John Benjamins Publishing Company This is a contribution from Diachronica 29:3 © 2012. John Benjamins Publishing Company This electronic file may not be altered in any way. The author(s) of this article is/are permitted to use this PDF file to generate printed copies to be used by way of offprints, for their personal use only. Permission is granted by the publishers to post this file on a closed server which is accessible to members (students and staff) only of the author’s/s’ institute, it is not permitted to post this PDF on the open internet. For any other use of this material prior written permission should be obtained from the publishers or through the Copyright Clearance Center (for USA: www.copyright.com). Please contact rights@benjamins.nl or consult our website: www.benjamins.com Tables of Contents, abstracts and guidelines are available at www.benjamins.com Contact-induced grammatical change Towards an explicit account Christopher Lucas SOAS, University of London Language contact plays a key part among the factors leading to change in grammars, and yet the study of syntactic change, especially in the generative or innatist tradition, has tended to neglect the role of contact in this process. At the same time, much work on contact-induced change remains descriptive, with theoretical discussion often restricted to the putative limits on borrowing. This article aims at moving beyond these restrictions by outlining a psycholinguistically-based account of some of the ways in which contact leads to change. This account takes Van Coetsem’s (1988, 2000) distinction between recipientlanguage and source-language agentivity as its starting point, building on this insight in the light of work on second language acquisition and first language attrition, and showing how these principles can be integrated into a unified acquisitionist model of syntactic change in general. The model is then applied to case studies of contact-induced syntactic change in Yiddish and Berber. Keywords: language contact, syntactic change, borrowing, imposition, second language acquisition, first language attrition 1. Introduction Arguably the key concern of modern historical linguistics is to understand the mechanisms underlying language change; and at least since Schuchardt (1884) it has been clear that change — typically convergence — is especially likely when languages come into contact. Despite this, theoretical work on the mechanisms which cause contact-induced change, as opposed to the sociolinguistic and other factors which constrain it, remains underdeveloped. This article contributes to this work by focussing on some of the mechanisms underlying contact-induced change in syntax. Opinions differ not only on how best to model syntactic change, but also on its definition. Following Andersen (1973), Hale (1998) and others, I view changes Diachronica 29:3 (2012), 275–300. doi 10.1075/dia.29.3.01luc issn 017–225 / e-issn 159–971 © John Benjamins Publishing Company 276 Christopher Lucas in the grammars acquired by individual speakers, rather than their performance or usage, as the ultimate object of investigation in diachronic syntax. However, since usage is the input to acquisition, any theory of contact-induced change must be based on the ways that contact can affect usage. As such, the arguments and evidence presented here should be relevant to historical linguists of various theoretical persuasions. Traditionally, research from an acquisitionist perspective focuses more on language-internal triggers of inter-generational changes in grammars than on triggers involving language contact. But in view of recent suggestions concerning the necessity for syntax-external causation in bringing about syntactic change (Longobardi 2001, Keenan 2002), this imbalance is unsatisfactory. This article moves in the direction of an acquisitionist model of contact-induced change by building on insights from outside the generative tradition, primarily in the work of Van Coetsem (1988, 2000), and adapting these insights in the light of work on second language (L2) acquisition and first language (L1) attrition. The article is structured as follows. §2 begins presents Van Coetsem’s (1988, 2000) model, and discusses how certain problematic elements can be recast and made more explicit by incorporating empirical findings on the nature of bilingualism. §3 explores L1 attrition as an input to syntactic change, exemplified in two case studies from Yiddish and Berber in §4. §5 discusses some of the psycholinguistic mechanisms underlying these kinds of changes, and §6 concludes. 2. Van Coetsem’s model As Winford (2005) points out, the key advantage of Van Coetsem’s model over competing approaches to contact-induced change is that it offers a principled means of distinguishing between the traditional notions of ‘borrowing’ and ‘interference’ (or ‘substratum influence’ or ‘transfer’). It does this by focussing on the bilingual speakers who are the agents of contact-induced change, and on the different types of agentivity they employ, depending on whether they are dominant (cognitively rather than socially speaking) in the source language (SL) or the recipient language (RL). In other well-known models of contact-induced change, (e.g. Thomason & Kaufman 1988, Johanson 2002), classification of contact processes is explicitly based on sociolinguistic factors only. While sociolinguistic approaches to language contact clearly have their advantages, they do not offer a suitable basis for a model of how contact initiates change in the grammars of individual speakers. In Van Coetsem’s framework ‘transfer’ is the general term for the introduction of patterns or material from one language into another. Transfer is, by definition, © 2012. John Benjamins Publishing Company All rights reserved Contact-induced grammatical change 277 always from the SL to the RL. The bilingual speakers of the RL and the SL are the ‘agents’ of contact-induced change. Ordinarily the agents of change will be ‘dominant’ in one of the languages involved in a given change. Typically a speaker’s dominant language is her1 L1, or ‘native’ language: the language acquired from birth (leaving aside for the moment the question of ‘balanced’ bilinguals). When an agent is dominant in the RL, a change she makes to the RL occurs under ‘RL agentivity’; that is, it is an instance of ‘borrowing’. When an agent of change is dominant in the SL, a change she makes to the RL occurs under ‘SL agentivity’, or ‘imposition’. On this definition of agentivity bilingual speakers can potentially employ both types of agentivity in a single contact situation: RL-agentivity on their dominant language, and SL-agentivity on their non-dominant language. We are now in a position to consider the cognitive mechanisms underlying these two transfer types, and the extent to which certain linguistic outcomes are more associated with one type than with the other. But first, certain more problematic aspects of Van Coetsem’s approach need to be addressed. The first issue concerns the precise nature of dominance. Van Coetsem (2000: 52) suggests that, while a bilingual speaker’s L1 and her dominant language typically align, “a speaker may become more proficient in a subsequently acquired language than in his native language (which comes first in time) … . Nativeness can thus be overruled by linguistic dominance”. Van Coetsem nowhere explains what he means by ‘proficiency’ in this context. If proficiency is calculated by means of fluency tests, vocabulary tests, etc., then clearly many people do become more proficient in an L2 than in their L1. Whether this has meaningful implications for the role of bilingualism in contact-induced change is doubtful, however. Instead, the question of whether an individual can become dominant in an L2 should be addressed by considering the nature of knowledge of an L2 versus knowledge of an L1. As is well known, L2 acquisition, particularly in adulthood, tends to be a much lengthier, and ultimately less successful, process than L1 acquisition. Although there is a huge literature modelling L2 performance (henceforth interlanguage) using the tools of generative grammar developed to describe L1 competence (cf. Schwartz 1998), it seems doubtful that either the process of acquiring an L2, or the character of the knowledge ultimately attained, are the same as in L1 acquisition. Instead, the observed discrepancies between L1 and L2 acquisition receive a principled explanation if we assume something like Bley-Vroman’s (1989) Fundamental Difference Hypothesis. This idea (also Bley-Vroman 2009, Meisel 2011) states that L1s are acquired on the basis of an innate, domain-specific Universal Grammar (UG), whereas L2s are acquired by means of more general cognitive learning strategies. That is, ‘poverty of the stimulus’ arguments imply 1. Here and throughout I adopt the arbitrary convention of making speakers female. © 2012. John Benjamins Publishing Company All rights reserved 278 Christopher Lucas the necessity of UG (whatever its precise nature might be) in order to constrain the set of hypotheses children must make concerning the grammar of their L1s; whereas the so-called ‘logical problem of L2 acquisition’ (how interlanguage has characteristics which transcend the input) requires only the availability of adults’ general problem-solving and analytical abilities along