large form-function distance, and are functionally and semantically more complex, less transparent and less explicit; 2) as far as the different types, frequency and restrictions of raising constructions are concerned, English is highly marked in a cross-linguistic perspective. In sum, raising constructions can be hypothesised as being underrepresented in the writing of advanced EFL learners, most likely due to avoidance (Kortmann 1998: 156). 5.

Methodology and data

The present study adopts a corpus-based approach to the analysis of advanced interlanguage on the basis of written data taken from the International Corpus of Learner English (ICLE, Granger et al. 2002), which consists of mostly argumentative essays produced by university students of English with different native languages. All of the informants that contributed essays to the corpus share

Raising constructions and information packaging in the advanced learner variety

215

the following characteristics: they are all university undergraduates in their twenties, have learned English in an EFL context involving classroom instruction, and are usually in their third or fourth year of studies. Thus, their English proficiency level ranges from higher intermediate to advanced, proficiency being assessed on external criteria, namely institutional status (ibid. 14). The corpora used for this study were sampled from ICLE and consist of more than 200,000 words each. They include only argumentative essays produced by students whose native languages are German and Polish respectively, and who studied at universities in Germany/Austria and Poland. The learner corpora will be referred to as GICLE and PICLE, respectively (see Table 3 below). L2 exposure, defined for the present purposes as the amount of time spent in an English-speaking country (for example an extended study-abroad period in the target culture), is generally assumed to positively affect a student’s overall proficiency. Therefore, those students who had spent more than 12 months either living or studying in an English-speaking country were excluded from the study in order to provide for a more homogeneous learner group. Table 3. Corpora used in the present study CORPUS NAME

TYPE

NO. OF ESSAYS

APPROX. NO. OF WORDS

Polish ICLE (PICLE) German ICLE (GICLE) US LOCNESS BRIT LOCNESS total LOCNESS

NNS NNS NS NS NS

350 395 176 165 342

224,000 210,000 150,000 78,000 228,000

The learner corpora were compared to a corpus of similar writing produced by native speakers of British and American English, sampled from the Louvain Corpus of Native English Essays (LOCNESS), which consists of a total of 300,000 words of (mostly) argumentative essays by British university students and grammar school students taking their A-levels, as well as US college students (ibid. 41). The major advantage of this control corpus is that it is comparable to the learner corpora both in terms of text type (argumentative essays) and, broadly speaking, participant age and educational background (graduating high-school and undergraduate university students). The corpus analysis focused on a selection of frequent verbs and adjectives that typically trigger raising constructions (see the discussion in Section 2 above): verbs of cognition (believe, consider, find) and desire and want (expect, like, need, prefer, require, want),8 and the four most frequent raising adjectives difficult, easy, hard and impossible. In addition, 34 less frequent toughadjectives (some of them listed in Rohdenburg 1993: 260) were also examined.9 The respective constructions were extracted on the basis of these triggering verbs and adjectives with the use of WordSmith Tools 3.0 (Scott 1999). The hits were then checked manually and submitted to a quantitative statistical and close qualitative textlinguistic analysis.

Marcus Callies

216 6.

Results and discussion

6.1

Quantitative results

Table 4 gives the frequencies of use of the three types of constructions (thatclause, SOR and SSR with passive matrix verbs) after selected raising verbs in the three corpora. Table 4. Frequency of use of variant constructions after selected raising verbs VERB

THAT-CLAUSE (%) LOCNESS

GICLE

10

SOR (%) LOCNESS

GICLE

PICLE

LOCNESS

GICLE

PICLE

51 60 (86.4) (89.6)

5 (3.2)

7 (11.9)

2 (3.0)

4 (2.6)

1 (1.7)

5 (7.5)

14 (24.1) 6 (26.1) 6 (35.3)

21 (67.7) 7 (33.3) 1 (50)

14 (24.1) 7 (43.8) 2 (28.6)

40 9 44 (69.0) (29.0) (75.9) 13 12 8 (56.5) (57.1) (50.0) 8 1 5 (47.1) (50) (71.4)

23 (44.2) 2 (100) 1 (100) 0

15 (78.9) 0

29 (87.9) 2 (100) 1 (100) 0

9 (17.3) 0

0 0

1 (3.0) 0

0

0

0

0

0

0

24 (100) 81 (38.9)

0

0

0

145 (94.2)

consider

4 (6.9) 4 (17.4) 3 (17.6)

1 (3.2) 2 (9.5) 0

20 (38.5) 0

4 (21.1) 0

3

need

0

0

0

prefer

0

0

0

want

0

0

0

total

176 (50.7)

require find like

PASSIVE / SSR (%)

PICLE

believe

expect

11

0 1 (6.3) 0

0

58 64 (34.9) (30.8)

40 (100) 97 (28.0)

0 1 (100) 33 (100) 85 (51.2)

74 23 63 (21.3) (13.9) (30.3)

The table shows that in LOCNESS the verb believe most often appears with a that-clause. By contrast, find, want and the infrequent verbs like and need regularly take SOR followed by a to-clause, while consider, expect and require predominantly occur with passive matrix verbs (see figures in bold print in Table 4). The latter result is in line with the observations by Mair (1990) and Biber et al. (1999) reported above, namely that, among others, with expect and consider the passive construction is more frequent than the active counterpart. These tendencies are largely mirrored in the learner corpora, but there are exceptions. The significance tests12 reveal several interesting differences that run counter to the initial hypothesis. Despite the assumed underrepresentation of all types of raising structures in advanced learner writing, with some verbs both SOR and SSR are in fact overrepresented in PICLE when compared to LOCNESS (log-likelihood score G2 of +4.13 and +4.78 respectively, both equalling p<0.05;

Raising constructions and information packaging in the advanced learner variety

217

based on total frequencies). This overrepresentation is most likely due to transfer of training, understood as resulting from “identifiable items in training procedures” (Selinker 1972: 216). In view of the structural differences between verbs and their complement clauses in English and Polish outlined in Section 4.3, which are believed to cause learning difficulties, both SOR and SSR with passive matrix verbs feature prominently in teaching materials and syllabi in Polish universities (Dr. Konrad SzczeĞniak, English Department at the University of Silesia, personal communication). While SOR is overrepresented in GICLE across the board (G2 = +16.14; p<0.01), SSR with passive matrix verbs seems to be underrepresented, the difference not being statistically significant, however (G2 = +3.50). Table 5 gives the frequencies of use for the three possible variants after raising adjectives in the three corpora. Table 5. Frequency of use of variant constructions after the four most common raising adjectives ADJECTIVE

difficult easy14 hard impossible total

EXTRAPOSED (%)

NON-EXTRAPOSED (%)

LOCNESS

PICLE

GICLE LOCNESS

24 (68.6) 14 (46.7) 18 (50.0) 13 (76.5) 69 (58.5)

50 (69.4) 45 (63.4) 25 (80.6) 32 (86.5) 152 (72.0)

32 (69.6) 23 (60.5) 18 (81.8) 21 (91.3) 94 (72.9)

PICLE

RAISED (%)

GICLE LOCNESS

0

0

0 0

2 (2.8) 0

1 (2.2) 3 (7.9) 0

1 (5.9) 1 (5.9)

3 (8.1) 5 (2.4)

2 (8.7) 6 (4.7)

11 (31.4) 16 (53.3) 18 (50.0) 3 (17.6) 48 (40.7)

13

PICLE

GICLE

22 (30.6) 24 (33.8) 6 (19.4) 2 (5.4) 54 (25.6)

13 (28.3) 12 (31.6) 4 (18.2) 0 29 (22.5)

The significance tests show a significant underrepresentation of TM in both learner corpora (G2 of -5.36 for LOCNESS/PICLE and -6.58 for LOCNESS/ GICLE, both scores equalling p<0.05). Similar results were found when checking the 34 less frequent tough-adjectives mentioned earlier. With these, 13 instances of TM constructions were found in the NS corpus, but only 7 and 4 in PICLE and GICLE, respectively. The reason for this underrepresentation will be discussed in the following. 6.2

Qualitative results

In both learner corpora, there are numerous cases that evidence learners’ problems with the use of raising constructions on three counts: weight distribution, information structure and thematic progression. This often leads to disfluency, thematic redundancy and an awkward style, most likely due to students’ unawareness of raised structures’ discourse functions. The following two examples show learners’ difficulties with the appropriate ordering of given and new information, and thus, the thematic

218

Marcus Callies

progression of discourse. The repeated mentioning of given information, indicated by the pronominal forms they/them, literally calls for the use of TM as shown in (25b) and (26b), which would optimise information flow (given before new) and textual cohesion. (25)

(26)

a. As I think that I have explained of what an enormous importance the joys of life are for me, I want to think of what they are. It is very difficult to define them, and it is also not easy to say what they really are. (GICLE) b. They are very difficult to define, ... a. The more and more popular use of illegal drugs is a serious problem of the contemporary world. What is most dangerous about it is the fact that drugs cause addiction. That is why it is extremely difficult to give them up. (PICLE) b. That is why they are extremely difficult to give up.

Students’ difficulty with the positioning of syntactically complex, thus heavy constituents is suggested by mere quantitative evidence. The number of nonextraposed clauses is much higher in both learner corpora (five and six in PICLE and GICLE, respectively, but only one in the NS corpus). The learner corpora contain several instances of untypically heavy, non-extraposed subject clauses, often containing new information (given in bold in the following examples). (27)

(28)

Crime results from a combination of various factors including desperation, deprivation or genetic disposition. To believe that executing a certain number of offenders who have been caught will seriously affect the reasoning of potential murderers is a simplistic response to a complex problem. (PICLE) This, however, is no acceptable solution to the problem because home schooling produces new difficulties; the task of teaching is not easy, and only qualified persons should be responsible for it. Whether parents are qualified to teach, and whether teaching at home is favourable to children's development, is not easy to decide. (GICLE)

While SSR with passive matrix verbs is underrepresented in GICLE, it has been shown above to be overrepresented in PICLE due to transfer of training. This also seems to be the reason for several instances of a textual (mis-)use of SSR in the Polish learner corpus. The passive variant is usually chosen to avoid mentioning the active subject, but this option often goes unused. By contrast, the mentioning of an agent in the by-phrases in the following examples is not used to emphasise a non-topical active subject, but leads to a disruption of text and an awkward style.15 Textual cohesion can easily be improved either by simply deleting the agent as in (29b), or using the SOR variant with an active matrix verb as in (30b) and (31b) if the agent is indeed supposed to be mentioned.

