Effects of the Second Language on the First

Chapter 1

Introduction: The Changing L1 in the L2 User's Mind

VIVIAN COOK

In 1953 Ulrich Weinreich talked about interference as ‘those instances of deviation from the norms of either language which occur in the speech of bilinguals as a result of their familiarity with more than one language’ (Weinreich, 1953: 1). This fits with everybody's common-sense belief that your first language (L1) has an effect on your second language (L2). The foreign accents we hear confirm this every day; an English speaker can tell whether someone is French or Japanese after a few words of English. In the fifty years since Weinreich's book, there has been extensive research into how the learning and use of a second language is affected by the first language, whether conceived as Contrastive Analysis, transfer, cross-linguistic influence, resetting of parameters or in many other ways.

Yet few people seemed to notice that Weinreich's definition concerned deviation from either language. As well as the first language influencing the second, the second language influences the first. Perhaps this effect is less detectable in our everyday experience: only complex instrumental analysis of a Spanish speaker's accent in Spanish will reveal whether the speaker also knows English. It becomes blatant only when the first language starts to disappear, for instance when a speaker brings more and more L2 words into his or her first language.

This volume is perhaps the first book to be devoted only to the effects of the second language on the first, sometimes called ‘reverse’ or ‘backward’ transfer. It arose out of an invitational workshop held in Wivenhoe House in 2001, at which all the papers included in this volume were delivered, apart from two (Porte, Chapter 6; Cook et al., Chapter 10). By using a variety of perspectives, methodologies and languages, the research reported here shows that the first language of people who know other languages differs from that of their monolingual peers in diverse ways, with consequences for second language acquisition research, linguistics and language teaching. The range of contributions shows the extent to which this question impinges not only on all the areas of language from vocabulary to pragmatics, but also on a variety of contemporary approaches currently being developed by second language acquisition (SLA) researchers.

The book is intended for researchers in second language acquisition research and bilingualism, students and teachers around the world. The breadth of the contributions in terms of countries, languages, aspects of language and theories means that it relates to most SLA courses at some point, whether at undergraduate or postgraduate level.

This introduction provides some background to the different contributions in this volume. It tries not to steal their thunder by anticipating their arguments and conclusions, but provides a more personal overview, with which of course not all of the writers will be in complete accord. It relies in part on a summary overview of issues provided to the writers by Batia Laufer after the conference. It does not attempt to deal with the vast areas of language transfer from L1 to L2 or with the field of language attrition, covered in such classic texts as Odlin (1989) or Weltens et al. (1986).

Multi-competence

For me, and for many of the contributors, the question of L2 effects on the L1 arose out of the notion of multi-competence. Initially the term was used almost as a convenience. While ‘interlanguage’ had become the standard term for the speaker's knowledge of a second language, no word existed that encompassed their knowledge of both the second language and their first: on the one hand the L1, on the other the interlanguage, but nothing that included both. Hence ‘multi-competence’ was introduced to mean ‘knowledge of two or more languages in one mind’ (Cook, 1991). For convenience we will mostly talk about ‘second language’ and bilingualism here, but this does not preclude multiple languages and multilingualism.

Since the first language and the other language or languages are in the same mind, they must form a language super-system at some level rather than be completely isolated systems. Multi-competence then raised questions about the relationship between the different languages in use. How do people code-switch fluently from one language to another? How do they ‘gate out’ one language while using the other (Lambert, 1990)? How do they manage more than one pragmatic and phonological system? Multi-competence also raised questions about cognition. Does an L2 user have a single set of ideas in the mind, more than one set of ideas, a merged set from different languages, or a new set of ideas unlike the sum of its parts? And multi-competence also led inevitably to questions about acquisition. What roles do the first language and the other language or languages play in the creation of knowledge of the second or later languages?

Multi-competence led me in particular to a re-valuing of the concept of the native speaker (Cook, 1999). While the concept of interlanguage had seemed to establish the second language as an independent language system, in effect SLA research still treated the L2 system in an L2 user as an approximation to an L1 system in someone else (i.e. a monolingual L1 user). SLAresearch methods compared knowledge of L2 syntax against the knowledge of native speakers (Cook, 1997). Whether L2 learners had access to Universal Grammar (UG) was seen as a matter of whether they learnt the same grammars as monolingual native speakers – ‘slightly over half of the non-native speakers typically exhibit the correct UG-based judgements on any given UG effect’ (Bley-Vroman et al., 1988: 24). Whether age affected L2 learning was seen in terms of how close people came to monolingual native speakers – ‘whether the very best learners actually have native-like competence’ (Long, 1990: 281). Whether they had an accent was a matter of how native-like they were – ‘the ultimate goal – perhaps unattainable for some – is, nonetheless, to sound like a native speaker in all aspects of the language’ (González-Nueno, 1997: 261). The independence of interlanguage was largely illusory, since the norm against which the L2 user was compared was almost universally the native speaker, whether overtly or covertly.