with knowledge of their L1s. Support for this idea comes from numerous empirical studies. For example, in a study of 57 L1 Hungarian immigrants to the US, DeKeyser (2000) found that a native-like score on an English syntactic acceptability judgment task was a statistically significant predictor of high scores on a general verbal analytical ability task (in Hungarian). This is in contrast to (non-pathological) L1 speakers, for whom knowledge of the grammar of the L1 appears to be independent of other cognitive abilities. Verbal analytical abilities are thus crucial to the eventual target-like acquisition of an L2, but largely irrelevant to the acquisition of an L1 (see also Johnson et al. 1996 on the inherent indeterminacy of L2 knowledge). If we accept that, in the majority of cases, there is a substantive difference between the nature of an L1 and an L2, where does this leave the notion of dominance and the idea that a speaker’s L2 can become dominant and her L1 nondominant? Two conflicting ways that the notion of dominance can be defined such that it is psycholinguistically more precise suggest themselves. Definition 1: A speaker’s dominant language is whichever of her L1, L2, L3, etc., is most accessible at any given time. Definition 2: A speaker’s dominant language is her L1. Any other language subsequently acquired is necessarily non-dominant. It is clear that for many bilinguals the accessibility or ‘availability’ of their L1s and L2s is shifting constantly. By Definition 1, then, an L2 can certainly become dominant with respect to an L1. However, this definition has the unwelcome consequence that transfer under RL agentivity into one’s L1 — i.e. the most intuitively obvious type of borrowing — becomes logically impossible. If, at a given time, a speaker transfers some L2 feature into her L1 performance, this is presumably at least partly because that feature is more accessible at that time than its L1 counterpart. But under Definition 1, this would be a case of imposition under SL agentivity — transfer from the dominant to the non-dominant language. For this reason, and given apparently reasonable assumptions about the respective natures of L1s and L2s, it is preferable to adopt Definition 2. Under this definition an L1 is necessarily always dominant with respect to an L2, regardless of its accessibility at a given time. This gives us a principled basis for explaining why borrowing and imposition each seem to have characteristically different consequences for an RL: to the ex- © 2012. John Benjamins Publishing Company All rights reserved Contact-induced grammatical change 279 tent that this is the case, it can be attributed, at least in part, to the very different processes of L2 acquisition and L1 attrition. Van Coetsem (1988, 2000) and Winford (2005), however, attribute the differing outcomes of borrowing and imposition to what Van Coetsem calls the ‘stability gradient’: “language does not offer the same degree of stability in all its parts, in particular … there are differences in stability among language domains, namely among vocabulary, phonology and grammar (morphology and syntax)” (Van Coetsem 1988: 25). It is claimed in particular that “phonology and grammar in general show greater stability than vocabulary” (1988: 26), and that in both borrowing and imposition the more stable domains are preserved. Winford (2005: 377) states that “this explains why borrowing tends to be mostly lexical, and does not usually affect the RL grammar, while imposition, on the other hand, tends to do so”. The claim associated with the stability gradient that is most relevant here is that, generally speaking, we should assume that contact-induced changes to the syntax of a given language are due to imposition and not borrowing (cf. the repeated use of this assumption as a diagnostic tool by Winford 2005: 390, 393, 407). Suffice to say, a number of studies of L2 acquisition (e.g. Stauble 1984, Clahsen & Muysken 1986, and many since) have found, contrary to the claims of the stability gradient, that basic elements of word order, for example, are not widely imposed, and where they are, this is typically only for a brief period in the earliest stages of L2 acquisition. Certainly, second language learners often produce target-deviant L2 word orders, but it seems only to be with more peripheral aspects of word order — such as the order of adverb and verb or the relative order of adverbs — that clear cases of imposition of L1 structures are found (e.g. White 1991). Where L2 learners characteristically make errors in the basic word order of the RL (e.g. in the order of subject, object and verb in V2 languages), this has been shown to result not from transfer of L1 structures but rather reanalysis of RL data (e.g. Möhring & Meisel 2003, Håkansson et al. 2002).2 We can make sense of these findings in terms of the approach to L2 acquisition advocated above. If L2 acquisition, unlike L1 acquisition, depends crucially 2. Another issue connected with dominance and the stability gradient concerns individuals who have simultaneously acquired two first languages. Clearly in the case of such individuals neither of their L1s is dominant and so the distinction between borrowing and imposition breaks down. Van Coetsem (2000) acknowledges this and talks of ‘neutralization’. He suggests (2000: 86) that neutralization permits the “free transfer” between the two L1s of elements from any linguistic domain. As Louden (1992, 1997) and others have shown, however, this appears to be incorrect. In ‘stable bilingualism’, where whole communities have two (or more) L1s, we observe widespread syntactic transfer (often called ‘convergence’) but typically little in the way of phonological transfer. Syntactic convergence in cases of stable bilingualism is an important sub-type of contact-induced change, but there is insufficient space to treat it in detail here. © 2012. John Benjamins Publishing Company All rights reserved 280 Christopher Lucas on general cognitive learning and reasoning abilities, then we expect it to involve reanalysis and simplification of the subtler aspects of RL syntax to a greater extent than is observed in L1 acquisition. However, there are no grounds to expect that L2 acquisition will necessarily involve wholesale imposition of all aspects of L1 syntax. Instead, the key factor determining the type of knowledge of the L2 that the learner acquires must be the nature and quantity of data that she is exposed to. Since virtually every utterance of the L2 that the acquirer hears will contain evidence of basic word order, this is likely to be more salient than other aspects of syntax, for which data are scarcer (Siegel 2003). As such we should only expect basic word order to be imposed in the earliest stages of acquisition, if at all. By contrast, for less salient aspects of L2 syntax it seems natural that acquirers should draw on their only other source of knowledge of language and hypothesize that the feature in question is identical to its L1 counterpart in the absence of sufficient evidence to the contrary. As noted in §1, syntactic change from the acquisitionist perspective is change in grammars, not merely variation in performance; and on this view interlanguage which is target-deviant is therefore a potential input to change, rather than change itself. From this perspective, change due to imposition (or other processes of L2 acquisition) occurs when a sufficiently high proportion of a child’s primary linguistic data (PLD) is made up of target-deviant interlanguage, such that that child acquires a grammar on the basis of those data which differs from that of previous acquirers. Given the conclusion that syntactic imposition, viewed in this way, is rather less prevalent than suggested by Van Coetsem and Winford (and straightforwardly explained in terms of the domain-general learning strategies involved in L2 acquisition), the remainder of this article focuses on syntactic borrowing. 3. Syntactic borrowing As we have just seen, even on a definition of dominance whereby an L2 is necessarily non-dominant, there is no doubt that an L2 can become the more accessible of a bilingual’s languages. Moreover, where there is prolonged exposure to and use of an L2 (and concomitant lack of exposure to and disuse of the L1) it is uncontroversial that linguistic performance in the L1 can be noticeably affected. This process is commonly referred to as ‘attrition’. From the acquisitionist perspective of change as change in grammars, however, it is important to know whether attrition is limited to performance, or whether it can also lead directly to changes in adult L1 competence. In what follows, I use the term ‘attrition’ as shorthand for ‘some bilingualisminduced alteration in a speaker’s L1 competence and/or performance’. Importantly, © 2012. John Benjamins Publishing Company All rights reserved Contact-induced grammatical change 281 ‘attrition’ here refers not only to deterioration, but to any alterations at all, including those best thought of as the addition of new options or rules.3 Unfortunately there has been little empirical investigation of whether competence as well as performance can be attrited in this way, not least, perhaps, because separating competence from performance in empirical data can be highly problematic (cf. Altenberg & Vago 2004, Schütze 1996). The usual a priori