SLA Topics     SLA Bibliography    Vivian Cook 

V.J. Cook
unpublished circa 1999

As de Saussure wryly remarked, 'Other sciences work with objects that are given in advance and that can then be considered from different viewpoints; but not linguistics' (de Saussure, trans. 1959, p.8). This paper argues that Second Language Acquisition (SLA) research should pay more attention to defining the object that second language acquisition is concerned with: what does it mean to know a second language?

Though not given in advance, the object of study of linguistics has been commonly agreed since the 1960s to be linguistic competence - the native speaker's knowledge of language (Chomsky, 1965a). All normal adult human beings have linguistic competence in their first language - competence is indeed defined as whatever it is that native speakers possess; any native speaker can then represent all native speakers. All children, with very few exceptions, succeed in learning their first language over the same period of time: they become native speakers. They do so in a language environment the properties of which are fairly familiar to all native speakers, usually surrounded by parents and other caretakers. The object at which first language acquisition is aiming is evidenced in every native speaker - the so-called steady state, Ss, of adult native competence (Chomsky, 1986). Research into first language acquisition therefore tries to describe the route by which the child reaches a destination that is relatively certain.

The object towards which L2 learners are progressing is far less obvious. There is no settled final state of knowledge, Ss, common by definition to all, or even most, learners, as there is in first language acquisition. Few L2 learners acquire a native-like competence in the L2; their final state ranges from near-beginners able to say a few L2 words in a classroom to writers of world-class literature, say Conrad or Nabakov; the terminal L2 state, St, is a variable not a constant (Cook, 1985). There is not therefore a 'normal' object of Second Language Acquisition to be found in any L2 learner at random. The environment of L2 learners also varies immensely, for example from schools to refugee camps, compared to the primal child-rearing situation. The age of L2 learners varies from early childhood to old age unlike the constant relationship between age and language acquisition in L1 children. Rather than being typical of all human beings, L2 learners are a far from random selection. So SLA research has to describe diverse routes to diverse destinations.

The field of SLA research has mostly dealt with the problem of the indefinite final state by implicitly accepting that the object that L2 learning is tending towards is identical to that of first language acquisition, that is to say, adult native speaker competence: language is what native speakers know. Anything which does not equate to native speaker standards is taken to be wrong. Consequently the only way of acquiring this object in an L2 is to become a bilingual capable of speaking an L2 as if it were an L1 - the so-called balanced bilingual.

But most non-native speakers conspicuously fail to reach this object. They make mistakes; they have a foreign accent; they are not like native speakers. Researchers agree that L2 learners are bad at L2 learning, diverse as their explanations for this may be. A few quotations will show the general trend of thinking. White (1989, p.41) says 'In L2 acquisition, on the other hand, it is common for the learner to fail to acquire the target language fully'; Bley-Vroman (1989, p.43) claims 'The lack of general guaranteed success is the most striking characteristic of adult foreign language learning'; Larsen-Freeman & Long (1991, p.153) lament 'Unfortunately, language mastery is not often the outcome of SLA'; Birdsong (1992, p.706) asserts 'failure to acquire the target language grammar is typical'; Towell and Hawkins (1994, p.14) claim ‘Very few L2 learners appear to be fully successful in the way that native speakers are’. In a well-known paper Selinker (1972) talked of 5% of L2 learners having 'absolute success'. Chomsky (1986, p.16) too insists 'We do not for example say that the person has a perfect knowledge of some language L similar to English but still different from it. What we say is that the child or foreigner has a "partial knowledge of English" or is "on his or her way" towards acquiring knowledge of English, and if they reach this goal, they will then know English'. Quality is therefore to be assessed by native speaker standards; 'Ideally, the end state represents a perfect command of the language' (Klein, 1986, p.50). Spolsky (1989, p.35) establishes a condition of L2 learning that 'Second language learner language approximates native speaker language'. The only good L2 learner is one who has become a native speaker - a logical impossibility since the one thing that the L2 learner is not, by definition, is a native speaker. Judged by the criterion of whether they have achieved native speaker linguistic competence, there are indeed very few successful L2 learners, some might say none.