The arguments against the native-speaker standard have been mounting over the past ten years. Let us first define the native speaker as ‘a monolingual person who still speaks the language they learnt in childhood’ (Cook, 1999). This combines the priority of the language in the development of the individual and the continuity of use by the individual with the usual simplifying assumption in linguistics that native speakers are monolingual. It does not preclude the possibility of a person being a native speaker of more than one language, if he or she acquired them simultaneously while a child. By this definition, however, it is impossible for an L2 user to become a native speaker – one reason why so many L2 users think of themselves as ‘failures’ and so many SLA researchers treat them in the same way: ‘learner's language is deficient by definition’ (Kasper & Kellerman, 1997: 5).

The main arguments against the use of native speakers as the norm against which L2 users should be measured are as follows:

The rights of L2 users

One group of human beings should not judge other people as failures for not belonging to their group, whether in terms of race, class, sex or language. People should be measured by their success at being L2 users, not by their failure to speak like native speakers. The object of acquiring a second language should be to become an L2 user, not to pass for a native speaker. SLA research has to do justice to its constituency – people who know two languages – not subordinate them to people who know only one language. The L2 user is a person in his or her own right (Cook, 1997; Grosjean, 1989), not an imitation of someone else.

The numbers of L2 users

It is hard to arrive at precise figures about the numbers of monolingual native speakers in the world. It is slightly easier in reverse to find some numbers for people who are learning or using second languages. Taking English as an example, the British Council (1999) claims that a billion people are studying English in the world, including all children over 12 in Japan. English is used everywhere for certain purposes (such as academic journals and the Internet); many people communicate with each other through English who have never met a native speaker (for example business people doing international deals). Some countries where English is hardly spoken at all natively (such as Singapore) deliberately use it as a ‘first language’; others (such as Nigeria, Cameroon, India and Pakistan) employ it as an official language. Turning away from English, most people in, say, Cape Town, Islamabad or Brussels switch from one language to one or more other languages in their daily lives. Monolingual native speakers are far from typical of human beings and are increasingly hard to find in the world (as we shall see in some of the contributions here), even in the highlands of Papua New Guinea. While it may be hard to prove that L2 users actually make up the majority of human beings, they at least form a very substantial group.

The usual resort in SLA and bilingualism research is to see the L2 standard in terms of the balanced bilingual or ‘ambilingual’. Toribio (2001: 215), for instance, defines a balanced bilingual as ‘a speaker who has native-like ability in two languages’ and sees the standard against which an L2 user is measured as being ‘an idealised bilingual's native speaker competence’. While the construct of native-speaker competence may be appropriate in first language acquisition as all human beings attain it, the concept of idealised bilingual competence can be extremely misleading since so few L2 users attain it. How many people have native-like skills in both languages in a reasonable range of their contexts of language use? They are the exception rather than the norm among L2 users, defined by their ability to function like native speakers in two languages not by their whole language ability to use two languages. The use of a native-speaker measure that is virtually impossible to achieve, even when disguised as the doublemonolingual native speaker of the balanced bilingual, will blind us in the future (as it has done in the past) to the overwhelming majority of L2 users who are far from native-like across two languages. First language acquisition research is about what most people achieve, not about the abilities of monolingual Shakespeares. Second language acquisition research should equally be about what typical L2 users achieve, not about bilingual Nabokovs. Hence I now try to avoid the word ‘bilingual’ in discussing people who know two languages, not only because of the plethora of confusing definitions, but also because they usually invoke a Platonic ideal of the perfect bilingual rather than the reality of the average person who uses a second language for the needs of his or her everyday life.

The distinctive characteristics of L2 users

If L2 users are different kinds of people, the interest of SLA research lies in discovering their characteristics, not their deficiencies compared with native speakers. In Cook (2002a: 4–8) the characteristics of L2 users are stated as four propositions:

(1) the L2 user has other uses for language than the monolingual;

(2) the L2 user's knowledge of the second language is typically not identical to that of a native speaker;

(3) the L2 user's knowledge of his or her first language is in some respects not the same as that of a monolingual;

(4) L2 users have different minds from those of monolinguals.