assumption, following Chomsky (1980), is that, upon maturation, the Language Acquisition Device reaches a ‘steady state’, and the grammar that instantiates this steady state does not undergo any significant alteration in non-pathological individuals during adulthood, whether or not an L2 is subsequently acquired (Köpke 2004, Tsimpli et al. 2004, Flores 2007, but see Sharwood Smith & Van Buren 1991 for a different view). In her study of L1 attrition among German Jews in Anglophone countries, for example, Schmid (2002: 18) finds concerning both morphology and syntax that “the question of whether … the actual knowledge of the L1 can become deteriorated … has not been conclusively resolved, but evidence overwhelmingly points towards what difficulties there are being only temporary”. This position is strengthened by studies (Ås 1962, Fromm 1970, Footnick 2007) where subjects who professed to having completely lost the language of their parents displayed native-like ability in both comprehension and production of that language while under hypnosis. It is natural to suppose that what holds for these extreme cases also holds for less extreme cases of L1 attrition: the appearance of fully-fledged loss of linguistic knowledge, considered as such by the subjects concerned, may in fact be merely the result of temporary difficulties with access and retrieval (see Altenberg 1991, Stolberg & Münch 2010 for experimental support). Let us proceed, therefore, on the assumption that in non-pathological individuals L1 attrition is restricted to performance and does not alter competence. What this means, again, is that from an acquisitionist perspective, L1 attrition is not change, but a potential source of perturbation to the PLD on the basis of which children acquire their own L1s. As mentioned in §1, however, it is the mechanisms underlying this kind of alteration in performance that a theory of contact-induced change must explain. Precisely how a child acquires a grammar on the basis of PLD is a matter for the theory of L1 acquisition in general: the processes involved will be the same whether or not the PLD has been affected by language contact. The main concern of what follows is therefore to elucidate the mechanisms leading to attrition in L1 performance — a topic of relevance to anyone interested in syntactic change, whether from an acquisitionist or a usage-based perspective. 3. Compare Seliger (1996: 614 fn. 6): “Language attrition is, in fact, a creative and often additive process”. Friedman & Joseph (forthcoming) adopt the term ‘reverse interference’ for this same concept. © 2012. John Benjamins Publishing Company All rights reserved 282 Christopher Lucas As an example of how L1 attrition, though restricted to performance, can nevertheless be systematic across a number of RL-dominant bilinguals, consider Gürel’s (2004) study of attrition in Turkish binding. Gürel tested 24 L1 Turkish speakers who were longterm residents in North America, having arrived as adults. The purpose was to test whether subjects’ interpretation of the binding properties of Turkish pronominal expressions had undergone influence from their English counterparts. Turkish has a third person singular pronoun o (not marked for gender) which, like English him/her, cannot be bound (i.e. must not be co-indexed with another referring expression) within its governing category: (1) Buraki o-nu*i/j beğen-iyor. Burak s/he-acc like-prog “Buraki likes him*i/j.” (Gürel 2004: 58) However, unlike in English, only whole matrix clauses — not embedded clauses or possessive noun phrases — are governing categories in Turkish. Hence in (2)–(3) John and o cannot co-refer as John and he/his can in the translations. (2) Johni o-nun*i/j gel-eceğ-i-ni söyle-di. John s/he-gen come-nmlz-3sg.poss-acc tell-pst “John said he would come.” (3) Johni o-nun*i/j karı-sı-nı öp-tü. John s/he-gen wife-3sg.poss-acc kiss-pst “John kissed his wife.” (Gürel 2004: 59) A co-referential reading is only possible in these contexts with an alternative pronoun kendisi, which may be bound within a matrix clause: (4) Johni kendi-si-nini/j karı-sı-nı öp-tü. John kendi-3sg-gen wife-3sg.poss-acc kiss-pst “John kissed his wife.” (Gürel 2004: 58) Three tasks in Gürel’s study were designed to test subjects’ judgments of the binding properties of these items: an untimed written interpretation task; an aural truth-value judgment task; and a picture identification/listening task. The test was also administered to a control group of 30 adult L1 Turkish speakers who had lived in Turkey since birth and had only beginner-level English. The untimed written interpretation task showed no significant difference between the tested and the control groups: the overwhelming majority of responses in both stated that the Turkish pronoun o only permitted a disjoint interpretation. However, in the other two tasks there was a significant difference between the two groups. In both tasks preference for a disjoint interpretation of o was at or near 100% in the control group, but 70–79% in the tested group. That is, in the tested © 2012. John Benjamins Publishing Company All rights reserved Contact-induced grammatical change 283 group at least 20% of responses to sentences in these two tasks stated that a bound (co-referential) interpretation of o as the subject of an embedded clause was preferred, against the judgments of unattrited L1 Turkish speakers. Thus we have clear evidence of L1 attrition among the 24 subjects in this study. The subjects appear (presumably unconsciously) to have established an equivalence between the English third person singular pronouns and o, with the result that, in processing at least, they interpret the governing category in which o must be free to be the same as for the English pronouns: not only whole matrix clauses but also embedded clauses and possessive noun phrases. In contrast, the control group, with minimal exposure to English, invariably interpret o as obligatorily free within the governing category of the whole matrix clause.4 Although Gürel did not conduct any production experiments, we may reasonably assume that these L1 attriters are at least as prone to interpreting o in this way in production as in processing. This being the case, Gürel’s study provides clear evidence that a range of bilinguals in sociolinguistically similar conditions can be the agents of contact-induced syntactic change under RL agentivity. If these speakers’ attrited Turkish performance makes up a significant proportion of the PLD of children acquiring their L1s, these children will be expected to acquire a grammar in which the governing category for Turkish o may be less than the whole matrix clause. At this point we can say from an acquisitionist perspective that a syntactic change due to borrowing has occurred. In §5 we consider what psychological mechanisms might underlie this kind of performance attrition. First, §4 presents two case studies of completed historical changes which appear to have been initiated by RL-dominant speakers in the same way. 4. Note that the discrepancy between the results in the first task, where there was no significant difference between the tested and control groups, and the other two tasks, where there was a significant difference, provides further support for the conclusion that attrition is restricted to performance and processing, and does not affect competence. The first task was an untimed written interpretation task in which subjects reflected directly on the referential properties of the Turkish pronouns. Here it is likely that there was less interference of performance factors than in the other two tasks, in which the target sentences were more contextualized and, in the case of the third task, heard only once. Gürel (2004: 75) concludes that “what is documented in this study is the alteration of the L1 grammar at some deeper syntactic competence level”, but this conclusion does not seem justified in view of the discussion in this section. © 2012. John Benjamins Publishing Company All rights reserved 284 Christopher Lucas 4. Syntactic borrowing in Yiddish and Berber 4.1 Yiddish expletive subjects The first case of syntactic borrowing to be considered comes from Prince’s (2001) study of expletive subjects in the Germanic language Yiddish. In (5) we have an example of a Yiddish structure, with the originally demonstrative element dos in initial position, that has no direct analogue in other Germanic languages. (5) Dos hot Leyb gezen Eriken. this.neut aux.3msg Leonard.nom see.ptcp Erica.acc “It’s Leonard who saw Erica.” (Yiddish, Prince 2001: 268) Yiddish also has a similar construction with the pronoun es in initial position, as in (6), which is directly inherited from Middle High German (7). The es- and dosconstructions have quite different discourse properties, however, as we will see. (6) Es hot Leyb gezen Eriken. it aux.3msg Leonard.nom see.ptcp Erica.acc “Leonard saw Erica.” (Yiddish, Prince 2001: 270) (7) Iz en-wart nie niehein keiser so here geboren. it neg-was never no emperor so noble born “There was never any emperor born so noble.” (Middle High German, Paul 1989: 368) The construction in (5) with an original demonstrative is unattested in both Middle High German (Behagel 1928: 470–471) and Old Yiddish (Prince 2001: 268), the Jewish variety of German spoken in Germany until the migration of Jews to eastern Europe (principally in the 13–15th centuries). There is little doubt, then, that this is an innovation within (Eastern) Yiddish.5 Against this background, Prince points out the direct parallel between (5) and a Slavic construction found, for example, in Russian (8) and Polish (9). (8) Èto Leonid uvidel Èriku. (Russian, Prince 2001: 268) (9) To Leonid widział Erykę. (Polish, R. Szczepaniak, p.c.) this.neut Leonard.nom see.pst.msg Erica.acc “It’s Leonard who saw Erica.” 