The independent grammars assumption and Second Language Acquisition research

It might seem that the problem of defining an object of study as the objective of learning was solved for SLA research in the 1960s by what has been called the independent grammars assumption (Cook, 1993). Influenced by Chomsky's ideas, first language acquisition researchers in the 1960s began to recognise the child as a speaker of a language of his or her own rather than as a defective speaker of adult language. The language systems of children were seen as systems in their own right, not as deficient versions of adult language. This led to L1 research that literally wrote independent grammars of children's languages, for example the pivot-open grammar of Braine (1963) or the negation grammars of Klima and Bellugi (1966). The independent grammars assumption was adapted to Second Language Acquisition by several people at roughly the same period - 'transitional idiosyncratic dialect' (Corder, 1971), 'approximative system' (Nemser, 1971), and 'interlanguage' (Selinker, 1972). It can be summed up by Corder, 'Only by treating language learners' language as a phenomenon to be studied in its own right can we hope to develop an understanding of the processes of second language acquisition, just as it is only by treating child language as a phenomenon to be studied in its own right that we can hope to understand something about the processes of first language acquisition and the use that infants make of language' (Corder, 1978, p.71).

One form of the independent grammars assumption in L2 learning was Nemser's idea of approximative system: 'Learner speech at a given time is the patterned product of a linguistic system, La [approximative language], distinct from Ls [source language] and Lt [target language] and internally structured' (Nemser, 1971). Learners have a knowledge of language that is neither L1 nor L2 but is something of its own, a true independent grammar. But what does this approximative system approximate to? Presumably to the target native speaker system; that is to say, the learner's system is still defined by reference to the target; L2 learners are moving towards native competence.

The notion of interlanguage put forward by Selinker (1972) potentially offered a solution to this problem. Interlanguage is an independent system of its own that may 'fossilise' in a form different from monolingual competence; it does not necessarily converge on the target language. In principle the interlanguage formulation gave SLA research the chance of specifying an object for L2 learning; the learner might know a language that was not the L2 but was nevertheless a language, an entity of its own. The concept of interlanguage was in many ways the foundation stone of modern SLA research; 'what gave SLA its excitement was the concept of interlanguage' (Davies, Criper, & Howatt, 1984, p.xii). But the concept of interlanguage was not in itself very clear. Did it stand for an intermediate system related to both L1 or L2, as the name seemed to suggest, or was it a state in its own right? A makeshift expedient or a solid object? Though the originator of the term 'interlanguage' continues to insist that interlanguages are not versions of the target language (e.g. Selinker, 1992), much discussion of interlanguage implicitly accepts a continuum of interlanguages leading towards native competence rather than in any other direction - that is to say an idea of approximation rather than of independence - even if nearly all learners become fossilised en route, as we see in the discussion of SLA methodology below.

Since Chomsky (1964) linguistic competence has been separated from linguistic performance, since Chomsky (1976) from pragmatic competence. Competence as knowledge of language is distinct from performance, in the sense both of the psychological processes the speaker uses to produce or understand speech and of the uses to which language knowledge is put. The independent grammars that researchers write for L1 children are intended to reflect the child's competence, even if Chomsky (1965b) warned of the problems of establishing this from performance data alone. But is L2 interlanguage a competence state of knowledge, or a set of performance processes for speech production, or something else? Selinker (1972) describes five 'central processes in IL performance': language transfer, overgeneralisation of TL rules, transfer of training, strategies of L2 learning, and communication strategies. These are not aspects of language knowledge - of competence as such - but processes for using or creating such competence. Many of them are ambiguous between a process of performance and a process of learning; transfer for example might be a 'synchronic' process happening at the moment that the learner speaks or a 'diachronic' process that occurs throughout the learner's learning experience, in Weinreich's terms the sand carried by the stream or the sand deposited on the lake bottom (Weinreich, 1953, p.11). In neither case are they knowledge of language in the sense of linguistic competence. They still do not settle what the object - the product of the processes - actually is. As Spolsky (1989, p.33) puts it, 'in spite of his use of competence terminology, Selinker seems to prefer a processing model'. The execution of the concept of interlanguage did not establish an L2 object equivalent to the L1 definition of competence, even if in principle it appeared to have this capability. Hence interlanguage led primarily to SLA research dealing with the strategies and processes of learning and use rather than to descriptions of interlanguage competences.