This book is thus primarily an expansion and justification of proposition (3) that L2 users differ from monolingual native speakers in their knowledge of their first languages. Inevitably it simultaneously provides further information about the distinctive nature of the L2 users’ uses for language, their knowledge of their second language, and their minds.

Multi-competence led to seeing the L2 user as a person in his or her own right, not as an approximation to a monolingual native speaker. This is why I prefer the term ‘L2 user’ to ‘L2 learner’ in recognition of the person's ability to use the language rather than remaining a learner in perpetuity, always recognising that the same person may be both ‘learner’ and ‘user’ in different aspects of his or her language identity.

The belief in the native-speaker standard is one reason why the effects of the L2 on the L1 were so little studied. If the L1 of the L2 user were different from that of monolingual native speakers, SLA research that used the native speaker as the target would be based on shifting sand. As argued in Cook (2002a), a comparison of the L2 user with the native speaker may be legitimate provided any difference that is discovered is not treated as a matter of deficiency. Persistent use of this comparison prevents any unique features of the L2 user's language being observed, since only those that occur in natives will be searched for. For many years this led, for example, to a view that code-switching in adults or children was to be deplored rather than commended; Genesee (2002), for instance discusses how young children's code-switching was interpreted as a sign of confusion rather than as skilful L2 use.

While this argument has been couched in terms of multi-competence, this is not the only approach for dealing with the effects of the L2 on the L1. In this volume we find general models such as the dynamic model of multilingualism of Jessner (Chapter 12), the Common Underlying Conceptual Base (CUCB) of Kesckes and Papp (Chapter 13), Karmiloff-Smith's Representational Redescription Model, and Bialystok's Analysis/Control Model (Murphy and Pine, Chapter 8), and variants on the Chomskyan Minimalist Program used by Balcom (Chapter 9) and Satterfield (Chapter 11). Most of these share the assumption that at some level the L2 user's mind is a whole that balances elements of the first and second language within it. Furthermore, as Satterfield argues, this is essentially the normal state that all human beings can reach, and so must form the basis for any account of human language. If monolingualism is taken as the normal condition of humanity, L2 users can be treated as footnotes to the linguistics of monolingualism. With most people in the world learners or users of second languages, however, monolinguals can be considered the exception, not only statistically but also in terms of human potential.

The Relationship of the First and Second Languages in the Mind: The Integration Continuum

What could the logical relationships actually be between the two or more languages in the mind? One possibility is that the languages are in watertight compartments, seen in the separation model in Figure 1.1, akin to the idea of coordinate bilingualism associated with Weinreich (1953); the L2 user speaks either one language or the other, with no connection between the different languages in the mind. The early SLA research controversies about the natural order of acquisition asserted a separation model in which the L2 interlanguage developed without drawing on the L1 to any great extent (e.g. Dulay & Burt, 1980). The separation model forms the basis for much language teaching methodology that teaches without reference to the first language and discourages its use in the classroom, hoping that the students will build up a new language system with no links to the first. This model sees no point to discussing the effects of the L2 on the L1, as they do not exist. Separation does not, by the way, imply anything one way or the other about universals of language whether language design (Hockett, 1960) or innate properties of the mind (Chomsky, 2000). Both separate languages might be similar because they are governed by the same constraints and potentials as any other language acquired by a human being.

Figure 1.1 Separation model

The opposite possibility is that the languages form a single system, shown in the integration model in Figure 1.2. In the area of vocabulary some people have claimed that, rather than two separate mental lexicons, the L2 user has a single lexicon where words from one language are stored alongside words from the other (Caramazza & Brones, 1980). In terms of phonology some have found that L2 users have a single merged system for producing speech, neither L1 nor L2 (Williams, 1977). Integration does not say that L2 users are unable to control what they do; they can still choose which language to use in a given context, just as a monolingual can choose which style or register to adopt in a particular situation. In this model, the discussion is not about the influence of L2 on L1, but about the balance between elements of a single language system. Indeed there is little point to counting ‘languages’ in a single mind – L1, L2, L3, Ln – as they form a single system.

Clearly neither of these two models can be absolutely true: total separation is impossible since both languages are in the same mind; total integration is impossible since L2 users can keep the languages apart. These possibilities represent the endpoints on the integration continuum (Cook, 2002a; Francis, 1999). In between these two extreme, and probably untenable, positions of total separation and total integration, there are many different degrees and types of interconnection, two of which are shown in Figure 1.3.