5. ‘Yiddish’ and ‘Eastern Yiddish’ are generally used synonymously. Strictly speaking, Eastern Yiddish refers to all varieties except Western Yiddish, that is, the dialects of those European Jews who never migrated east from Germany. © 2012. John Benjamins Publishing Company All rights reserved Contact-induced grammatical change 285 Prince (2001: 268) notes the following parallels: both (5) and (8)–(9) have an expletive pronominal element in initial position; in both cases this is (homophonous with) the neuter demonstrative pronoun; neither sentence contains an embedded clause; and both sentences are most appropriately translated with an English it-cleft. This last point represents an important distinction with respect to the Yiddish es-sentence in (6). Prince shows that dos-sentences (and their Slavic counterparts) mark a proposition as hearer-old, except for the particular instantiation of one of its variables (usually the subject term), which is hearer-new and contrastively focussed. Hence in (5) and (8)–(9) the hearer knows that somebody saw Erica, but the fact that it was Leonard who saw her is treated as new information. Leonard could, however, have been referred to in the immediately preceding discourse in some other connection. By contrast, es-sentences are used only in contexts where the subject of a proposition is entirely hearer-new (in the current discourse segment). This distinction is seen from the fact that subject pronouns (which necessarily refer to an entity in the immediately preceding discourse) are common in dos-sentences but infelicitous or ungrammatical in es-sentences, in which the subject must be hearer-new. Compare (10), in which the proposition is not hearer-old (and ‘everything’ is an inappropriate target of contrastive focus), and (11), in which the proposition (except the instantiation of its subject term) is discourse-old and the pronoun subject is necessarily hearer-old (and can be the target of contrastive focus). (10) a. Es geyt alts azoy geshvint. b. # Dos geyt alts azoy geshvint it go.pres.3sg everything so fast “Everything’s going so fast.” (11) a. # Es shlogst du di puter? b. Dos shlogst du di puter? it beat.pres.2sg you the butter “Is it you that’s churning the butter?” (Yiddish, Prince 2001: 272) A construction with the form and meaning of that in (10b) and (11b) is found among the Germanic languages only in Yiddish; we know that the development of this construction postdates the migration of Jews to eastern Europe; it has a direct parallel in Polish and other Slavic languages; and we know that Yiddish was in extensive contact with Polish and other Slavic languages for hundreds of years, as evidenced by the high proportion of Slavic-origin loanwords in Yiddish. These considerations favour an explanation of the development of dos-sentences as triggered by contact with Polish and perhaps other Slavic languages with a parallel © 2012. John Benjamins Publishing Company All rights reserved 286 Christopher Lucas construction.6 This development was probably favoured by the prior existence in Yiddish of the structurally similar es-sentences. However, the fact that dossentences have the same discourse function as their Slavic counterparts — and a clearly different discourse function to es-sentences — militates against seeing dossentences as a simple Yiddish-internal extension of a pre-existing pattern. This is the position defended by Prince (2001), but she is reluctant to view this development as a case of strictly syntactic transfer. Prince (2001: 268–270) points out that the Yiddish dos-construction is possible within an embedded clause, while it appears not to be in Slavic; and the V2 rule, which operates quite generally in Yiddish, also operates in dos-sentences, while V2 is not a feature of Slavic. It is fair to say, then, that the syntax of the Slavic and Yiddish versions of this construction is not identical: basic properties of Yiddish word order are maintained in dosconstructions, not totally restructured on the Slavic model. It is not clear, however, that this should disqualify the development in question from counting as a case of syntactic transfer. The key is that we have transfer of an abstract grammatical nature, not simply of lexical material. We will see in §5 that this transfer corresponds closely to what Heine & Kuteva (2003, 2005) have called ‘contact-induced grammaticalization’. To give an explicit account of how contact led to innovation in this case (and any other), it is essential to establish as far as possible which of the two types of agentivity was involved: borrowing (RL agentivity) or imposition (SL agentivity). Unfortunately, in contrast to the attrition studies mentioned previously, well-established historical changes such as this one tend to have occurred sufficiently long ago that the sociolinguistic background of the change cannot be known for certain and must be inferred. Nevertheless, it seems useful to offer an explicit (and falsifiable) account of this and other historical changes within the present framework, even if that account is based on reasonable assumptions rather than certainties. Although we cannot identify the individual agents of change responsible for introducing dos-sentences into Yiddish, what we know about the sociolinguistics of Yiddish–Slavic bilingualism in this period suggests that contact-induced changes to Yiddish were predominantly due to borrowing rather than imposition. The basis of this assumption is that knowledge and use of Yiddish was, and remains, very rare outside of Jewish communities. Yiddish-speaking communities in eastern Europe used the languages of their non-Jewish neighbours in their interactions with them (e.g. Jacobs 2005: 267–269), while Yiddish was used only among Jews. For this reason Yiddish will not, in general, have been acquired as an L2 by adults (but see below). Rather, Yiddish–Slavic bilinguals will have been 6. Birnbaum (1979: 77–78), however, shows that Polish is the only Slavic language to have had a significant influence common to the whole of Eastern Yiddish. © 2012. John Benjamins Publishing Company All rights reserved Contact-induced grammatical change 287 predominantly native speakers of Yiddish (i.e. Yiddish-dominant or balanced bilinguals, but not Slavic-dominant). Thus, contact-induced changes in Yiddish initiated by these speakers will have been under RL, not SL, agentivity. A problem for this scenario is that there were already significant numbers of Jews, known as Knaaim, living in eastern Europe at the time of the migration of Yiddish speakers from Germany. Louden (2000: 100) points out that these Jews, most likely native speakers of local Slavic varieties, gradually merged with the more numerous Yiddish-speaking communities, “with Yiddish winning out as the dominant intra-community medium of communication”. As such, prior to their total assimilation many Knaaim would have been adult L2 acquirers of Yiddish and therefore in a position to impose on it elements of their native varieties. Importantly, however, there seem to have been very few Knaaim in Poland prior to their assimilation with Yiddish speakers — they were instead concentrated in Russia and Lithuania (Louden 2000: 101–102). Louden suggests that this explains why certain simplificatory phonological changes (putatively the result of L2 acquisition of Yiddish by L1-Slavic Knaaim) are restricted to dialects spoken east of Poland. Prince (2001: 268) cites dos-sentences as a general (Eastern) Yiddish feature, however. If the innovation of dos-sentences had been triggered by Knaaim L2 acquirers of Yiddish, we would expect this construction to have a similar dialectal restriction to the simplificatory phonological changes cited by Louden (2000). Since it does not, the most reasonable assumption is that this development is, in fact, the result of borrowing under RL agentivity by L1 Yiddish speakers.7 4.2 Berber negation A second case comes from Berber. Many Berber varieties spoken across north Africa have undergone a change in their expression of negation whereby the original negative construction with a single preverbal marker ur has been replaced by a bipartite construction, apparently on the model of Arabic (Lucas 2007, 2009). This is illustrated in (12)–(13). (12) represents the conservative Berber negative construction found in Tuareg, spoken by nomadic peoples in Mali, Niger, southern Algeria and southern Morocco. This contrasts with the (Moroccan) Central Atlas Tamazight Berber construction in (13), which features a postverbal element ša in addition to the original preverbal ur found also in Tuareg. In (14) we have 7. Similarly, the fact that dos-sentences are a general feature of Eastern Yiddish makes it unlikely that they were triggered by the L2 acquisition of Yiddish by Jews in communities shifting away from Yiddish. Under this scenario one would rather predict dos-sentences to have a patchy