Raising constructions and information packaging in the advanced learner variety (29)

(30)

(31)

219

a. Those Eastern criminals apply very violent means. They attack cars in the highways taking them away from their owners. They demand ransoms from the proprietors of restaurants and hotels. Finally, they are considered by the police to set the bombs which have exploded in Warsaw. (PICLE) b. Finally, they are considered to have set the bombs which have exploded in Warsaw. a. The device is designed so as to be built in the ordinary TV-set in order to encode and/or block certain television broadcasts, especially those considered by the adults unsuitable for children. (PICLE) b. ..., especially those that adults consider (to be) unsuitable for children. a. The other group, consists of people who go to the cinema not only in order to have fun but also to consider and discuss important issues raised in the film. By means of visual pictures they use their imagination. Krzysztof Kieslowski is considered to be one of the best film directors by this part of audience [sic]. (PICLE) b. This part of the audience considers Krzysztof Kieslowski to be one of the best film directors.

The fact that even advanced learners encounter problems with the structuring of given and new information, often resulting in thematic redundancy, may simply be considered a stylistic deficiency. However, I think that the underlying reason for this is a more fundamental interlanguage phenomenon, namely an avoidance strategy. It appears that many students simply avoid using TM constructions, possibly because they believe that the infinitive construction after a raising adjective is incomplete and lacks an object. The following are only two of many such examples in the learner corpora where the object NP represents given information and seems redundant, thus could have been left out completely (raised structures are given in bold, underlined constituents are co-referential). (32)

(33)

An hour before the examination, that we have once sat, a friend asks to tell him or her something about it. We know that this particular examination was hard and that it was not, at all, easy to pass it. (PICLE) People, working as hard as other people, do not understand why they earn less money than others. Why does it seem so easy for others to amass money? Isn’t it much easier to obtain it illegally than to save it for years? (GICLE)

Finally, and most interestingly, both learner corpora exhibit several instances of “pseudo-tough-movement” (TM structures with overpassivisation), i.e. interlanguage structures that Yip (1995) observed for Chinese EFL learners: (34)

The children are not smacked but still able to recognize the boundaries and limits that cannot be overcome. Such young people provided with help

220

(35)

Marcus Callies develop proper reaction techniques; they are much more eager to cooperate and their emotions are easier to be controlled. (PICLE) On balance, I would say that corporal punishment is no appropriate means to fight against criminality. Although corporal punishment may have a deterrent effect on certain criminals and although it is easier to be executed than imprisonment, it would not really solve the current crimewave. (GICLE)

The reason why these occur in advanced learners’ production with native languages that are typologically so different from Chinese is yet unclear and needs further examination. Admittedly, they are rare compared to the mostly target-like use of TM in general, but these interlanguage forms may be remnants of a (possibly universal) developmental stage in the acquisition of TM. 7.

Conclusion

This paper has examined the frequency of occurrence and contextual use of raising constructions in the written production of advanced German and Polish learners of English. The findings of this study strongly support the view that advanced interlanguage is indeed the result of a very complex interplay of many factors. I have shown that several interlanguage phenomena are at play in the advanced learner variety: avoidance, resulting in an underrepresentation of linguistic structures, transfer of training which may explain the overrepresentation of SOR and SSR with several verbs, and unawareness or lack of knowledge of raised structures’ discourse functions, resulting in a non-target-like textual use of these structures. In addition, the occurrence of pseudo-tough-movement in the writing of advanced learners with typologically very different native languages may be traces of a universal developmental stage in the acquisition of TM. I have shown that TM is significantly underrepresented in the writing of two groups of advanced EFL learners. By adopting a typologically-oriented view on contrastive linguistics and SLA, I have argued that one reason for this underrepresentation is avoidance due to the high degree of typological markedness of English TM constructions, as well as their functional and semantic complexity, and comparatively little transparency and explicitness due to a large form-function distance in terms of argument realisation. Hence, this result supports previous findings that advanced learners avoid loose-fit, that is less explicit and semantically opaque variants (for example non-agentive subjects or mediopassives), even when they do exist in the L1. Generally speaking, however, it has to be noted that for a study of verb complementation all three corpora are comparatively small. Thus, the frequency counts for most raising verbs and adjectives are low and have to be interpreted with caution. For more reliable quantitative results, the study should be replicated on the basis of larger corpora. With all types of raising constructions, the learners face difficulties with respect to their effective use in terms of weight distribution, information structuring and textual cohesion, which often leads to disfluency, thematic

Raising constructions and information packaging in the advanced learner variety

221

redundancy and an awkward style. In particular, these are problems with an appropriate ordering of sentence constituents with respect to both syntactic complexity and information status, confirming earlier findings that information structure management is problematic even for advanced EFL learners. The underlying reason for this I believe to be unawareness in combination with avoidance, possibly because general discourse-functional principles, and the textual functions of raising constructions in particular, seem to be unclear or completely unknown even to advanced EFL learners. This is hardly surprising given that teaching materials have been shown to exhibit deficits. In particular, discourse-functional concepts of language use that relate to information structure tend to be underrepresented in materials for (advanced) L2 learners. Textbooks more often deal with formal syntactic aspects, and only infrequently deal with concepts such as information status or weight (see the discussion in Callies 2006a). Notes 1

This paper is dedicated to Rüdiger Zimmermann on the occasion of his 68th birthday. I would like to thank David Smith and the editors of the present volume for helpful comments on an earlier version. All remaining errors are my own.

2

See for example Birdsong (1999) for studies that investigate advanced learners’ success in areas such as phonology, grammar and discourse.

3

See Hinkel (2002) for a comprehensive and detailed analysis of English native speakers’ and second language writers’ texts. Studies that focus on specific linguistic features are for example Granger & Tyson (1996) on connectors, Granger (1997) on participle clauses, Lorenz (1998, 1999) on adjective intensification, Boström Aronsson (2001, 2003) on clefts and extraposition, and Callies (2006a) on lexical and syntactic focusing devices.

4

The term ‘raising’ is merely used as a convenient descriptive label with no theoretical implications intended.

5

Three groups of predicates typically occur with SSR (Mair 1990, Biber et al. 1999, Givón 2001): “Verbs of probability or simple fact” (Biber et al. 1999: 695, 731) such as seem, appear, turn out, “degree of certainty adjectival predicates” (ibid. 716, 718), of which “the adjectives likely and unlikely are especially common”, and “passive voice mental verbs” (ibid. 731), for example be believed, be assumed and be found.

6

See also the detailed study by Noël (1998) for more such verbs.

222

Marcus Callies

7

While it may be true that this group represents the most frequent raising adjectives, the list of such triggers is potentially much longer, see e.g. Rohdenburg (1993: 260).

8

Despite their relatively high frequency, the verbs allow and enable were not included because with these verbs, the “raised” object can be interpreted to be both object of the matrix clause and subject of the toclause, depending on whether it is an animate or inanimate entity (see Biber et al. 1999: 696f.).

9

These adjectives were expensive, inexpensive, cheap, costly, harmless, dangerous, safe, pleasant, unpleasant, nice, awful, rotten, all right and o.k. (in various spellings), desirable, bad, good, depressing, complicated, cumbersome, tough, useful, handy, convenient, boring, tedious, exciting, interesting, painful, embarrassing, fun, terrible, strange and simple.

10

This count includes instances of truncated that-clauses, e.g. I really do believe (that) they like it.

11

This count includes instances where the to + infinitive sequence has been deleted, e.g. as in Most men consider themselves (to be) superior to women.

12

Statistical measures were calculated using the Log-likelihood calculator provided by Paul Rayson at Lancaster University, available at http://lingo.lancs.ac.uk/llwizard.html (last accessed on May 7th, 2008). The corpus data were also processed and cross-tabulated in contingency tables using SPSS, hence double-checked for statistically significant differences using the Likelihood Ratio chi-square.

13

This count also includes “raised” structures related to TM, i.e. instances of so-called pseudo-tough-movement (Mair 1990: 72ff.) of the type an easy decision to make, and phrases in which infinitives are used as if they were ordinary attributive adjectives, e.g. simple, easy-to-understand facts. Note that Mair’s use of the term is quite different from Yip’s introduced in Section 4.1.

14

In the Polish learner corpus there is an overrepresentation of the adjective easy because of numerous essays on the topic “English is an easy language to learn/teach”. The title itself contains a TM structure and is likely to have caused priming effects. For this reason, all raised structures of the type “X is (NP) easy to learn/teach” have been excluded.

15

This may also be an effect of transfer of training. See the undeleted byphrase in example (20) taken from Fisiak et al. (1978), which is still widely used as the standard textbook in contrastive grammar classes in English Departments in Poland.

Raising constructions and information packaging in the advanced learner variety

223

References Anderson, D.L. (2005), The Acquisition of Tough-Movement in English. Unpublished PhD dissertation, University of Cambridge. Biber, D., S. Johansson, G. Leech, S. Conrad and E. Finegan (1999), Longman Grammar of Spoken and Written English. Harlow: Longman. Birdsong, D. (ed.) (1999), Second Language Acquisition and the Critical Period Hypothesis. Mahwah, NJ: Erlbaum. Bongaerts, T. (1983), ‘The comprehension of three complex English structures by Dutch learners’, Language Learning, 33: 159-182. Boniewicz, A. (1982), ‘Properties of raised constructions in English and Polish’, Papers and Studies in Contrastive Linguistics, 15: 95-110. Boström Aronsson, M. (2001), ‘The use of extraposition in writing produced by Swedish advanced learners of English’, in: G. Byrman, H. Lindquist and M. Levin (eds) Corpora in Research and Teaching. Papers from the ASLA symposium Corpora in Research and Teaching, Växjö, 11-12 November 1999. Uppsala: FUMS. 32-42. Boström Aronsson, M. (2003), ‘On clefts and information structure in Swedish EFL writing’, in: S. Granger and S. Petch-Tyson (eds) Extending the Scope of Corpus-Based Research. New Applications, New Challenges. Amsterdam: Rodopi. 197-210. Callies, M. (2006a), Information highlighting and the use of focusing devices in advanced German learner English. Unpublished PhD dissertation, Philipps-Universität Marburg. Callies, M. (2006b), ‘Why money can’t buy you anything in German: A functional-typological approach to the mapping of semantic roles to syntactic functions in SLA’, in: J. Arabski (ed.) Cross-linguistic Influences in the Second Language Lexicon. Clevedon: Multilingual Matters. 111129. Carroll, M., J. Murcia-Serra, M. Watorek and A. Bendiscioli (2000), ‘The relevance of information organization to second language acquisition studies. The descriptive discourse of advanced adult learners of German’, Studies in Second Language Acquisition, 22: 441-466. Chiang, J.S. and J.R. Costello (1983), ‘The acquisition of syntax in first and second language learning’, International Review of Applied Linguistics, 21(1): 19-33. Chomsky, C. (1969), The Acquisition of Syntax in Children from 5 to 10. Cambridge, MA: MIT Press. Comrie, B. and S. Matthews (1990), ‘Prolegomena to a typology of tough movement’, in: W. Croft, K. Denning and S. Kemmer (eds) Studies in Typology and Diachrony. Papers presented to Joseph H. Greenberg on his 75th Birthday. Amsterdam: John Benjamins Publishing Company. 43-58. Cook, V.J. (1973), ‘The comparison of language development in native children and foreign adults’, International Review of Applied Linguistics, 11(1): 1328.