In principle the independent grammars assumption became part and parcel of SLA research through interlanguage or similar concepts. Yet this was not reflected in much research methodology; the mainstream SLA research of the 1970s and 1980s mostly depended implicitly on reference to native competence. Take the notion of obligatory context. In first language acquisition research, Brown (1973) scored presence of grammatical morphemes in obligatory contexts - occasions on which a native speaker is obliged to use particular morphemes in a sentence; he divided these into the four categories of linguistic context, nonlinguistic context, linguistic prior context, and linguistic subsequent context. Given a 'rich' interpretation of native speaker speech, this may be justified; native speakers can spot when a particular grammatical morpheme should appear in sentences of their L1 with hardly any problem. The concept of obligatory context was taken over by SLA researchers within several different research paradigms, usually only in the sense of linguistic context rather than in the four categories used by Brown. Let us take three examples where obligatory contexts are explicitly mentioned: Dulay and Burt (1973) scored whether eight grammatical morphemes occurred in the English of L2 learners; Schumann (1978) scored the percentage with which structures such as negation occurred in obligatory contexts in Alberto's speech; and Meisel, Clahsen, and Pienemann (1981) looked at the occurrence of word order structures in obligatory linguistic contexts for migrant workers learning German. Thus obligatory contexts play their role in work ranging from grammatical morpheme studies to acculturation theory to the Multidimensional Model. But what do obligatory contexts test except approximation to native speaker competence? The obligatory context is based on when the native is presumed to use an item. L2 learners are being measured against native speakers; their own interlanguage does not have status as a system in its own right. The obligatory contexts research technique reveals nothing about the nature of the interlanguage per se, only about its deviations from the native target. True, a researcher could hope to build up a picture of the interlanguage indirectly from such data and could try to see what unique interlanguage system could explain these patterns of deviancy. But this second stage deduction constrains the extent to which the peculiarities of the interlanguage itself can be established to its deviations from native speech; it is like studying the reflections of a tree on the surface of a moving stream. If interlanguage is indeed an independent language, scoring learner speech for obligatory native contexts is as absurd as scoring English for presence of Italian morphemes.

To be faithful to the concept of independent grammars, researchers need to establish the patterns in the interlanguage itself, rather than measuring L2 speech against that of natives. A major claim of SLA research has been that there are similarities in the sequence of development across L2 learners, demonstrated in their different ways by Dulay and Burt (1973), Schumann (1978), and Meisel, Clahsen, and Pienemann (1981). Since obligatory contexts are native-related, the claim has been made with flawed evidence that ignores the independent grammars assumption: we simply do not know if the interlanguage grammars form a consistent series in their own right; we only know that they form a series as measured by the native grammar. This stricture could be overcome by using other forms of evidence; the research with European migrant workers outlined in Klein and Perdue (1992) established a set of five common principles operating at the base stage of L2 acquisition in four different L2s involving five L1s by adopting an Error Analysis style approach to looking at learners' productions in their own terms rather than in native-biassed obligatory contexts.