distribution in the Yiddish dialects, and to be absent in regions where the language was most strongly maintained. © 2012. John Benjamins Publishing Company All rights reserved 288 Christopher Lucas the north African Arabic bipartite negative construction with preverbal mā and a postverbal element -š(i). (12) ur igle neg leave.past.3msg “He didn’t leave.” (13) ur iffiɣ ša neg exit.past.3msg neg “He didn’t go out.” (Tuareg Berber, Chaker 1996: 16) (Central Atlas Tamazight, Boumalk 1996: 36) (14) ma-katəsmaʕ-š l-klām-i neg-listen.pres.2msg-neg to-speech-my “You’re not listening to what I’m saying.” (Moroccan Arabic, Adila 1996: 103) Several factors indicate that the similar structures in (13) and (14) resulted from contact-induced change rather than independent parallel development, and that this bipartite construction spread from Arabic to Berber primarily through borrowing on the part of Berber-dominant speakers. These are detailed in Lucas (2007, 2009) and can be summarized as follows. First, the geographical distribution of the bipartite construction is such that, for a Berber variety to have it, it is a necessary condition that that variety should be spoken in a region where the local Arabic variety also has the construction. The bipartite construction found in many spoken Arabic varieties is an innovation relative to Classical Arabic. The varieties with this construction are spoken across coastal north Africa, from Morocco to Palestine, and in the southern Arabian Peninsula. Berber varieties spoken outside this region, such as Tuareg (12), universally lack the bipartite construction. Second, note that the postverbal negator ša in (13) is derived from kra “(any) thing” ([š] < [k] being a common sound change in many Berber varieties, Chaker 1996: 16). This is directly parallel to the development of the Arabic postverbal marker -š(i) from a free word šayʔ “(any)thing”, in another apparent case of contact-induced grammaticalization (see §5 for further discussion). However, in different Berber varieties the postverbal element takes various forms. Many of these derive from kra (as in (13)) but others (e.g. ara in (16) below) can be shown to derive from a separate word wara also meaning “(any)thing” (Chaker 1996: 16). This suggests that the development was not a single early Berber-internal development, but rather that on two or more occasions contact with Arabic resulted in a Berber variety grammaticalizing its word for “(any)thing” as a postverbal negator on the model of Arabic. Third, while this development of a postverbal negator from a word formerly meaning “(any)thing” is not uncommon (particularly in western Europe), © 2012. John Benjamins Publishing Company All rights reserved Contact-induced grammatical change 289 idiosyncratic features of the syntax of the construction in Berber and Arabic suggest a single innovation that spread through contact rather than independent parallel developments. For example, in all Berber and Arabic varieties that typically feature the bipartite construction in unmarked contexts, the newer postverbal negator is odd or ungrammatical in contexts where negation is emphasized by oaths invoking God: (15) wəḷḷāh ma-ngūl-ha-lu(*?-š) by.God neg-say.pres.1sg-it-to.him “By God, I won’t tell him it.” (Moroccan Arabic, Caubet 1996: 86) (16) wəḷḷəh ur t-swiɣ (*?ara) by.God neg it-drink.past.1sg “By God, I didn’t drink it.” (Kabyle Berber, Mettouchi 1996: 192) There is no general crosslinguistic trend for this kind of negative sentence to be specially marked; and those Arabic and Berber varieties which lack the bipartite negative construction altogether mark negative oaths no differently to other negative sentences. Hence this commonality between the relevant Arabic and Berber varieties would constitute a bizarre coincidence if it had been innovated separately in each case. Finally, what we can infer about the sociolinguistic situation in north Africa from the time of the Arab conquest until the present day points to this change being primarily the result of borrowing, rather than imposition, though the lack of detailed information on this score makes this conclusion somewhat tentative. Historical evidence (e.g. Abun-Nasr 1971, Julien 1961) shows that ethnic Arabs have always been a small military–political elite in north Africa, particularly west of Egypt. Numbers grew in the eleventh century when conglomerations of Arab tribes such as the Banu Hilal migrated into the region, but these remained a minority. This historical picture is confirmed by recent population-genetic studies, which show little or no significant genetic difference between those in northwest Africa today who speak Berber and those who do not, or between those who identify themselves as culturally Berber or culturally Arab. At the same time, all of these groups differ significantly and in the same ways from today’s inhabitants of the Middle East and Europe, including southern Spain (Bosch et al. 2000, Fadlaoui-Zid et al. 2004, Manni et al. 2002). Thus the importance in this region of the Arabic language, Islam and other aspects of Arab culture is first and foremost the result of a cultural shift by autochthonous Berbers, rather than large-scale demographic replacement by, and intermarriage with, individuals of Arab genetic inheritance. As such, the numbers of genetically Arab L1 Arabic speakers will always have been small, and, given their status as an elite in a society whose cultural, © 2012. John Benjamins Publishing Company All rights reserved 290 Christopher Lucas religious and political institutions were being steadily arabized, the putative acquisition of Berber as an L2 by a subset of these is unlikely to have been the source of significant structural effects on the L1 Berber varieties of the masses. However, ethnic origin is clearly not a sure predictor of linguistic dominance. It is likely that, in the period when many Berbers started to shift to Arabic, at least some individuals were exposed to insufficient Berber during childhood to acquire it as an L1, but subsequently went on to learn it as an L2. Such individuals could then impose elements of their L1 Arabic onto their L2 Berber. However, to the extent that such individuals existed, they are likely to have been few in number and relatively uninfluential for the following reasons. Although many peripheral varieties of Berber are in serious decline, there remain over ten million native speakers of varieties with the bipartite negative construction in Morocco and Algeria.8 A case could perhaps be made concerning some peripheral varieties, for example in Tunisia and Libya, that numbers of L1 speakers have been sufficiently small for sufficiently long that those who failed to acquire Berber as an L1 but did then learn it as an L2 could have represented a significant proportion of speakers of that variety and thus have made a nontrivial contribution to the PLD of children acquiring that variety as an L1. But this scenario is implausible for the more widely-spoken varieties. Any such imperfect learners will always have been on the peripheries of these speech communities (usually having migrated from Berber-speaking rural areas to Arabic-speaking urban centres, cf. Crawford 2002, Hoffman 2008) and until very recently the Berber language has been actively repressed rather than encouraged in the political, commercial and educational spheres (Hoffman 2006). It is thus unlikely that historically there have been large numbers of L1 Arabic speakers learning Berber as an L2 and making extensive use of their L2 Berber, whereas it is certain that the L2 acquisition of Arabic by L1 speakers of Berber has been an increasingly widespread phenomenon in northwest Africa for almost a millennium. Hence we may reasonably conclude that the development of bipartite negation in Berber was the result of borrowing from L2 Arabic on the part of L1 Berber speakers. Together with the discussion in §§2–3, these case studies show, contrary to the stability gradient, that a default assumption that contact-induced syntactic change is due to imposition and not borrowing is unjustified. Instead it is preferable to make the available sociolinguistic information about the contact situation criterial in establishing whether a change was due to borrowing, imposition, or perhaps both. 8. See http://www.ethnologue.com/show_family.asp?subid=57-16 (accessed on 18/06/2011). © 2012. John Benjamins Publishing Company All rights reserved Contact-induced grammatical change 291 5. Mechanisms of borrowing We now turn to the cognitive mechanisms underlying the Berber and Yiddish developments described in §4, and how these mechanisms can be integrated into the acquisitionist model of change. This models states that children acquiring their L1 are the locus of change in grammars. Change occurs when a child’s PLD (or ‘trigger experience’) differs significantly from the PLD of older speakers who acquired their grammars earlier. These differences cause the child to abduce a different grammar to that of older groups. This alteration in PLD could have various syntax-external causes, such as phonological change, change in frequency of use of various structures, and, most importantly for our purposes, language contact. The focus of what follows is on how L1 attrition can bring about changes in PLD, leading to change in grammars through L1 acquisition. 