224

Marcus Callies

Cooper, R., E. Olshtain, G.R. Tucker and M. Waterbury (1979), ‘The acquisition of complex English structures by adult native speakers of Arabic and Hebrew’, Language Learning, 29(2): 255-275. Cromer, R.F. (1970), ‘“Children are nice to understand”: Surface structure clues for the recovery of a deep structure’, British Journal of Psychology, 61(3): 397-408. d’Anglejan, A. and G.R. Tucker (1975), ‘The acquisition of complex English structures by adult learners’, Language Learning, 25(2): 281-296. de Haan, P. (1997), ‘How ‘native-like’ are advanced learners of English?’, in: A. Renouf (ed.) Explorations in Corpus Linguistics. Amsterdam: Rodopi. 5565. Eckman, F. (1977a), ‘On the explanation of some typological facts about raising’, in: F. Eckman (ed.) Current Themes in Linguistics. New York: Halsted Press. 195-214. Eckman, F. (1977b), ‘Markedness and the contrastive analysis hypothesis’, Language Learning, 27: 315-330. Eckman, F. (1996), ‘A functional-typological approach to second language acquisition theory’, in: W. Ritchie and T. Bhatia (eds) Handbook of Second Language Acquisition. San Diego: Academic Press. 195-211. Eckman, F., E. Moravcsik and J. Wirth (1989), ‘Implicational universals and interrogative structures in the interlanguage of ESL learners’, Language Learning, 39: 173-205. Fisiak, J, M. LipiĔska-Grzegorek and T. Zabrocki (1978), An Introductory English-Polish Contrastive Grammar. Warszawa: PaĔstwowe Wydawnictwo Naukowe. Giacalone Ramat, A. (2003), ‘Introduction’, in: A. Giacalone Ramat (ed.) Typology and Second Language Acquisition. Berlin & New York: Mouton de Gruyter. 1-18. Givón, T. (1991), ‘Markedness in grammar: Distributional, communicative and cognitive correlates of syntactic structure’, Studies in Language, 15: 335370. Givón, T. (2001), Syntax. An introduction. Vol. 2. Amsterdam: John Benjamins Publishing Company. Granger, S. (1997), ‘On identifying the syntactic and discourse features of participle clauses in academic English: native and non-native writers compared’, in: J. Aarts, I. de Mönnink and H. Wekker (eds) Studies in English Language and Teaching. Amsterdam: Rodopi. 185-198. Granger, S. (2004), ‘Computer learner corpus research: Current status and future prospects’, in: U. Connor and T.A. Upton (eds) Applied Corpus Linguistics: A Multidimensional Perspective. Amsterdam: Rodopi. 123145. Granger, S., E. Dagneaux and F. Meunier (2002), The International Corpus of Learner English. Handbook and CD-ROM. Louvain-la-Neuve: Presses Universitaires de Louvain.

Raising constructions and information packaging in the advanced learner variety

225

Granger, S. and S. Tyson (1996), ‘Connector usage in the English essay writing of native and non-native EFL speakers of English’, World Englishes, 15: 19-29. Hawkins, J.A. (1986), A Comparative Typology of English and German: Unifying the Contrasts. Austin, London and Sydney: Croom Helm. Hinkel, E. (2002), Second Language Writers’ Text. Linguistic and Rhetorical Features. Mahwah, NJ: Erlbaum. Jordens, P. (1977), ‘Roles, grammatical intuitions and strategies in foreign language learning’, Interlanguage Studies Bulletin, 2: 5-76. Jordens, P. (1978), ‘Sprachspezifisch oder sprachneutral? Zur Anwendung einer Strategie im Fremdsprachenerwerb’, Deutsch als Fremdsprache, 15: 367370. Kellerman, E. (1979a), ‘The problem with difficulty’, Interlanguage Studies Bulletin, 4: 27-48. Kellerman, E. (1979b), ‘Transfer and non-transfer: Where we are now’, Studies in Second Language Acquisition, 2(1): 37-57. Kellerman, E. (1983), ‘Now you see it, now you don’t’, in: S.M. Gass and L. Selinker (eds) Language Transfer in Language Learning. Rowley, MA: Newbury House. 112-134. König, E. (1971), Adjectival Constructions in English and German: A Contrastive Analysis. Heidelberg: Groos. König, E. (1995), ‘Einzelaspekt: Raising’, in: R. Ahrens, W.-D. Bald and W. Hüllen (eds) Handbuch Englisch als Fremdsprache. Berlin: Erich Schmidt. 155-157. Kortmann, B. (1998), ‘Kontrastive Linguistik und Fremdsprachenunterricht’, in: W. Börner and K. Vogel (eds) Kontrast und Äquivalenz. Beiträge zu Sprachvergleich und Übersetzung. Tübingen: Narr. 136-167. Legenhausen, L. and G. Rohdenburg (1995), ‘Kontrastivierung ausgewählter Strukturen im Englischen und Deutschen’, in: R. Ahrens, W.-D. Bald and W. Hüllen (eds) Handbuch Englisch als Fremdsprache. Berlin: Erich Schmidt. 133-139. Lorenz, G. (1998), ‘Overstatement in advanced learners’ writing: Stylistic aspects of adjective intensification’, in: S. Granger (ed.) Learner English on Computer. London & New York: Addison Wesley Longman. 53-66. Lorenz, G. (1999), Adjective Intensification – Learners versus Native Speakers. A Corpus Study of Argumentative Writing. Amsterdam: Rodopi. Mair, C. (1987), ‘Tough-movement in present-day British English: A corpusbased study’, Studia Linguistica, 41(1): 59-71. Mair, C. (1990), Infinitival Complement Clauses in English. A Study of Syntax in Discourse. Cambridge: Cambridge University Press. Mair, C. (1992), ‘‘Raising’ in English and German: Formal explanation, functional explanation, or no explanation at all?’, in: C. Mair and M. Markus (eds) New Departures in Contrastive Linguistics. Vol. 1. Innsbruck: Verlag des Instituts für Sprachwissenschaft. 167-176.

226

Marcus Callies

Noël, D. (1997), ‘The choice between infinitives and that-clauses after believe’, English Language and Linguistics, 2(1): 271-284. Noël, D. (1998), ‘Infinitival complement clauses in English: Explaining the predominance of passive matrix verbs’, Linguistics, 36(6): 1045-1063. Odlin, T. (1989), Language Transfer. Cross-linguistic Influence in Language Learning. Cambridge: Cambridge University Press. Odlin, T. (2003), ‘Cross-linguistic influence’, in: C.J. Doughty and M.H. Long (eds) The Handbook of Second Language Acquisition. Oxford: Blackwell. 436-486. Quirk, R., S. Greenbaum, G. Leech and J. Svartvik (1985), A Comprehensive Grammar of the English Language. London: Longman. Rohdenburg, G. (1993), ‘Scheinbare Konstituenten und Verwandtes im Englischen’, Arbeiten aus Anglistik und Amerikanistik, 18(2): 249-271. Scott, M. (1999), WordSmith Tools 3.0. Oxford: Oxford University Press. Selinker, L. (1972), ‘Interlanguage’, International Review of Applied Linguistics, 10: 209-230. Trosberg, A. (1983), ‘The acquisition of some complex syntactic structures in L1 and L2 learners’, Rassegna Italiana da Linguistica Applicata, 2(3): 261283. Van Valin, R. (2001), An Introduction to Syntax. Cambridge: Cambridge University Press. Yamaoka, T. (1988), ‘A semantic and prototype discussion of the ‘be easy to V’ structure: A possible explanation of its acquisition process’, Applied Linguistics, 9(4): 385-401. Yip, V. (1995), Interlanguage and Learnability. Amsterdam: John Benjamins Publishing Company.

Thematic choice in the written English of advanced Spanish and Dutch learners Mike Hannay Vrije Universiteit, Amsterdam Elena Martínez Caro Universidad Complutense, Madrid Abstract This paper is about an important aspect of discourse competence in advanced learners of English. It reports the first results of an ICLE-based study of how Spanish and Dutch learners of English construct the theme zone of their sentences, defined as the complex of constituents up to and including the subject and its non-restrictive postmodifiers in the first declarative main clause of the sentence. The theme zone is interesting for discourseoriented learner research since thematic material links up with the previous content, gives an orientational frame for the message to come, and adds background information for understanding the upcoming message. Dutch and Spanish are interesting in terms of their basic clause patterns: with regard to the beginning of the clause, they differ from each other and they also differ from English. The results show that there is a large degree of similarity in how the two groups construct their sentences, but the Dutch group exhibits both a higher degree of grammatical competence and a richer development of discourse competence.

1.

Introduction

The application of corpus analysis to learner language, stimulated for English by the International Corpus of Learner English (ICLE) (Granger et al. 2002), has resulted in a number of studies oriented at discourse aspects of learner writing (e.g. Altenberg & Tapper 1998, Smits 2002).1 This paper extends interest in discourse aspects of learner language by considering the discourse-building potential of the theme, seen in the Hallidayan sense (cf. Halliday 2004: Ch. 3, Downing & Locke 2006: Ch. 6), in written English. It also makes use of a contrastive element by comparing the writing of two different groups of advanced learners of English, with Dutch and Spanish as their first language. Our interest in the developing discourse competence of these learners is geared firstly towards determining the extent to which the groups have acquired the structural patterns available in English, given that the basic patterns for clause building differ in these two languages precisely in the thematic area. And secondly, our interest is to develop a picture of students’ syntactic, semantic and pragmatic exploitation of the thematic area, which we will call the theme zone (Hannay 1994, forthcoming).