The same criticism applies to the research technique in which learners are asked to evaluate whether sentences are grammatical in one way or another. For first language acquisition, this might indeed be one means of showing that native speakers or children actually possess some aspect of grammatical knowledge, though in fact the technique is seldom employed. Grammaticality judgments are for example the mainstay of SLA researchers into Universal Grammar theory, with other techniques hardly being used. To take some representative examples, Bley-Vroman, Felix, and Ioup (1988) used them to test L2 learners' knowledge of subjacency as did Johnson and Newport (1991) and White (1985); White (1986) used them to test pro-drop, Gass (1979) to test relative clauses. There are many problematic issues over grammaticality judgments, discussed in Cook (1993); the issue here is that in SLA research the technique of grammaticality judgments implicitly measures whether L2 learners conform to putative native standards. The L2 learner is being measured against native intuitions of grammaticality, contrary to the independent grammars assumption.

Hence, while these two major sources of data in SLA research are seemingly at opposite poles of objective facts and subjective opinions, they are at one in relying on native competence. As are most other sources of evidence. The technique of elicited imitation has measured successful imitation in terms of native use from Cook (1973) to Flynn (1987). Indeed in general, production data by L2 learners has almost always been interpreted as deviant native speech. Error Analysis for example requires the analyst to reconstruct what the learner wants to say from the actual sentence used; 'the methodology of description is, needless to say, fundamentally that of a bilingual comparison' (Corder, 1971). Despite warnings against the Comparative Fallacy of basing interlanguage on the monolingual native (Bley-Vroman, 1983), SLA research still has a built-in bias towards native speaker competence rather than accepting the unique language system of the L2 learner; Birdsong (1993, p.717) describes his own study as construing 'ultimate attainment in L2A in terms of whether nonnatives can display evidence of possessing native linguistic norms'. In general the independent grammars assumption is honoured in the breach, however much lip-service is paid to the concept of interlanguage. To sum up, whether one takes Contrastive Analysis, Error Analysis, or interlanguage, 'the variety of language used by a second language learner is intended to be, and is usefully to be considered as, an approximation to that of a native speaker' (Spolsky, 1989, p.34).

The Object of Bilingualism Research

Alongside the newcomer of SLA research, there has existed a long tradition of research into bilingualism, which to a large extent forms a separate field. Is the object of bilingualism studies different from that in Second Language Acquisition research? Most texts on bilingualism debate at length the varying definitions of bilingualism. One parameter of variation is the amount of language that needs to be known, from maximal - the balanced bilingual 'with native-like control of two languages' (Bloomfield, 1933), otherwise called the ambilingual (Halliday, McIntosh and Strevens, 1964) - to minimal - 'the point where a speaker can first produce complete meaningful utterances in the other language' (Haugen, 1953, p.7). Other parameters concern whether bilingualism is successive or simultaneous, the relationship of learner and target groups, and whether bilingualism should be based on proficiency or use.

Bilingualism researchers are much more reluctant to subordinate their subjects to monolinguals than are SLA researchers. Romaine (1989) insists 'it is clear that a reasonable account of bilingualism cannot be based on a theory which assumes monolingual competence as its frame of reference'. Appel and Muysken (1987, p.3) lament 'All too often imposing Bloomfield's criteria on bilinguals has led to their stigmatisation as being somehow deficient in their language capacities.' Hoffman (1991) states 'For the vast majority of bilinguals, 'bilingual competence' is not measurable in terms of monolingual standards'. Many bilingualism researchers describe the additional benefits of bilingualism over monolingualism, such as cognitive and social advantages, and extra skills such as code-switching. In short the ability to function through two languages rather than one cannot be measured just by monolingual competence: a person who can juggle two balls cannot be limited by the standards of somebody who can juggle only one. The overall emphasis in bilingualism studies is an acceptance of the L2 user as an L2 user to be measured by the standards of L2 users, not by those of monolinguals. The L2 user is not an imitation monolingual in the L2 but a genuine bilingual - a type all of its own: bilingualism is not double monolingualism but a different state. The 'ultimate attainment' of second language is not, and could never be, monolingual competence.