5.1 Why attrition? We saw in §3 that, despite considerable evidence of L1 performance attrition in bilingual speakers, L1 competence is typically not attrited in non-pathological individuals. If syntactic change is change in mental grammars, then this means that attrition is an input to change, in the form of perturbation to PLD, rather than the change itself. But what causes attrition in performance? Plausibly, L1 performance attrition is the manifestation of an inherent cognitive tendency to minimize the processing effort associated with the use of two (or more) languages (Weinreich 1953: 8, Nadkarni 1975: 381). This idea might be implemented as follows. Even if one accepts the ‘massively modular’ view of cognition (Carruthers 2005), there is still clearly an important interaction between language and ‘habit’, or procedural memory, potentially expressible in terms of strength of neural connections. We might hypothesize a neural network associated with, say, a morpheme or syntactic structure in a speaker’s L2. If the strength of the connections between the units in this network is high due to repeated exposure to, and use of, the relevant morpheme/structure, one would expect there to be decreased processing costs associated with involving this same, already strongly activated network in L1 use too. Put simply, it pays to carry over habits formed in an L2 into one’s L1. Anyone who has spent time immersed in an L2 environment will have anecdotal examples of the resulting urge (sometimes irresistible) to indulge L2 habits during L1 speech. Similarly, where a speaker makes little use of, and has little exposure to her L1 over an extended period, this will result in decreased activation of the neural networks associated with communication in the L1, manifesting itself to the consciousness of the speaker as a decrease in the L1’s accessibility (Köpke 2007). At the same time, the neural networks associated © 2012. John Benjamins Publishing Company All rights reserved 292 Christopher Lucas with communication in an L2 that is used frequently will be strongly activated and therefore that L2 will be readily accessible. However, if the speaker in question wishes to express herself in the deactivated L1, this must be reactivated, and the strongly activated L2 must be inhibited. This simultaneous reactivation and inhibition will put pressure on the cognitive mechanisms controlling speech production, such as working memory. Where this pressure is moderate, normal L1 production will result. Where it is excessive, however, one would expect the full reactivation of the L1 and inhibition of the L2 to be unsuccessful and for L2-influenced performance attrition of the L1 to result. Bearing these ideas in mind, we return to the developments described in the previous section. 5.2 Performance attrition in Berber and Yiddish As noted in §4, both the Berber and Yiddish developments appear to instantiate Heine & Kuteva’s (2003, 2005) notion of contact-induced grammaticalization (CIGr), to which we now turn. Heine & Kuteva (HK) present CIGr as an extension of grammaticalization theory from cases of internal to cases of external grammatical change (2005: 1). HK (2003: 533, 539) distinguish two varieties of CIGr, which they label ‘ordinary’ CIGr and ‘replica grammaticalization’. We focus on replica grammaticalization: (17) Replica grammaticalization9 a. Speakers notice that in language M there is a grammatical category Mx. b. They create an equivalent category Rx in language R, using material available in R. c. To this end, they replicate a grammaticalization process they assume to have taken place in language M, using an analogical formula of the kind [My > Mx]: [Ry > Rx]. d. They grammaticalize Ry to Rx. The process outlined in (17) captures the case of Berber negation fairly well. As we saw in §4, north African varieties of Arabic have, in addition to their preverbal negative marker ma, a postverbal negator -š(i), derived (grammaticalized) from an independent word šayʔ “thing”. Apparently as a result of contact with these varieties of Arabic, many varieties of Berber spoken in the same region have developed a postverbal negator in addition to the preverbal one. In the case of Central Atlas Tamazight, for example, this is ša, derived (grammaticalized) from an independent word kra, also meaning “thing”. Thus Arabic is the SL and Berber the RL, and Berber 9. HK use the terms M(odel) language and R(eplica) language equivalently to our SL and RL, respectively. © 2012. John Benjamins Publishing Company All rights reserved Contact-induced grammatical change 293 speakers could be argued to have replicated a grammaticalization process they assume to have taken place in Arabic in order to create a new negative construction. There are significant difficulties with HK’s formulation of CIGr, however. In particular, it is hard to see precisely what is meant in (17) by ‘category’ and ‘notice’. If ‘category’ is being used in something like its standard sense (equivalent to ‘part of speech’), then (17) appears to exclude any case, such as Berber negation, where no new category is created, only a new construction to express an existing category. Presumably HK do not intend such a restriction. Regarding ‘notice’, if we take this literally it suggests that the processes described in (17) are necessarily conscious and intentional, with speakers engaged in these processes having a great deal of metalinguistic knowledge at their disposal (cf. Karatsareas 2007, Gast & van der Auwera 2009). This hardly seems plausible. Finally, it is not clear, given these reservations, that the formulation of CIGr in (17) has more explanatory power than traditional notions of calquing and loan translation. Rather, its chief value (as is arguably also the case with grammaticalization more generally, e.g. Campbell 2001) is more as a means of classification than of explanation (see HK 2005 for examples from a wide range of languages). As such, I propose this reformulation, summarizing the ideas developed here: (18) Contact-induced grammaticalization (revised) a. The general cognitive goal of minimizing processing costs associated with bilingualism results in the recreation of an SL construction Sx in the RL performance of RL-dominant speakers. b. This new RL construction Rx does not reflect the underlying competence of the RL-dominant speakers, but does form part of the PLD for children acquiring the RL, who thus acquire Rx as part of their underlying L1 competence. In the case of Berber negation the RL is Berber and the SL Arabic. Construction Sx is the Arabic bipartite negative construction in (14). Construction Rx is the Berber bipartite negative construction shown in (13). In this example the first stage of (18) presumably occurred roughly as follows. Berber-dominant speakers make extensive use of Arabic generally and thus also of the Arabic bipartite negative construction. This construction maps neatly onto the Berber construction, but with the addition of a postverbal negator which is (virtually) homophonous with an Arabic word meaning “thing”.10 In their L1 Berber performance these speakers 10. The fact that the postverbal negator in several north African Arabic varieties remains an unreduced form ši/šey to this day (Lucas 2009: 23), and thus fully homophonous with a word for “thing”, suggests that this was more generally the case in the period of transfer of the bipartite construction to Berber. © 2012. John Benjamins Publishing Company All rights reserved 294 Christopher Lucas therefore recreate the Arabic bipartite construction by also adding a postverbal negator homophonous with a word meaning “thing”. Once this construction has made it into these speakers’ L1 Berber performance as a result of this process, the possibility of children acquiring this construction as part of their L1 Berber is opened up as in (18b). We can also capture the development of Yiddish dos-sentences in terms of (18). The RL is Yiddish and the SL is Polish (perhaps also other Slavic varieties). Construction Sx is the Slavic sentence type illustrated in (8)–(9), in which there is an expletive neuter demonstrative pronoun in initial position, the proposition communicated is interpreted as hearer-old, except for the instantiation of one of the variables, and there is contrastive focus on this element. Construction Rx is the Yiddish equivalent with dos in initial position, illustrated in (5). The first stage of (18) in this case presumably proceeded in directly parallel fashion to the development of bipartite negation in Berber. Yiddish-dominant speakers’ extensive use of Polish includes use of the construction in (8)–(9). This construction is perceived as structurally similar to the existing Yiddish (and German) es-construction, but discourse-functionally distinct. Speakers also perceive (consciously or unconsciously) that the expletive element in initial position is the neuter demonstrative pronoun (or an item homophonous with it). The recreation of the discourse function of the Polish construction in these speakers’ L1 Yiddish performance is therefore achieved by extending the use of the Yiddish neuter demonstrative dos to the initial position occupied by es in es-sentences and interpreting it in this initial position as an expletive signal of the appropriate discourse function. Again, if the demands of efficient processing result in this