228

Mike Hannay and Elena Martínez Caro

Differences between the groups might be expected to indicate a measurement of developing discourse competence. One valuable application is in the area of language skills assessment using the Common European Framework of Reference (CEF). It has been noted, for instance, that users of the CEF should consider “how qualitative progress in the pragmatic component can be characterized” (Council of Europe 2001: 130). With regard to Dutch learners of English, there is evidence that first year university learners use the thematic zone in a rather limited way, while more advanced learners have learned to use specific patterns in a way which approximates the use by experienced native speaker writers (Smits 2002). Although what we report on here are just the first stages of a contrastive study, the approach we adopt aims to give a more encompassing view of students’ discourse competence by studying the range of syntactic structures applied by two groups of advanced learners to fulfil a range of semantic and pragmatic functions. 2.

A take on theme in English

The thematic area of the English text sentence is of considerable importance in terms of text building. On the one hand, thematic information provides an orientation for the interpretation of the content of the sentence to come. On the other hand, the beginning of the sentence is a typical location of constituents which forge cohesive ties with the previous context; connectives, for instance, typically, though by no means exclusively, occur in thematic positions. In the last few years there has been a considerable amount of learner corpus work into the use of connectives in different learner groups (e.g. Granger & Tyson 1996, Altenberg & Tapper 1998). However, we are not acquainted with any learner language research into theme which takes an all encompassing approach and which also contrasts learner languages. The basic approach to the scope of the theme in English is that it extends up to the first element in the experiential structure of the sentence (Halliday 2004: 66). However, there are many different approaches to the phenomenon (for different views, see for instance Berry 1996, and for an overview of the issues involved, see Butler 2003: 129ff.). The approach that we adopt here owes much to Downing (1991). For Downing, Freud in (1) belongs to the theme even though it is the second experiential element in the clause: (1)

Towards the end of his life, Freud concluded that he was not a great man, but he had discovered great things.

Downing distinguishes between the circumstantial frameworks set up by spatiotemporal and situational elements and individual frameworks set up by participants. This means that in declarative clauses with an S V structure the subject will be part of the theme. But we have also extended the scope of the theme further to include non-restrictive postmodification of the subject. The main

Thematic choice in written English

229

reason for this is that an earlier study (Hannay forthcoming) revealed that the functions of elements which non-restrictively postmodify the subject are essentially the same as those which modify pre-subject thematic elements, and can in general be characterised as contributing to the orientational part of the sentence. Consequently, what we will call the theme zone, following Hannay (1994, forthcoming) and Gómez-González (2004), extends up to, but does not include, will enable as finite verb of the main clause in (2): (2)

But if our lords and masters have their way, the proposed amendment to the defamation bill, which has just completed its passage to the Lords, will enable MPs to waive parliamentary privilege.

Before looking at students’ exploitation of the various options available in the theme zone, we need a more detailed characterisation of the structural patterns of English compared to Spanish and Dutch. 3.

The functional patterns of English, Spanish and Dutch

One motivation for our study came from a consideration of syntactic patterning in Spanish and English, within the framework of Functional Discourse Grammar (FDG) (Hannay & Martínez Caro forthcoming). FDG is not concerned with the basic word order patterns of languages in terms of some underlying constituent order, but rather with the partially functionally oriented patterns according to which sentences are actually constructed. The central mechanism in the grammar involves underlying structures of a pragmatic and semantic structure which, based on structural and functional specifications of the constituents concerned, are mapped onto a functional pattern, or template. Dutch and Spanish differ from each other in terms of their basic structural patterning and are both different from English in this regard. Our first aim was to investigate the extent to which the sentences constructed by advanced learners in running prose (a) made use of the patterns available in English, and (b) showed vestiges of L1. 3.1

English

The basic pattern for English declarative clauses is given in (3) (cf. Hannay & Martínez Caro forthcoming): (3)

P2, P1 S

Vf Vi O

X

PØ, P3

In this pattern, S stands for subject, Vf for finite verb, Vi for non-finite verb, O for object and X stands for non-subject, non-object constituents such as “oblique” arguments or all kinds of adjuncts (Dik 1997: 420). In an extended pattern there are also slots for all kinds of minor, non-argument constituents, but they do not concern us here (for a detailed account, see Connolly 1991). There are two pragmatically relevant positions within the clause and two outside it. Inside the

230

Mike Hannay and Elena Martínez Caro

clause pattern we have the initial position P1, which can only be filled by one constituent, and which may house all kinds of constituents with a special pragmatic status, such as Topic or Focus. Then there is a clause-final position PØ which under certain discourse conditions is home to Focus constituents. As far as the external positions are concerned, P2 is home to left-dislocated elements, while P3 houses so-called tail constituents which clarify or modify the content of the clause, although these final positions do not concern us here. The pattern is illustrated in (4): (4)

Clearly, in her next book Mary will have to take P1 S Vf Vi P2 the problem of obesity, won’t she? P3 PØ

into account X

In addition to the basic pattern it is important to recognise the value of two other functional patterns in English. First, we need to recognise a pattern for presentative structures. This is given in (5) and illustrated in (6). (5) (6)

P1 Vf Vi S a. On every table b. Along with this P1

had has Vf

been placed a ring of small vases. come a new fear of violence. Vi S

Finally, inversion structures in English follow the pattern of wh-interrogatives, given in (7) and exemplified in (8): (7) (8)

P1 Vf S Vi O X Not until today did Vf P1 had said.

he S

realise the importance of what he Vi O

The dominant P1 S V pattern is indeed very frequent. Inversion structures are limited to a relatively small set of lexical realisations and inversion plus presentative constructions are in general relatively rare in English (for figures concerning fronting in general, cf. Biber et al. 1999: 909; for figures concerning inversion in general, cf. Biber et al. 1999: 926). 3.2

Spanish

Martínez Caro (2006) argues that Spanish is best characterised as conforming to two basic clause patterns. The first of these is given in (9): (9)

P2, P1 c

V

S

O

X

PØ, P3

Small c stands for clitics and other set particles like the passive or impersonal se, all of which occur in Spanish in the immediately preverbal position. The basic

Thematic choice in written English

231

clause pattern in (9) is assumed to capture the following main types of constructions in Spanish: (a) SVO orderings, the dominant clause order in Spanish, where the subject is placed in P1 by virtue of its status as Topic (cf. Dik 1997: 409); (b) constructions with a fronted non-subject constituent in P1, the subject either being presented postverbally or left unmentioned; and (c) presentative constructions in clauses which present a fronted constituent functioning as the setting. The examples in (10-12) illustrate the use of these constructions: (10)

Estos diez libros contienen algunas claves para vivir mejor. (HV, 15)2 some keys to live better These ten books contain:3PL V O P1 ‘These ten books contain some clues for a better life’

(11)

Lo mismo The same P1 ‘I think the same’

(12)

por eso apareció el aburrimiento. (HV, 25) appeared:3SG the boredom for that:N V S P1 ‘For that reason, boredom appeared.’

pienso think:1SG V

yo. (HV, 22) I S

The P1 V S ordering in (12) is common not only with presentatives but also with experiencer verb constructions and se-passives. Note further that in verb-initial clauses in Spanish the special position P1 is assumed to be left empty and the verb occupies its standard slot position, as in (13): (13) Existen tres mil especialidades matemáticas. (HV, 95) three thousand specialisations mathematical.PL Exist:3PL V S ‘There exist three thousand specialisations within mathematics.’ Finally, a characteristic feature of Spanish (and some other Romance languages, like Italian and Portuguese) is its ability to omit the subject when this represents given information or when the morphological encoding of the verb makes its presence unnecessary. This is illustrated in (14): (14)

Creo que las mujeres valoran menos Think.1SG that the women value less V O que los hombres. (HV, 22) than the men

el ingenio the wit

232

Mike Hannay and Elena Martínez Caro ‘I think that women value wit to a lesser extent than men do.’

In order to account for those constructions in Spanish which maintain an S V order with the fronting of another constituent (adjunct or non-subject argument) in P1, the following second basic clause pattern is proposed: (15)

P2, P1 S

c

V

O

X

PØ, P3

Thus, the examples in (16-17) may be captured by the pattern in (15): (16)

Para nosotros eso es afectación, hipocresía y falta de sinceridad. For us that is pretension, hypocrisy and lack of sincerity. S V X P1 ‘For us that is pretension, hypocrisy and lack of sincerity.’ (HV, 124)

(17) Como ya he dicho, en Galicia uno de los platos típicos es As already have.1SG said, in Galicia one of the dishes typical.PL is P1 S V P2 el pulpo. the octopus X ‘As I already said, in Galicia one of the typical dishes is octopus.’ Note that this cannot be analysed as an example of a P1 V S structure. 3.3

Dutch

In Dutch, independent declarative clauses are formed according to one functional pattern which captures all word order variation. This is given in (18): (18)

P2, P1 Vf S

O

X, P3

Dutch is very much a verb-second language. Although there is relative freedom in the filling of P1, this position very often houses the grammatical subject. And whereas in English initial positioning of conjuncts, disjuncts and adverbial clauses may relate to the P2 slot and thereby still leave space for a P1 constituent, this is much less the case in Dutch. The verb-second status of Dutch is exemplified in (19), where the initial adverbial subclause is followed by the finite verb of the main clause: (19)

Toen hij arriveerde heeft zij haar zus When he arrived has she her sister Vf S O P1 ‘When he arrived, she phoned her sister’

gebeld. phoned X

Thematic choice in written English

233

However, it should be noted that Dutch in fact has the same functional options for these structures as English has (cf. Hasereyn et al. 1997, Smits 2002: 169). Consider (20-21), taken from Smits (2002: 171-172): (20)

Integendeel: qua kansen (op arbeid, inkomen of onderwijs) On the contrary as for chances (of work, income or education) P1 P2 was er eerder van verbetering dan van verslechtering sprake. was there rather of improvement than of deterioration talk Vf S ‘On the contrary, with respect to opportunities (for jobs, income or education) the situation improved rather than worsened.’