Figure 1.2 Integration model

The linked languages model A in Figure 1.3 captures the idea of influence between two essentially separate language systems in the same mind, i.e. it is a variant of the separation model in which the two separate language components interact with each other. This is perhaps the typical model assumed in much SLA research; development and use of the L2 is affected by the already-existing L1, as surveyed say in Odlin (1989). Studies of language ‘transfer’ or ‘influence’ have assumed an interconnection model by seeing how the development of interlanguage (the L2 element in multi-competence) takes advantage of the first language (the L1 element in multi-competence). The linked languages model does, however, allow the links to go in both directions: the Revised Hierarchical Model of the lexicon (Kroll & Tokowicz, 2001) for instance assigns unequal strengths to the links between L1 and L2 according to direction and stage of acquisition.

The variant called the partial integration model (B in Figure 1.3) captures the idea of partial overlapping of the two language systems in the same mind; i.e. it is a limited version of the integration model. Inevitably this is bi-directional in a particular area since, like the integration model, it does not distinguish between languages in the areas of overlap but shows how the single conjoined system differs from monolingual versions of either language. There may be shared or overlapping vocabulary, syntax, or other aspects of language knowledge. Van Hell and Djikstra (2002: 1), for example, show that ‘the multilingual's processing system is profoundly non-selective with respect to language’ and that the lexicon of the language that is not overtly in use is nevertheless still active.

One question is whether the differences between the two conceptualisations of interconnection (linked languages and partial integration) are alternative wordings of the same idea, as implied in Cook (2002a), or represent different types of relationship. Is there a difference between saying that the languages in the mind are merged towards the integration end of the continuum, or that the links between them are stronger? This may partly be settled by the research in this volume, or indeed by neurolinguistic approaches such as that of Fabbro (2002).

Figure 1.3 Interconnection models

Figure 1.4 displays the integration continuum as a whole, and is equivalent to the figure in Cook (2002a: 11). The continuum does not necessarily imply a direction of movement. It may be that some people start with separation and move towards integration or vice versa, or the languages might stay permanently separate. Paradis and Genesee (1996), for example, see bilingual children as having ‘autonomous’ languages from early on rather than ‘interdependent’ languages.

The integration continuum does not necessarily apply to the whole language system (Cook, 2002a); a person's lexicon might be integrated, but the phonology separate. Nor does it necessarily affect all individuals in the same way; some may be more integrated, some not, a factor of individual variation subsuming Weinreich's types of bilingualism. The point on the continuum may also vary from moment to moment in the individual according to his or her perception of language mode (Grosjean, 2001), level of tiredness and other personal factors; or indeed, as Porte suggests in Chapter 6, because of fondness for language play by L2 users. The continuum might also be related to different stages of L2 development; children may move from an integrated single lexicon to a less integrated double lexicon (Taeschner, 1983). One important aspect of this question may indeed be to establish which areas of the L1 are not affected by the L2, for example the lexical diversity or productivity studied by Dewaele and Pavlenko (Chapter 7). The concept of the integration continuum is discussed elsewhere in more detail in the context of bilingual concepts in (Cook 2002a, 2002b).

Figure 1.4 The integration continuum of possible relationships in multi-competence

Again there are several alternatives to the integration-continuum way of conceptualising L2 influence on L1. Pavlenko (Chapter 3) describes five forms of ‘transfer’; Kecskes and Papp (Chapter 13) describe six variations on the L2→L1 effect. Many of these variations concern whether the second language is the majority or minority language in the community or in the individual (Li Wei, 1994), whether it is taught in a classroom or learnt outside, whether it acts as a lingua franca, and so on.

The integration continuum differs in certain ways from the concept of language modes. Grosjean (2001) puts any language use by an L2 user on a continuum of monolingual and bilingual modes according to the proportions of language A and language B, or indeed language C, that are involved. ‘Language mode is the state of activations of the bilingual's languages and language processing mechanisms at a given point in time’ (Grosjean, 2001: 3). An L2 user decides the proportions of the two languages to employ at any given moment in the light of multiple factors on a continuum between effectively activating only one language and activating both simultaneously. The language mode continuum is not then about which language to use but about how much of each. It is like a mixer tap that merges hot and cold water, but neither tap can be completely turned off. The way in which the language mode concept is phrased still implies the existence of two (or more) languages, two (or more) separate mental entities that strike a balance for each language use rather than integration of the two languages. Grosjean (2001), for instance, questions the methodologies that have produced signs of integration in semantics and word perception by pointing out that they paid insufficient attention to language mode. The partial integration model denies, however, that there are two languages as such and favours a single mental system within which a balance can be struck between elements of a particular aspect of language in a particular situation. This is one of the reasons why this introduction has tried to avoid the word ‘transfer’ (the other reasons being its confusing multiple definitions and its emotive connotations): it is questionable whether links between elements in the same mind can be considered transfer without in effect ‘counting’ languages (Cook, 2000).