extension systematically appearing in the performance of these speakers, despite not being a part of their L1 grammar, then the potential is there for the construction to form part of a child’s PLD and to be acquired as part of their L1 grammar of Yiddish. Finally, we can also see the performance attrition in Turkish binding described in §3 in these terms. Here the RL is Turkish and the SL English. Sx represents the binding properties of the English third person anaphors him/her, in particular that embedded clauses and possessive noun phrases are the minimal governing categories within which him/her cannot be co-indexed with another referring expression. The strong activation of the neural networks associated with Sx in the Turkish-dominant speakers’ L2 English persists in their L1 Turkish performance, such that Sx is recreated in that performance as a new RL structure Rx. In Rx the Turkish third person anaphor o is interpreted as being obligatorily free within the same governing categories as English, whereas in unattrited Turkish the minimal governing category for o is the whole matrix clause. As with Berber and Yiddish, the presence of Rx in these speakers’ performance opens up the possibility of Rx being acquired as part of children’s L1 grammar of Turkish, as in (18b). Until we © 2012. John Benjamins Publishing Company All rights reserved Contact-induced grammatical change 295 can establish that this has in fact occurred, however, this development is better seen as a potential rather than an actual change. 6. Conclusion The aim of this article has been to move towards a more explicit, cognitively-based model of contact-induced syntactic change. I have argued that contact-induced syntactic change can be integrated into a unified model for grammatical change in general, based on Andersen’s (1973) notion of change through abduction. Although the sociological factors that that feed into contact-induced change are often so complex and varied as to lie beyond the scope of such a model, the change itself need not and should not. If Lightfoot (1997: 269) is right that “there can be no change in grammars without change in trigger experiences”, then both internal and externally-caused syntactic change must consist in children abducing their grammatical competence on the basis of a set of PLD which is not fully consistent with the competence of older groups of native speakers. The task of the historical linguist is then to account for how the trigger experience for these children came to be different from that of the older groups. One common way for this to happen is through contact with another language. I suggest that we cannot understand how contact leads to change in trigger experiences without Van Coetsem’s (1988, 2000) distinction between RL and SL agentivity. Speakers who have learnt an RL as a second language will typically produce utterances which are ungrammatical for native speakers of the RL. This appears to be because the general learning strategies they use to acquire the RL differ markedly from the way in which children acquire their L1. Sometimes second language learners draw on some mental representation of their L1 and thus transfer elements of it onto the RL (imposition), and on other occasions their (mis)analysis of the RL leads to their producing structures which are grammatical neither in the RL nor in their L1. In either case these hitherto ungrammatical structures may then form part of the trigger experience for children acquiring the RL, thus resulting in a change under SL (or L2) agentivity. By contrast, native speakers of an RL who have a high degree of exposure to an SL will tend, in a variety of ways, to recreate elements of the syntax of that SL in their RL speech, resulting in utterances that are ungrammatical with respect to their own underlying competence. This appears to be because the (often involuntary) recreation of elements of their L2 in their L1 performance is a means of minimizing the processing costs associated with the extensive use of two languages. Again, if these utterances form part of the trigger experience for children acquiring the RL, this could result in a change at the level of competence, this time © 2012. John Benjamins Publishing Company All rights reserved 296 Christopher Lucas due to RL agentivity (borrowing). This, I suggest, is a useful way of understanding how contact leads to change in grammars. References Abun-Nasr, Jamil. 1971. A History of the Maghrib. Cambridge: Cambridge University Press. Adila, Aziz. 1996. “La négation en arabe marocain”. La négation en berbère et en arabe maghrébin ed. by Salem Chaker & Dominique Caubet, 99–116. Paris: L’Harmattan. Altenberg, Evelyn. 1991. “Assessing First Language Vulnerability to Attrition”. First Language Attrition ed. by Herbert Seliger & Robert Vago, 189–206. Cambridge: Cambridge University Press. Altenberg, Evelyn & Robert Vago. 2004. “The Role of Grammaticality Judgments in Investigating First Language Attrition”. First Language Attrition: Interdisciplinary perspectives on methodological issues ed. by Monika Schmid, Barbara Köpke, Merel Keijzer & Lina Weilemar, 105–129. Amsterdam: John Benjamins. Andersen, Henning. 1973. “Abductive and Deductive Change”. Language 49.765–793. Ås, Arvid. 1962. “The Recovery of a Forgotten Language Through Hypnotic Age Regression: A case report”. The American Journal of Clinical Hypnosis 5.24–29. Behagel, Otto. 1928. Deutsche Syntax: Eine geschichtliche Darstellung, vol. 3. Heidelberg: Carl Winter. Birnbaum, Solomon. 1979. Yiddish: A survey and grammar. Manchester: Manchester University Press. Bley-Vroman, Robert. 1989. “What is the Logical Problem of Foreign Language Learning?” Linguistic Perspectives on Second Language Acquisition ed. by Susan M. Gass & Jacqueline Schachter, 41–68. Cambridge: Cambridge University Press. Bley-Vroman, Robert. 2009. “The Evolving Context of the Fundamental Difference Hypothesis”. Studies in Second Language Acquisition 31.175–198. Bosch, Elena, Francesc Calafell, Anna Pérez-Lezaun, Jordi Clarimón, David Comas, Eva Mateu, Rosa Martínez-Arias, Bernal Morera, Zahra Brakez, Omar Akhayat, Abdelaziz Sefiani, Ghania Hariti, Anne Cambon-Thomsen & Jaume Bertranpetit. 2000. “Genetic Structure of North-west Africa Revealed by STR Analysis”. European Journal of Human Genetics 8.360–366. Boumalk, Abdallah. 1996. “La négation en berbère marocain”. La négation en berbère et en arabe maghrébin ed. by Salem Chaker & Dominique Caubet, 35–48. Paris: L’Harmattan. Campbell, Lyle. 2001. “What’s Wrong with Grammaticalization?”. Language Sciences 23.113– 161. Caubet, Dominique. 1996. “La négation en arabe maghrébin”. La négation en berbère et en arabe maghrébin ed. by Salem Chaker & Dominique Caubet, 79–97. Paris: L’Harmattan. Carruthers, Peter. 2005. “The Case for Massively Modular Models of Mind”. Contemporary Debates in Cognitive Science ed. by Robert Stainton, 205–225. Oxford: Blackwell. Chaker, Salem. 1996. “Remarques préliminaires sur la negation en berbère”. La négation en berbère et en arabe maghrébin ed. by Salem Chaker & Dominique Caubet, 9–22. Paris: L’Harmattan. Chomsky, Noam. 1980. Rules and Representations. New York: Columbia University Press. © 2012. John Benjamins Publishing Company All rights reserved Contact-induced grammatical change 297 Clahsen, Harald & Pieter Muysken. 1986. “The Availability of Universal Grammar to Adult and Child Learners: A study of the acquisition of German word order”. Second Language Research 2.93–119. Crawford, David. 2002. “Morocco’s Invisible Imazighen”. Journal of North African Studies 7:1.53–70. DeKeyser, Robert. 2000. “The Robustness of Critical Period Effects in Second Language Acquisition”. Studies in Second Language Acquisition 22.499–533. Fadhlaoui-Zid, K., S. Plaza, F. Calafell, M. Ben Amor, D. Comas & A. Bennamar El Gaaied. 2004. “Mitochondrial DNA Heterogeneity in Tunisian Berbers”. Annals of Human Genetics 68.222–233. Flores, Cristina. 2007. “Sprachverlust im Kontext deutsch–portugiesische Remigration”. Wildern in luso-austro-deutschen Sprach- und Textgefilden. Festschrift für Erwin Koller ed. by Orlando Grossegesse & Cristina Flores, 87–100. Braga: Cehum, Universidade do Minho. Footnick, Rosalie. 2007. “A Hidden Language: Recovery of a ‘lost’ language is triggered by hypnosis.” Language Attrition: Theoretical perspectives ed. by Barbara Köpke, Monika S. Schmid, Merel Keijzer & Susan Dostert, 169–187. Amsterdam: John Benjamins. Friedman, Victor & Brian D. Joseph. Forthcoming. The Balkan Languages. Cambridge: Cambridge University Press. Fromm, Erika. 1970. “Age Regression with Unexpected Reappearance of a Repressed Childhood Language”. International Journal of Clinical and Experimental Hypnosis 18:2.79–88. Gast, Volker & Johan van der Auwera. 2009. “What is Contact-induced Grammaticalization? Evidence from Mayan and Mixe-Zoquean languages”. Ms., Freie Universität Berlin and Universiteit Antwerpen. Gürel, Ayse. 2004. “Selectivity in L2-induced L1 Attrition: A psycholinguistic account”. Journal of Neurolinguistics 17.53–78. Håkansson, Gisela, Manfred Pienemann & Susan Sayehli. 2002. “Transfer and Typological Proximity in the Context of Second Language Processing”. Second Language Research 18:3.250–273. Hale, Mark. 1998. “Diachronic Syntax”. Syntax 1:1.1–18. Heine, Bernd & Tania Kuteva. 2003. “On Contact-induced Grammaticalization”. Studies in Language 27:3.529–572. Heine, Bernd & Tania Kuteva. 