(21)

Anders gezegd, opnieuw is het moreel gehalte van het Amerikaanse Put differently, again is the moral content of American P1 Vf S P2 buitenlandse beleid in het geding. foreign policy at issue X ‘Put differently, again the moral content of American foreign policy is at issue.’

Here both P2 and P1 are filled, and in both cases by adverbial constituents. Nevertheless, structures with an adverbial expression in P2 remain relatively rare in written Dutch. 3.4

Comparison

What can we conclude from this basic presentation of the functional patterns in the three languages? First of all, as far as the target language of English is concerned, the advanced learner should feel fully confident with the dominant pattern but will also have to be aware of the options which the inversion and presentative patterns offer, in terms of variation and communicative effect. As far as the dominant pattern is concerned, it is also relevant that English is a potentially front-heavy language when it comes to constructing text sentences. That is to say, it is not uncommon for two, three or more constituents to occur before the subject. By contrast, a verb-second language will clearly tend to have a limited number of constituents before the verb. With regard to Dutch learners, one might expect the strong S V orientation in the secondary school teaching of English to have become well entrenched. At the same time, given the relative freedom that Dutch has in the syntactic filling of the P1 position, one might expect that the advanced learner will make considerable use of P1 S V structures when producing the kind of formal, argued text that we are looking at here. However, there might be only limited use of P2 P1 S V structures since it is very unusual in Dutch to have even two elements before the finite verb.

234

Mike Hannay and Elena Martínez Caro

As for Spanish learners, one might again expect the strong S V orientation of English language teaching to be well entrenched, but a difficulty compared to Dutch lies in the use of the subject. First of all, Spanish does not have a dummy subject and does not always require the understood subject to be specified. Second, the position of the subject before or after the verb is highly dependent on its pragmatic status. On the other hand, one would expect no major problems with mastering P2 P1 S V structures: first, P1 S V structures occur in Spanish as well, for instance in the case of object fronting; and second, Spanish also allows structures with two elements before the preverbal subject, as illustrated in (17) above. 4.

Data and method of analysis

For the purposes of the present study we compiled two subcorpora from the ICLE corpus, using the parameters given in Table 1.3 Table 1. Parameters for defining the subcorpora Text type Length Sex Country Native language Age Second language at home Years of English at university Timing Reference tools Examination

argumentative 400 to 1,500 words all Spain, the Netherlands Spanish, Dutch max. 25 none max. 2 no timing yes no

This produced 63 Spanish texts with a total of 1,904 sentences (approx. 44,000 words) and 45 Dutch texts with a total of 2,153 sentences (approx. 45,000 words). All the sentences were entered into an Access database. We coded for the clause type of the core communicative unit in each sentence, distinguishing declarative, imperative, interrogative and a rest category. This latter category includes presentatives, fragments and ungrammatical sentences where the clause type of the communicative core was not formally identifiable by means of the relevant sequence of mood elements. An example of the latter is (22): (22)

When someone has a bit money usually will be an ambitious person. (ICLE-SP-UCM-0020.3)4

We return to cases like this in Section 6.3 below. The overall figures for clause type are given in Table 2.

Thematic choice in written English

235

Table 2. Sentences in the Dutch and Spanish subcorpora according to clause type Dutch learners Spanish learners

Declarative 2,058 (95.6%) 1,784 (93.7%)

Interrogative 29 (1.3%) 34 (1.8%)

Imperative 8 (0.4%) 11 (0.6%)

Rest 58 (2.7%) 75 (3.9%)

Total 2,153 (100%) 1,904 (100%)

We also coded for a variety of syntactic features occurring in the theme zone. These are given in Table 3. Table 3. Syntactic features in the theme zone Code Ac Ap Cc Ci Cp Cr Np Ob Pc Qq Sd Sn Sp St Vb Wh Xx Zz 5.

Meaning adjunct or conjunct or disjunct consisting of a clause adjunct or conjunct or disjunct consisting of a phrase comment clause independent clause participial clause non-restrictive relative clause independent noun phrase object predicate complement apposition subject with dummy it referential noun phrase as subject subject with pronominal head there as existential subject finite verb in interrogative clauses or verb in imperative clauses wh- word in interrogative clauses absolute construction Conjunction Thematic patterns in sentences with declarative main clauses

One way of seeing how writers in the two groups exploited thematic potential in sentences with declarative main clauses is to consider all the different patterns developed. Table 4 gives an overview of the top 10 in each group, according to frequency. In determining the patterns, sub-specifications of individual categories were disregarded. To illustrate, the category ASQ means that (a) the initial element A is an adjunct, conjunct or disjunct consisting of a clause or phrase, followed by (b) a grammatical subject S which may be either a full NP, a pronoun, dummy it or existential there, which in turn is followed by (c) an apposition Q.

236

Mike Hannay and Elena Martínez Caro

Table 4. The top 10 thematic patterns in sentences with declarative main clauses in the Dutch and Spanish subcorpora 1 2 3 4 5 6 7 8 9 10 Rest Total no. of declaratives

Dutch learners S 1,116 (54.23%) AS 661 (32.12%) AAS 63 (3.06%) ZS 55 (2.67%) ZAS 39 (1.90%) SA 27 (1.31%) SQ 17 (0.83%) ASQ 10 (0.49%) CS 9 (0.44%) AAAS / SC 7 (0.34%) 54 (2.62%) 2,058 (100%)

Spanish learners S 967 (54.20%) AS 541 (30.33%) ZS 74 (4.15%) AAS 60 (3.36%) ZAS 33 (1.85%) SQ 16 (0.90%) SA 15 (0.84%) CS 13 (0.73%) ACS 9 (0.50%) AAAS / SC 6 (0.34%) 50 (2.80%) 1,784 (100%)

Table 4 shows that nine of the top ten patterns in each subcorpus are the same, the exception being ASQ (adverbial-subject-apposition) in the Dutch subcorpus and ACS (adverbial-participial clause/relative clause/comment clause-subject) in the Spanish subcorpus, and that the relative frequency of these top patterns also does not differ very much. In general, then, Spanish and Dutch learners appear to behave very similarly in terms of thematic patterning. Another way of looking at the patterns is to distinguish across the whole subcorpora according to levels of complexity. We distinguished three levels. Level 1 involves no thematic material before the subject, with the exception of conjunctions, and includes any postmodification of the subject. Level 2 involves one adverbial or nominal element such as an object before the subject, with the possible addition of a conjunction, and again including any postmodification. Finally, level 3 involves two or more elements before the subject, again with the possible addition of a conjunction and a form of postmodification. The three levels are exemplified in (23-25): (23)

(24)

(25)

The ONU (Organization of United Nations) claims that all men and women are equal, that is, that they have the same rights. (ICLE-SP-UCM0008.5) To break this vicious circle, a plan should be framed to integrate delinquents in society by helping them to get jobs. (ICLE-DB-KVH0006.3) However, if we consider Ritzen’s idea more closely, it becomes clear that lecturing in English at Dutch Universities has many advantages and that the disadvantages do not weigh up to them. (ICLE-DN-NIJ-0002.7)

The categorisation according to levels gives the following results:

Thematic choice in written English

237

Table 5. Theme patterns according to level Level 1 Level 2 Level 3 Total

Dutch texts 1,221 (59.3%) 751 (36.5%) 86 (4.2%) 2,058 (100%)

Spanish texts 1,076 (60.3%) 605 (33.9%) 103 (5.8%) 1,784 (100%)

The distribution of the three levels over the two languages is statistically significant (chi square, p”0.05), but the differences actually appear quite small. A closer look at the standardised residuals and the lambda value reveals such low counts that there is a considerable likelihood that the significance found in fact comes from the sample size.5 Note at least that if we collate levels 2 and 3 so as to compare the languages simply in terms of subject-initial vs. non-subject-initial sentences, then there is indeed no significant difference between the two. Table 5 provides confirmation of what we found in Table 4 and thus yields a first important observation, namely that the two language groups are highly comparable in their use of the theme zone. However, we will see in Sections 6, 7 and 8 that when one looks in more detail, significant differences emerge. The next step is to compare the frequencies for structural patterns with frequencies for English. Although comparison with native speaker learners with similar writing experience is generally preferred (cf. Granger 1996), here we compiled a small sample of English texts in five genres for comparative purposes, since our aim is to provide a picture of developing discourse competence in the context of professional competence. The sample contains 39 texts taken from five different genres, in total 2,627 sentences, as further specified in Table 6. The comparison is given in Table 7. Table 6. Sample of professional English texts, according to genre Text type Political commentaries Charity mail Academic text Literary text In-flight magazine articles Total

No. of texts 9 17 4 4 5 39

No. of sentences 424 430 912 659 202 2,627

With the addition of native speaker texts we now see that the distribution becomes highly significant (chi square, p”0.001). The strongest difference lies in the comparatively higher use of level 1 themes by native speakers, whereas Dutch and Spanish learners both exhibit a higher use of level 2 themes than native speakers. What Table 7 also shows is that learners’ exploitation of thematic potential for level 3 is similar to the exploitation in the native speaker texts and for level 2 even exceeds it. It should, however, be noted that there is substantial

238

Mike Hannay and Elena Martínez Caro

variation amongst the different genres of the native speaker texts, and more in general, the differences between the learner texts and the native texts could possibly be attributable to genre differences. Nevertheless, all of this does suggest in general something like full acquisition by both learner groups of the structural means available for thematic development within at least the main functional pattern for English declaratives. Table 7. Thematic complexity according to level in Dutch and Spanish learners’ texts as compared to native speaker writers Level 1 Level 2 Level 3 Total no. of declaratives

Native texts 1,746 (67.38%) 734 (28.32%) 111 (4.28%) 2,591 (100%)

Dutch texts 1,221 (59.3%) 751 (36.5%) 86 (4.2%) 2,058 (100%)

Spanish texts 1,076 (60.3%) 605 (33.9%) 103 (5.8%) 1,784 (100%)

However, we would now like to go into further detail to investigate possible differences between the Dutch and Spanish groups. In Section 6 we look more closely at formal correctness in the use of all the declarative patterns. Then in Sections 7 and 8 we look at syntactic, semantic and pragmatic features of the themes employed by the two groups. For present purposes we have left out of consideration the exploitation of theme at level 1, since this is essentially a matter of postmodification of the subject. 6.