The Integration Continuum fits with the Dynamic Model of Multilingualism (Jessner, Chapter 12) in trying to see the language system of the L2 user as a whole rather than as an interaction between separate language components. It is similar to the Common Underlying Conceptual Base (Kecskes and Papp, Chapter 13) in seeing the effects of the second language as affecting the whole mind. It is also compatible with the integrated neurolinguistic theory of bilingualism (Paradis, 2001) in unifying both L1 and L2 within the same architecture of the mind. The Integration Continuum does not spell out the separate L1 and L2 components of pragmatics, semantics, morphosyntax and phonology that are part of the Paradis model but, without naming the components, implies that the relationship of integration versus separation varies from component to component. It differs, however, in extending the continuum to concepts, whereas Paradis (2001) has a single unvarying conceptual system. If speakers of different languages do in fact have different concepts, such as relative versus absolute orientation (Levinson, 1996) and form versus substance categorisation of objects (Imai & Gentner, 1997), models of L2 users clearly must accommodate variation in cognition and some relationship between the specifically L1 and specifically L2 concepts (Cook, 2000, 2002b). For example Athanasopoulos (2001) showed that Greeks who knew English had a different perception of the colour ‘blue’ than Greeks who did not.

Positive and Negative Effects on the L1?

The integration continuum has been presented here positively as the separation or integration of two languages in the same mind. However these L2→L1 effects could be evaluated in at least three ways: positive effects on the L1, negative effects on the L1, and effects that are essentially neutral.

The L1 can be enhanced by the use of an L2

It seems obvious that in some sense knowing another language benefits your use of your first language; language teaching classically invoked the concept of ‘brain-training’ to justify the teaching of Latin for example. Hungarian children who know English use measurably more complex sentences in their first language than those who do not (Kesckes & Papp, 2000). Extensive research into bilingual development shows overall that L2 user children have more precocious metalinguistic skills than their monolingual peers (Bialystok, 2001). English children who are taught Italian for an hour a week read English better than those who are not (Yelland et al., 1993). So far as the general use of the first language is concerned, it is an advantage to know a second language, as attested by many celebrated bilingual writers, ranging from Chinua Achebe to John Milton, Samuel Beckett to Rabindranath Tagore. In the present volume the enhancing effects of the second language can be found in the contribution by Murphy and Pine (Chapter 8) on the development of bilingual children in England.

The L1 can be harmed by the use of an L2

The usual context for discussing possible harmful effects of the L2 on the L1 is language loss or attrition. We will set aside here the effects on the L1 of other factors than the L2, such as aphasia caused by brain damage (Paradis, 2000). As a person gains the ability to use a second language, so he or she may to some extent lose the ability to use the first language. In circumstances where one language becomes less and less used, people do lose their command of it, whether as a group or as individuals. Perhaps this is familiar to everybody whose school-learnt language has effectively vanished from their lives. Research into this has mostly been carried out in the context of the loss of the first language by people who are spending their lives in a situation where it is not used for their major everyday social and professional purposes, whether as immigrants or expatriates. Examples of such research in this volume are Porte's account of expat teachers of English (Chapter 6) and Laufer's account of Russian L1 speakers in Israel (Chapter 2).

The L1 is different from the L2, without being better or worse

Positive and negative evaluations of differences are to some extent problematic in that they rely on a value judgement about what is good and what is bad. Enhanced metalinguistic ability is valuable only if it is useful in some definable way; losing some aspect of the first language is a disadvantage only if it prevents the L2 user from carrying out some activity successfully.

Many of the effects of the L2 on the L1 simply amount to differences. The L2 user mind is bound to have differences in the first language element because of its more complex linguistic organisation, whether through linking or integration. In phonology the extensive literature on Voiced Onset Times in L2 users, surveyed in Watson (1991), reveals time and again the differences in the first language of L2 users for plosive consonants such as/p/and/b/or/k/and/g/across pairs of languages such as Spanish/English (Zampini & Green, 2001), French/English (Flege, 1987), and Hebrew/English (Obler, 1982), which are essentially undetectable in normal language use. In this volume the contribution by Cook et al. (Chapter 10) shows differences in L1 sentence-processing by Japanese L2 users that are hard to regard as either