2005. Language Contact and Grammatical Change. Cambridge: Cambridge University Press. Hoffman, Katherine. 2006. “Berber Language Ideologies, Maintenance, and Contraction: Gendered variation in the indigenous margins of Morocco”. Language & Communication 26:2.144–167. Hoffman, Katherine. 2008. We Share Walls: Language, land and gender in Berber Morocco. Oxford: Blackwell. Jacobs, Neil. 2005. Yiddish: A linguistic introduction. Cambridge: Cambridge University Press. Johanson, Lars. 2002. Structural Factors in Turkic Language Contact. London: Curzon. Johnson, Jacqueline S., Kenneth D. Shenkman, Elissa L. Newport & Douglas L. Medin. 1996. “Indeterminacy in the Grammar of Adult Language Learners”. Journal of Memory and Language 35.335–352. Julien, Charles-André. 1961. Histoire de l’Afrique du Nord. Paris: Payot. Karatsareas, Petros. 2007. “Contact-induced Morphological Change: The ‘leap’ from fusional to inflectional morphology in Cappadocian”. M.Phil. dissertation, University of Cambridge. © 2012. John Benjamins Publishing Company All rights reserved 298 Christopher Lucas Keenan, Edward. 2002. “Explaining the Creation of Reflexive Pronouns in English”. Studies in the History of the English Language: A millennial perspective ed. by Donka Minkova & Robert P. Stockwell, 325–354. Berlin: Mouton de Gruyter. Köpke, Barbara. 2004. “Neurolinguistic Aspects of Attrition”. Journal of Neurolinguistics 17.3–30. Köpke, Barbara. 2007. “Language Attrition at the Crossroads of Brain, Mind and Society”. Language Attrition: Theoretical perspectives ed. by Barbara Köpke, Monika S. Schmid, Merel Keijzer & Susan Dostert, 9–37. Amsterdam: John Benjamins. Lightfoot, David. 1997. “Shifting Triggers and Diachronic Reanalyses”. Parameters of Morphosyntactic Change ed. by Ans van Kemenade & Nigel Vincent, 253–272. Cambridge: Cambridge University Press. Longobardi, Giuseppe. 2001. “Formal Syntax, Diachronic Minimalism, and Etymology: The history of French chez”. Linguistic Inquiry 32.275–302. Louden, Mark. 1992. “Old Order Amish Verbal Behavior as a Reflection of Cultural Convergence”. Diachronic Studies on the Language of the Anabaptists ed. by Kate Burridge & Werner Enninger, 264–278. Bochum: Brockmeyer. Louden, Mark. 1997. “Linguistic Structure and Sociolinguistic Identity in Pennsylvania German Society”. Languages and Lives: Essays in honor of Werner Enninger ed. by James R. Dow & Michèle Wolff, 79–91. New York: Peter Lang. Louden, Mark. 2000. “Contact-induced Phonological Change in Yiddish: Another look at Weinreich’s riddles”. Diachronica 17:1.85–110. Lucas, Christopher. 2007. “Jespersen’s Cycle in Arabic and Berber”. Transactions of the Philological Society 105:3.398–431. Lucas, Christopher. 2009. “The Development of Negation in Arabic and Afro-Asiatic”. Ph.D. dissertation, University of Cambridge. Manni, Franz, Pascal Leonardi, Abdelhamid Barakat, Hassan Rouba, Evelyne Heyer, Michael Klintschar, Ken McElreavey & Luis Quintana-Murci. 2002. “Y-chromosome Analysis in Egypt Suggests a Genetic Regional Continuity in North-Eastern Africa”. Human Biology 74:5.645–658. Meisel, Jürgen. 2011. First and Second Language Acquisition: Parallels and differences. Cambridge: Cambridge University Press. Mettouchi, Amina. 1996. “La négation dans les langues du Maghreb: Synthèse”. La négation en berbère et en arabe maghrébin ed. by Salem Chaker & Dominique Caubet, 177–195. Paris: L’Harmattan. Möhring, Anja & Jürgen Meisel. 2003. “The Verb–Object Parameter in Simultaneous and Successive Acquisition of Bilingualism”. (In)vulnerable Domains in Multilingualism ed. by Natascha Müller, 295–330. Amsterdam: John Benjamins. Nadkarni, Mangesh V. 1975. “Bilingualism and Syntactic Change in Konkani”. Language 51.672–683. Paul, Hermann. 1989. Mittelhochdeutsche Grammatik ed. by Peter Wiehl & Siegfried Grosse. 23rd ed. Tübingen: Niemeyer. Prince, Ellen. 2001. “Yiddish as a Contact Language”. Creolization and Contact ed. by Norval Smith & Tonjes Veenstra, 263–289. Amsterdam: John Benjamins. Schmid, Monika. 2002. First Language Attrition, Use and Maintenance: The case of German Jews in Anglophone countries. Amsterdam: John Benjamins. Schuchardt, Hugo. 1884. Slawo-deutsches und Slawo-italienisches. Graz: Leuschner & Lubensky. Schwartz, Bonnie D. 1998. “The Second Language Instinct”. Lingua 106.133–160. © 2012. John Benjamins Publishing Company All rights reserved Contact-induced grammatical change 299 Schütze, Carson. 1996. The Empirical Base of Linguistics: Grammaticality judgments and linguistic methodology. Chicago: University of Chicago Press. Seliger, Herbert. 1996. “Primary Language Attrition in the Context of Bilingualism”. Handbook of Second Language Acquisition ed. by William C. Ritchie & Tej K. Bhatia, 605–626. New York: Academic Press. Sharwood Smith, Michael & Paul van Buren. 1991. “First Language Attrition and the Parameter Setting Model”. First Language Attrition ed. by Herbert Seliger & Robert Vago, 17–30. Cambridge: Cambridge University Press. Siegel, Jeff. 2003. “Substrate Influence in Creoles and the Role of Transfer in Second Language Acquisition”. Studies in Second Language Acquisition 25.185–209. Stauble, Anne-Marie. 1984. “A Comparison of a Spanish–English and a Japanese–English Second Language Continuum: Negation and verb morphology”. Second Languages: A crosslinguistic perspective ed. by Roger Andersen, 323–353. Rowley, Mass.: Newbury House. Stolberg, Doris & Alexandra Münch. 2010. “ ‘Die Muttersprache vergisst man nicht’ — or do you? A case study in L1 attrition and its (partial) reversal”. Bilingualism: Language and cognition 13:1.19–31. Thomason, Sarah G. & Terence Kaufman. 1988. Language Contact, Creolization, and Genetic Linguistics. Berkley: University of California Press. Tsimpli, Ianthi, Antonella Sorace, Caroline Heycock & Francesca Filiaci. 2004. “First Language Attrition and Syntactic Subjects: A study of Greek and Italian near-native speakers of English”. International Journal of Bilingualism 8:3.257–277. Van Coetsem, Frans. 1988. Loan Phonology and the Two Transfer Types in Language Contact. Dordrecht: Foris. Van Coetsem, Frans. 2000. A General and Unified Theory of the Transmission Process in Language Contact. Heidelberg: Winter. Weinreich, Uriel. 1953. Languages in Contact: Findings and problems. New York: Linguistic Circle of New York. (Reprinted 1986, The Hague: Mouton.) White, Lydia. 1991. “Adverb Placement in Second Language Acquisition: Some effects of positive and negative evidence in the classroom”. Second Language Research 7:2.133–161. Winford, Donald. 2005. “Contact-induced Changes: Classification and processes”. Diachronica 22:2.373–427. Résumé Il est évident que le contact des langues est un des facteurs principaux menant au changement de grammaires, et pourtant les recherches sur le changement syntaxique, surtout dans la tradition générative, ont eu tendance à négliger le rôle de contact dans ce processus. En même temps, les travaux sur les changements provoqués par le contacte sont restés largement descriptifs, la discussion théorique se cantonnant dans une large mesure à la portée de l’emprunt linguistique. Cet article vise à dépasser ces restrictions en présentant un compte-rendu de quelques-uns des mécanismes cognitifs par lesquels le contact conduit au changement linguistique. Le point de départ de ce compte-rendu est la distinction de Van Coetsem (1988, 2000) entre ‘recipientlanguage agentivity’ et ‘source-language agentivity’ et l’article développe cette idée à la lumière des recherches empiriques sur l’acquisition des langues secondes et l’attrition des langues maternelles, tout en montrant comment ces principes peuvent être intégrés dans un modèle © 2012. John Benjamins Publishing Company All rights reserved 300 Christopher Lucas acquisitioniste unifié du changement syntaxique en général. Le modèle est aussi appliqué à des études de cas en yiddish et berère. Zusammenfassung Unter den Faktoren, die zu grammatischem Wandel führen, spielt Sprachkontakt eindeutig eine Schlüsselrolle. Dennoch wurde der Einfluss von Sprachkontakt auf diesen Prozess in Arbeiten über syntaktischen Wandel bisher vernachlässigt, insbesondere in der generativen Tradition. Zugleich ist die Forschung über kontakt-induzierten Sprachwandel bisher weitgehend deskriptiv und die theoretische Diskussion beschränkt sich größtenteils auf die vermeintlichen Grenzen der Entlehnung. Ziel dieses Artikels ist es, diese Beschränkungen durch die Entwicklung eines pyscholinguistisch basierten Modells einiger Aspekte des kontakt-induzierten Sprachwandels zu überwinden. Dieses Modell nimmt Van Coetsems Unterscheidung zwischen Empfängersprachen-Agentivität (‘recipient-language agentivity’) und QuellsprachenAgentivität (‘source-language agentivity’) als Ausgangspunkt. Van Coetsems Einsichten werden unter Berufung auf empirische Forschung zu Zweitspracherwerb und Muttersprachverlust weiterentwickelt. Es wird gezeigt, wie Ergebnisse aus diesen verschiedenen Bereichen in ein einheitliches Modell syntaktischen Wandels integriert werden können. Dieses Modell wird anschließend auf Fallstudien zu kontakt-induziertem syntaktischen Wandel im Jiddischen und den Berber-Sprachen angewandt. Author’s Address Department of Linguistics SOAS, University of London Thornhaugh Street Russell Square London WC1H 0XG cl39@soas.ac.uk © 2012. John Benjamins Publishing Company All rights reserved