The acquisition of functional patterns

We start off this section by looking at the use of presentative constructions and constructions requiring inversion. These are structures which do not fit the dominant declarative pattern for English. Then we go on to consider a number of ungrammatical structures which for various reasons do not fit the functional pattern for declaratives but which the writers probably intended as declaratives. 6.1

Presentatives

For present purposes we used the label “presentative” to cover focus constructions involving subject-verb inversion with either a fronted adverbial, commonly locative, as in (26), or with a fronted predicative expression, as in (2728): (26) (27) (28)

On one long wall hung a row of Van Goghs. (Biber et al. 1999: 912) Also billed to appear as a special mystery guest is Vivacious Val. (Biber et al. 1999: 906) Far more serious were the severe head injuries, in particular a bruising of the brain. (Biber et al. 1999: 902)

Thematic choice in written English

239

There are a small number of presentatives in each learner group: 10 in the Dutch group and 3 in the Spanish group. None of the Spanish examples is well-formed, in that the P1 V S pattern of Spanish is transferred incorrectly to English. The three examples are given below: (29)

(30)

(31)

Thanks to science and its consecuences, technology and industrialisation, appeared the big factories and the capitalism system. (ICLE-SP-UCM-0007.3) Because of this, in some countries such as the United States (not in all the states, only in some of them) was passed the death penalty. (ICLESP-UCM-0041.4) In Western Society dominated a different vision of religion. (ICLE-SPUCM-0035.4)

The nature of the verb and the pragmatic status of the subject would admittedly allow a presentative structure in English, but only with the dummy subject there, thus following the dominant syntactic pattern and not the presentative one. And even then, the verbs in these examples are by no means the most common verbs in presentative constructions, which most typically contain copula be or locative existential verbs like lie and stand, and might only be expected to occur in quite formal or literary contexts. Moreover, (29) does not have an initial locative adverbial, and both (29) and (30) have definite subject noun phrases. All in all, these cases lack a number of typical features of presentatives. Of the Dutch cases, seven are well-formed and are used appropriately in the context. Examples are (32-34): (32)

(33)

(34)

But then came technology, because of machines the work became much lighter, people could go on holiday one in while and had plenty of time to fantasize and dream. (ICLE-DB-KVH-0037.3) At one end are a few highly selective colleges that admit only the most highly qualified students; at the other end most community colleges practise open admission […] (ICLE-DN-AVU-0016.1) Closely related to this is the high concentration of drug users. (ICLE-DNNIJ-0005.7)

While Dutch has a constructional equivalent of (32), the other two examples cannot be translated into Dutch without adapting the syntax or changing verb selection. This suggests that the students concerned have a command of the English presentative construction and the functional pattern involved. However, there are also three presentatives with a fronted predicative element, all produced by the same student, which, while appearing to be well-formed, do not in fact fulfil the requirements for presentatives. These are presented in (35-37): (35)

Below mentioned will be people who are incurably ill and people who are unable to make the final decision themselves. (ICLE-DN-AVU-0013.1)

240

Mike Hannay and Elena Martínez Caro

(36)

Secondly, discussed below will be the patient who is in coma after, for example, having had an accident or a celebral hemorrhage. (ICLE-DNAVU-0013.1) Meant here are aged people who can no longer take care of themselves and are no longer able to communicate with the normal and outer world. (ICLE-DN-AVU-0013.1)

(37)

In English, if the initial element in presentative constructions is a predicative phrase rather than an adverbial, then it must contain an explicit element which ties that element to the previous context, whereas in Dutch this is not necessary. Thus while (35) and (36) are out, (35’) and (36’) are fine, with adjustment of the adverb position in (35’) and also providing the cohesive tie needed. (35’) Also mentioned below will be people who are incurably ill and people who are unable to make the final decision themselves. (36’) Also discussed below will be the patient who is in coma after, for example, having had an accident or a celebral hemorrhage. 6.2

Constructions requiring subject-operator inversion

There are a very small number of constructions occurring in the subcorpus which require inversion of the subject and the operator. There are two cases in the Dutch data, both of which are formed correctly, and three in the Spanish data, one of which is formed correctly and two of which are not. We will restrict our comments to the latter. They are given in (38-39): (38)

(39)

Not only, in modern life, there is not too much spare time for imagination and dreams but the technology have replaced it too. (ICLE-SP-UCM0007.3) Not always the penalty is served totally – reduction of penalties. (ICLESP-UCM-0036.4)

In both cases the writers concerned have not picked up on the inversion signal given by the initial negative element, and have applied the dominant declarative pattern rather than the interrogative pattern. 6.3

Declarative misfits in the Spanish data

The Spanish data include three categories of ungrammatical sequences which can best be analysed in terms of incomplete acquisition of the English functional pattern. By contrast, the Dutch data do not contain any such misfits. For each of the three categories in the Spanish data, there seems to be interference from L1. First, there are nine cases where the writer presumably intended a declarative structure but did not produce an S V sequence. Examples are given in (40-43):

Thematic choice in written English

241

(40)

Finally must be added that in our days it is necessary for a country to be provided with a good army […] (ICLE-SP-UCM-0013.4) (41) Talking about the rehabilitation is important to consider two points: the first one is the kind of crime the person has been condemned by, and the second one, the society in which this subject is situated. (ICLE-SPUCM-0018.4) (42) When someone has a bit money usually will be an ambitious person. (= ex. 22) (ICLE-SP-UCM-0020.3) (43) In consequence of this inactiveness, together with the fact that we possess everything we need, does not allow us to develop our power of dreaming or imagination, as we can see in the next instances (ICLE-SPUCM-0003.5) These are sentences which simply contain no grammatical subject in the main clause. In (40-41) English requires dummy it, while in (42-43) the intended subject is in fact mentioned in the initial constituent (someone and this activeness). These structures appear to have been transferred from Spanish: the Spanish equivalents of (40-41) indeed require no grammatical dummy and the equivalents of (42-43) would not need a specified subject. The Spanish data also contained a second category of ungrammatical declaratives, with 18 examples. In this second category, students succeeded in producing an S V sequence but repeated a subject phrase later in the clause. We concluded that this constituted a lack of acquisition of the basic functional pattern. Here are some of the examples: (44)

(45) (46)

In my opinion it is very logic the idea of having voluntary soldiers in the army since a citizen can have a lot of reasons for not wanting to carry a gun in his hands. (ICLE-SP-UCM-0004.3) It is not strange for us the scene of a family dining watching TV and without speaking. (ICLE-SP-UCM-0007.3) When objectifying the economic compensation for one’s life (murder) or for the physical integrity of the person (injuries), - objectivity which is quantified through ready reckiners,- it is not taken into account the significance of the subjective elements that the victim gives to what he no longer owns. (ICLE-SP-UCM-0027.3)

The Spanish learners seem to transfer the postverbal subject of the Spanish construction incorrectly in these sentences, and once they have done so, they apply the rule of obligatory subject in English by filling in the preverbal slot with dummy it, as in extrapositions. The third and final category of misfits in the Spanish data relates to the occurrence of thematic structures containing independent noun phrases. This category is illustrated in (47-49):

242

Mike Hannay and Elena Martínez Caro

(47)

Children instead of reading books, they play computer games. (ICLESP-UCM-0003.5) Besides rich men when they have a family and, I think, that they are tired of them they look for a young pretty woman, destroying the family and besides these women are in love with the money but not with that man. (ICLE-SP-UCM-0037.3) Under these circumstances, the periods of time that children wasted “sticked to the box”, other generations may have devoted them to traditional games, going for a walk or even studying any subject. (ICLESP-UCM-0034.3)

(48)

(49)

Of the 11 cases of independent NPs found, almost half showed a common pattern where the independent NP and its co-referent pronoun are split up by a subordinate clause modifying the initial NP. The first two examples represent the standard case, where the pronoun is the subject of the main clause, but in the case of (49) it is the object.6 7.

Exploitation of thematic potential: level 2

In the first stage of our research we have chosen to concentrate on the use of clauses at level 2 rather than phrases. A pilot study of Dutch university students of English in their first year of studies revealed that writers at this point of development showed an elaborational style rather than an orientational style with regard to the use of subordination; in other words, subordinate clauses tended to follow the main clause rather than precede it, as in (50 vs. 51). (50) (51)

Politicians tend to tell lies, because they want to keep everyone happy. (elaborational) Because politicians want to keep everyone happy, they tend to tell lies. (orientational)

Moreover, there was a very limited range of use in terms of semantic function, with conditional clauses being prominent. In this study we wanted to see how much the thematic potential in this area was exploited by a larger group of learners in a broader band with up to two years of study behind them. First of all, Table 8 distinguishes between the use of clauses with finite and non-finite verb forms. It shows that the distribution of finite and non-finite forms is very similar in the two groups. Table 8. Finite and non-finite adverbial clauses at level 2 Finite Non-finite Total

Dutch 142 (69.3%) 63 (30.7%) 205 (100%)

Spanish 85 (68.0%) 40 (32.0%) 125 (100%)

Thematic choice in written English

243

Next, Table 9 presents an overview of clauses in terms of representational function vs. textual and attitudinal function. What is interesting here is the significantly greater use of textual and attitudinal elements in the Spanish data (chi square, p”0.001). Examples of textual, attitudinal and representational elements are given in (52) to (54). Table 9. Adverbial clauses within level 2 according to representational vs. textual/attitudinal function Representational Textual / attitudinal Total (52) (53)

(54)

Dutch 180 (87.8%) 25 (12.2%) 205 (100%)

Spanish 93 (74.4%) 32 (25.6%) 125 (100%)

TEXTUAL: To sum up, there is still something more to add. (ICLE-SPUCM-0028.4) ATTITUDINAL: As we all know, Military service requires masculine population in order to do one’s military instruction. (ICLE-SP-UCM0016.4) REPRESENTATIONAL: If two governors of different countries want to dispute something like a part of a land, they should solve the problem themselves without involving the rest of the population. (ICLE-SP-UCM0004.3)

Interestingly, the Spanish data include eleven cases of the textual element to sum up, which on the one hand is not a common way of introducing a summarising section in English academic writing, but which on the other hand nevertheless suggests that cohesion-oriented writing instruction might be partly responsible for the figures for Spanish. But a fuller understanding of how initial adverbial elements are used would require us to consider non-clausal expressions as well, and that is outside the scope of the present paper. The use of expressions like the non-finite clause in (52) suggests that there may be a relation between the form of the subordinate clause and its function. Table 10 shows how finite clauses were used by the two groups. More precisely, it shows that both groups use the vast majority of finite clauses for representational purposes, as one would expect from Table 9, but nevertheless that Spanish students use finite clauses significantly more often than Dutch students with textual and attitudinal functions (p”0.025). Table 11 shows how non-finite clauses were used. It appears that both groups use the majority of non-finite clauses for representational purposes as well, but again that Spanish students use such clauses significantly more often than Dutch students with textual and attitudinal functions (p”0.05). We conclude that the significantly greater use of textual and attitudinal clauses by Spanish students is reflected in their use of both finite and non-finite clause forms.

244

Mike Hannay and Elena Martínez Caro

Table 10. Finite adverbial clauses according to function Finite clauses with textual + attitudinal function Finite clauses with representational function Total

Dutch 11 (7.7%)

Spanish 15 (17.6%)

131 (92.3%)

70 (82.4%)

142 (100%)

85 (100%)

Table 11. Non-finite adverbial clauses according to function Non-finite clauses with textual + attitudinal function Non-finite clauses with representational function Total

Dutch 14 (22.2%)

Spanish 17 (42.5%)

49 (77.8%)

23 (57.5%)

63 (100%)

40 (100%)

We now turn briefly to the major group at level 2 in terms of function, namely representational themes. The make-up of this category in terms of semantic function is given in Table 12. Table 12. Classification of representational adverbial clauses at level 2 according to semantic function Conditional Temporal Concessive Circumstantial Purpose Reason Contrast Conditional-concessive7 Additive Manner Total

Dutch 44 (24.4%) 43 (23.9%) 28 (15.6%) 23 (12.7%) 15 (8.3%) 13 (7.2%) 5 (2.8%) 4 (2.2%) 4 (2.2%) 1 (0.6%) 180 (100%)

Spanish 31 (33.3%) 27 (29.0%) 7 (7.5%) 16 (17.2%) 7 (7.5%) 2 (2.2%) 2 (2.2%) 0 (0.0%) 0 (0.0%) 1 (1.1%) 93 (100%)

Although some of the functions here are used too infrequently to make statistical testing possible, a number of observations can be made. First, we see that temporal and conditional functions are predominant in both learner groups, although in Spanish they account together for 62.3% of all the data while in Dutch they account for only 48.3%. Second, we see that the functions of concessive, circumstantial and purpose all figure well in both groups too, accounting for 36.6% of the Dutch data and 32.2% of the Spanish data. Third,

Thematic choice in written English

245

there is a slightly wider range of functions employed by the Dutch learners, with two functions occurring in the Dutch data which do not occur in the Spanish data (conditional-concessive and additive). Finally, Table 13 gives a breakdown of the different formal types of nonfinite clauses used by the two groups. Table 13. Types of non-finite adverbial clauses Conjunction + participle Infinitive Participle Reduced Total

Dutch 28 (44.4%) 21 (33.3%) 12 (19.0%) 2 (3.2%) 63 (100%)

Spanish 5 (12.5%) 20 (50.0%) 15 (37.5%) 0 (0.0%) 40 (100%)

Although some of the cells here are also too small to test for significance, there are some clear differences. First, half of the non-finite clauses in the Spanish data are infinitive clauses, whereas this is true for only a third of the Dutch data. When we look at what these infinitive clauses are doing, we see that they are predominantly used to express textual functions in the Spanish data, as noted above in relation to example (52), and are used with a representational function in the Dutch data. Second, Dutch writers use many more participial clauses with conjunctions, as in (55-56): (55) (56)

Besides being expensive, the voyage was full of hardships. (ICLE-DNAVU-0005.1) If applied correctly, positive action will abolish itself. (ICLE-DN-GRO0003.1)

Third, the Dutch data include two cases of reduced clauses, illustrated in (57): (57)

If not, the family has to decide what is best to do. (ICLE-DN-AVU0013.1)

The use of these last two clause types permits a wider range of semantic functions than simple participial clauses, and given the variation that we also saw in the Dutch texts in terms of semantic function, this suggests that Dutch learners may have a somewhat wider range of application at level 2, at least for clauses. We will come back to all these points below when discussing the implications of the analysis, but first we need to look at level 3.

246

Mike Hannay and Elena Martínez Caro

8.

Exploitation of thematic potential: level 3

Level 3 involves cases where the subject is preceded by at least two other elements, excluding conjunctions. The major combinations are of two adverbial elements. Smits (2002) conducted a wide-scale analysis of such complex beginnings in English, Dutch and Dutch learner-English texts, finding a total of 496 examples in the English native speaker corpus (for more discussion of such complex thematic patterns, see Gómez-González 1998).8 All cases could be analysed as belonging to one of five patterns. Here is a brief overview, with percentages for the frequency of each pattern in the English native speaker corpus: Stepwise The most frequent pattern is called Stepwise (38.7%). Here, the second expression provides an orientation for the main clause, but both the orientation and the clause fall inside the scope of the first orientational expression; crucially, the second expression does not modify the first, as in (58): (58)

On the St Petersburg waterfront, if you don’t pay off the right people, you may find that the crane operator will drop your cargo in the water. (Smits 2002: 76)

Complex Almost as common as the Stepwise pattern is the Complex pattern (33.8%). Here, unlike Stepwise, the second expression modifies the first, and the two orientations together provide a complex orientation for what is to follow. Crucially, the second orientation provides an instruction to the reader to interpret the element concerned in terms of the first element; as such, the second element has a kind of focalising function, as in (59): (59)

By the early 70’s, however, this attitude was changing […] (Smits 2002: 84)

Grounded 20% of all cases belong to the Grounded pattern. With this pattern, the second expression further specifies the first, often by mentioning an event which allows the reader to gain a greater understanding of the relevance of the particular setting for understanding the message to come. An example is (60): (60)

Later, with England converted to Christianity, the daughters of the great Anglo-Saxon noblemen were sent abroad to France to be educated. (Smits 2002: 84)

Thematic choice in written English

247

Composite Then we have the Composite pattern (4.8%), in which typically time and place adjuncts are fused, as exemplified in (61): (61)

Two years ago in Dublin I said if you don’t have something that is perceived to be inclusive, you’ve had it. (Smits 2002: 84)

Compound The least frequent pattern is the Compound pattern (3.4%), which involves two pre-subject elements which could be coordinated, as in (62): (62)

Obediently, unquestioningly, Wang does the move again. (Smits 2002: 64)

For present purposes we selected all cases of level three structures which involved adverbials in the first two positions. This gave us 69 of the 86 level 3 cases in Dutch and 63 of the 103 level 3 cases in Spanish. The distribution over Smits’s five types of complex beginnings is given in Table 14. Table 14. Patterns of complex beginnings in Dutch and Spanish texts Stepwise Complex Grounded Composite Compound Total

Dutch 51 (73.9%) 8 (11.6%) 6 (8.7%) 2 (2.9%) 2 (2.9%) 69 (100%)

Spanish 53 (84.1%) 0 (0.0%) 7 (11.1%) 2 (3.2%) 1 (1.6%) 63 (100%)

The first thing that stands out here is the great predominance in both groups of the Stepwise category. This agrees with the findings of Smits (2002) for first and second year Dutch learners, and Spanish learners show the same tendency. The second point is the occurrence of the specific subtype of Complex beginnings in the Dutch data but not the Spanish. Smits’s analysis shows that Dutch learners in their third and fourth years of university study use more complex beginnings than first and second year learners, and indeed start to approximate native writers in terms of frequency. The complex type is argued to be cognitively more demanding to produce, and the experiments conducted by Smits permit the conclusion that it is discourse competence rather than language competence (Smits 2002: 312) which explains the relative underuse of the complex type amongst advanced learners.9

248

Mike Hannay and Elena Martínez Caro

9.

Discussion

If we look at the outcomes of the different frequency comparisons, then a general picture of the two groups of learners begins to emerge with regard to how they use the thematic potential of the English sentence. The backdrop against which we would like to view the differences between the two language groups is formed by the overall similarity in the way that the two groups of learners use the thematic structures available in English: there is comparatively little difference in the frequency of structures belonging to the three thematic levels which we distinguished. When compared to the structures used in professional native speaker texts, it appeared that both learner groups overused level 2 structures and underused level 1 structures, which suggests more complex structures are being applied by the learners. This is reminiscent of findings in other learner language research relating to the overuse of connectives in initial position (cf. Altenberg & Tapper 1998), which may be strongly instruction-driven. But in the broader sense this overuse of level 2 structures may well be L1-driven: in Dutch, adverbials appear more often in initial position than in English, with the subject occurring post-verbally (Smits 2002: 175), while the dominant functional pattern in Spanish also allows postverbal subjects. As far as acquisition is concerned, Dutch learners at this stage appear to have mastered the use of the presentative construction in English, both in terms of its structural constraints and in terms of its contextual conditions of use, although it must be noted that our subcorpus contained relatively few data. The problematic cases encountered were all produced by the same individual learner, and were all due to a lack of understanding of a particular information structure requirement. Spanish learners also use presentative-like constructions, but on an even more limited scale than Dutch learners, and all the examples in the corpus were ill-formed. The development of the Spanish learners’ grammatical competence in this respect would appear to be hampered by the lack of a full understanding of the semantic aspects of the construction. In addition to illformed presentatives, we also saw that the Spanish learners’ texts contained 27 instances (1.51% of sentences with a declarative main clause) of an ungrammatical declarative structure, with possible transfer from L1. As a group, the Spanish learners have therefore not achieved the same high degree of acquisition as the Dutch learners. We can form a better picture of the development of discourse competence when looking at thematic exploitation at level 2 and level 3. In terms of discourse functions, Spanish learners use clausal structures at level 2 much more than Dutch learners, while Dutch learners exhibit a broader range of subclauses both in syntactic and semantic terms. With regard to level 3, we see a strong resemblance in the patterns adopted by the two groups, with the stepwise category standing out. However, the fact that the Spanish learners’ texts do not contain any examples of the complex subtype again suggests that they have not developed the same richness of functionalities that the Dutch learners have, although this must

Thematic choice in written English

249

be a tentative conclusion since we are dealing with a small number of data. The picture that emerges from level 2 and level 3 analysis is thus similar to that for acquisition of structural patterns. The one unusual aspect here is the comparative overuse by Spanish learners of adverbial clauses with a textual and attitudinal function. As we suggested above, however, these constructions are dominated by the presence of a small number of fixed expressions such as to sum up, which may result from an instructional focus on cohesive elements in writing. Overall, the Dutch learner group thus appears to make rather more elaborate use of the thematic options of English, seen in terms of the structural possibilities, the related grammatical construction types, and the semantic and rhetorical exploitation of the structural options. Overall, we tentatively conclude from the analysis that (a) Spanish learners need to pay more attention to developing the grammatical competence involved in producing the dominant functional pattern, and (b) both groups would benefit from writing courses which pay attention to the theme zone in terms of both functionality and syntactic realisation. This latter conclusion constitutes strong support for the “syntax for writing” approach discussed and promoted for instance by Hinkel (2002, 2004). Such an approach may mean paying attention to structures which are not always looked at in any great detail in pedagogical grammars, but perhaps more importantly, it is an approach which allows learners to develop an understanding of how grammatical structures function in running prose. 10.

Conclusion

The theme zone as defined in this study is an area of the English text sentence where writers can apply a range of syntactic devices to perform important orientational tasks for the message at hand. First of all there is the choice of basic syntactic pattern, but there is also the matter of how to use the syntactic structures, in particular those before the subject, to express a variety of semantic and textual relations. The more richly one can exploit the options available, the more developed one’s discourse competence is. The analysis we have presented is only a first step in considering the data from the ICLE corpus in terms of advanced learners’ exploitation of the thematic potential of the English clause. The results reported here suggest that Dutch learners have a more highly developed grammatical competence and in some respects a more developed discourse competence than their Spanish counterparts. The picture emerging can be expanded by studying how phrases are used at levels 2 and 3 and by looking at how learners use structures like relative clauses and appositions to provide grounding for thematic elements at all levels, including level 1, which was left out of consideration here. More qualitative analysis of learner text may also be fruitful with regard to understanding how learners use the theme zone. For instance, one might focus on such matters as the contextual fit of the various level 3 structures chosen, as well as the lack of backgrounding devices in the theme zone where such devices might be warranted (cf. Hannay

250

Mike Hannay and Elena Martínez Caro

forthcoming). By taking such additional features into account it should be possible to provide a detailed description of one important component of discourse competence. Notes 1

We would like to thank Chris Butler and the editors of the present volume for useful comments on an earlier version of this paper.

2

The great majority of the examples in Spanish have been extracted from the book Hablemos de la vida (Let’s talk about life, marked as HV), based on a series of formal conversations between a Spanish journalist (Nativel Preciado) and a philosopher (José Antonio Marina). The coding of the examples also gives the page number.

3

After conducting the analysis, our attention was drawn to a note by Kaszubski (2001: 317n) to the effect that the Spanish subcorpus in ICLE belongs to the upper intermediate level rather than the advanced level, due to the relatively great number of grammatical mistakes and the poor level of the vocabulary. We stress that while grammaticality is also an issue in the present study, the core of the analysis is geared towards the semantic and rhetorical exploitation of the structures available. Correctness of expression is not relevant for this part of the analysis.

4

The codes used here are those given in the ICLE corpus. SP indicates that the sentence comes from the Spanish subcorpus, while DB/DN refers to the Dutch subcorpus.

5

Standardised residuals and lambda value are relevant for detecting error in the predictive model.

6

Some of these structures containing independent noun phrases in Spanish, and especially those with a thematic object as in (49), seem to be related to the common use of the left dislocation construction, highly grammaticalised in this language. However, left dislocation structures in Spanish are mostly associated with the informal spoken language (cf. Downing 1997).

7

Conditional-concessive clauses, a mix of conditional and concessive, are typically whether clauses as in Sarah is always intense, whether working or playing (cf. Quirk et al. 1985: 1100).

8

The English native speaker corpus consisted of approximately 500,000 words and comprised academic texts, literary fiction, and newspaper and magazine texts.

9

The experiments involved writing tasks, unscrambling, and judgments tasks concerning textual fit. The subjects were (a) English native speakers

Thematic choice in written English

251

who were experienced writers of English, (b) English native speaker students regarded as novice writers, (c) Dutch native speakers who were experienced writers of Dutch, and (d) first and second year Dutch students of English, regarded as novice writers. For further details of the experimental design, see Smits (2002: Ch. 8). References Altenberg, B. and M. Tapper (1998), ‘The use of adverbial connectors in advanced Swedish learners’ written English’, in: S. Granger (ed.) Learner English on Computer. London: Longman. 80-93. Berry, M. (1996), ‘What is Theme? A(nother) personal view’, in: M. Berry, C.S. Butler, R. Fawcett and G. Huang (eds) Meaning and Form: Systemic Functional Interpretations. Meaning and Choice in Language: Studies for Michael Halliday. Norwood, NJ: Ablex. 1-64. Biber, D, S. Johansson, G. Leech, S. Conrad and E. Finnegan (1999), Longman Grammar of Spoken and Written English. Harlow: Pearson. Butler, C.S. (2003), Structure and Function. A Guide to Three Major Structural Functional Theories. Part 2: From Clause to Discourse and Beyond. Amsterdam & Philadelphia: John Benjamins Publishing Company. Connolly, J.H. (1991), Constituent Order in Functional Grammar: Synchronic and Diachronic Perspectives. Berlin & New York: Foris. Council of Europe (2001), Common European Framework of Reference for Languages. Learning, Teaching, Assessment. Cambridge: Cambridge University Press. Dik, S.C. (1997), The Theory of Functional Grammar. Vol. 1: The Simple Clause. Berlin & New York: Mouton de Gruyter. Downing, A. (1991), ‘An alternative approach to theme: a systemic-functional perspective’, Word, 42(2): 119-143. Downing, A. (1997), ‘Discourse pragmatic functions of the Theme constituent in spoken European Spanish’, in: J.H. Connolly, R.M. Vismans, C.S. Butler and R.A. Gatward (eds) Discourse and Pragmatics in Functional Grammar. Berlin & New York: Mouton de Gruyter. 137-161. Downing, A. and P. Locke (2006), English Grammar: A University Course. Second edition. London & New York: Routledge. Gómez-González, M.L.A. (1998), ‘A corpus-based analysis of extended multiple themes in PresE’, International Journal of Corpus Linguistics, 3(1): 81-113. Gómez-González, M.L.A. (2004), ‘Functional Grammar and the dynamics of discourse’, in: J.L. Mackenzie and M.L.A. Gómez-González (eds) A New Architecture for Functional Grammar. Berlin & New York: Mouton de Gruyter. 211-242. Granger, S. (1996), ‘From CA to CIA and back: an integrated approach to computerized bilingual and learner corpora’, in: K. Aijmer, B. Altenberg and M. Johansson (eds) Languages in Contrast. Papers from a Symposium

252

Mike Hannay and Elena Martínez Caro

on Text-based Cross-linguistic Studies, Lund, 4-5 March 1994. Lund: Lund University Press. 37-51. Granger, S., E. Dagneaux and F. Meunier (eds) (2002), The International Corpus of Learner English. Handbook and CD-ROM. Louvain-la-Neuve: Presses Universitaires de Louvain. Granger S. and S. Tyson (1996), ‘Connector usage in the English essay writing of native and non-native EFL speakers of English’, World Englishes, 15: 1929. Halliday, M.A.K. (2004), An Introduction to Functional Grammar. Third edition revised by C.M.I.M. Matthiessen. London: Arnold. Hannay, M. (1994), ‘The theme zone’, in: R. Boogaart and J. Noordegraaf (eds) Nauwe betrekkingen. Voor Theo Janssen bij zijn vijftigste verjaardag. Amsterdam: Stichting Neerlandistiek Vrije Universiteit and Munster: Nodus Publikationen. 107-117. Hannay, M. (forthcoming), ‘Patterns of multiple theme and their role in developing English writing skills’, in: C.S. Butler, R. Hidalgo and J. Lavid (eds) Functional Perspectives on Grammar and Discourse. Amsterdam & Philadelphia: John Benjamins Publishing Company. 257-278. Hannay, M. and E. Martínez Caro (forthcoming), ‘Last things first: a FDG approach to clause-final focus in Spanish and English’, in: M.L.A. Gómez González, J.L. Mackenzie and E. González Álvarez (eds) Languages and Cultures in Contrast: New Directions in Contrastive Linguistics. Amsterdam & Philadelphia: John Benjamins Publishing Company. Hasereyn, W. K. Romijn, G. Geerts, J. de Rooij and M.C. van den Toorn (1997), Algemene Nederlandse Spraakkunst. Second, revised edition. Groningen: Martinus Nijhoff. Hinkel, E. (2002), Second Language Writers’ Text. Mahwah, NJ: Lawrence Erlbaum. Hinkel, E. (2004), Teaching Academic ESL Writing: Practical Techniques in Vocabulary and Grammar. Mahwah, NJ: Lawrence Erlbaum. Kaszubski, P. (2001), ‘Tracing idiomaticity in learner language: the case of BE’, in: P. Rayson, A. Wilson, T. McEnery, A. Hardie and S. Khoja (eds) Proceedings of the Corpus Linguistics 2001 Conference. Lancaster: University Centre for Computer Corpus Research on Language. 312-322. Martínez Caro, E. (2006), ‘Constituent order in Spanish: a Functional Grammar perspective’, in: M. Carretero, L. Hidalgo Downing, J. Lavid, E. Martínez Caro, J. Neff, S. Pérez de Ayala and E. Sánchez-Pardo (eds) A Pleasure of Life in Words. A Festschrift for Angela Downing. Madrid: Universidad Complutense. 187-213. Preciado, N. and J.A. Marina (2002), Hablemos de la vida (‘Let’s talk about life’). Madrid: Temas de Hoy. Quirk, R., S. Greenbaum, G.N. Leech and J. Svartvik (1985), A Comprehensive Grammar of the English Language. London: Longman.

Thematic choice in written English

253

Smits, A.M. (2002), How Writers Begin their Sentences: The Discourse Functions of Complex Beginnings in Written English. LOT Dissertation Series 69. Utrecht: LOT.