So bilingualism researchers generally deny that the description of bilingualism is parasitic on the monolingual: L2 users are not failed monolinguals but successes in their own right. This leads to a flexible notion of the object of study; 'From whatever angle we look at it, bilingualism is a relative concept' (Hoffman, 1991, p.31). In a sense there is not one object but a shifting range of objects according to the perspective one adopts. This departs from the view of linguistic competence as a uniform phenomenon implied earlier. The reliance on use and process as aspects of bilingualism implies something more akin to performance or to pragmatic competence than to linguistic competence as knowledge of language. Most researchers concur with Mackey (1970) that the definition should be based on use rather than on knowledge; 'Bilingualism is not a phenomenon of language; it is a characteristic of its use'. What counts is whether the speaker can use two languages, not whether he or she 'knows' them. Bilingualism research allies itself more with a sociolinguistic notion of variability. Romaine (1989) explicitly rejects a competence-based approach in favour of use and blames the Saussurean structuralist tradition for 'the belief that an entity, whether it is a society, language, or so forth, can be viewed as a structured self-contained whole, an autonomous entirety, which is consistent with itself' (p.286).

We seem to be no nearer finding a definite object of study. On the one hand we have SLA research using interlanguage derivatives. The interlanguage construct in practice has not maintained the independent grammars assumption of individual linguistic competence; the field is pervaded by the assumption that the knowledge of the L2 is an ineffective version of the monolingual's L1. As Sridhar and Sridhar (1986) point out, 'Paradoxical as it may seem, Second Language Acquisition researchers seem to have neglected the fact that the goal of SLA is bilingualism'. On the other hand we have bilingualism research. This insists on the validity of the L2 state but recognises a concatenation of effects rather than a coherent L2 state of language knowledge. What is needed is an object which preserves the respect for the L2 user while preserving an independent state of knowledge.

A key issue to reconcile these two opposite views is the concept of a language; what is the 'language' that the native speaker or the non-native speaker knows? One use of the term "language" refers to a collection of sentences: English is a set of sentences uttered by people. Linguistics can be seen as an attempt to capture regularities in this objective 'language' corpus; a collection of sentences produced by an L2 learner would constitute an interlanguage, say 'the utterances which are produced when the learner attempts to say sentences of a TL [target language]' (Selinker, 1972). But such description of external reality does not go very far if the aim is to capture the grammar internal to the speaker's mind. Though the concept of language as a set of external data is relevant to the learnability arguments of the Subset Principle (Cook, 1993), it is not now taken as a central issue of linguistics, which is directly concerned with the mental representation of language in the speaker's mind rather than with patterns in the external data. Hence remarks such as Bialystok (xx) 'adults speaking a second language which they have not perfectly mastered' are very hard to interpret; what is this external "language" which L2 learners have failed to 'master'? It appears to be an absolute object separate from knowledge of language

The word "language" is also used to refer to the general phenomenon of the English language, the French language, and so on; a language in this sense is a social possession of a body of speakers who communicate with each other - the speakers of English, of French, etc. This shades into the use of "language" to refer to a socio-political concept; Chinese is a language despite having at least eight mutually incomprehensible spoken forms called 'dialects'; Danish and Swedish are different languages despite being mutually comprehensible. But an interlanguage lacks any community of mutually intelligible speakers, except more or less by accident where a group of students in the same environment have similar interlanguages. An interlanguage is peculiar to a particular L2 learner at a particular time; possibly noone else speaks it in the world. So interlanguage cannot be a language in the general sense of a social phenomenon.

One alternative is to relate interlanguage to some existing language. Weinreich (1953, p.7) said that 'A structuralist theory of communication which distinguishes between speech and language ... necessarily assumes that "every speech event belongs to a definite language"'. In this view whatever somebody says belongs either to language A or to language B. So what language does the L2 learner's interlanguage belong to in this broad sense? In practice interlanguages have been spoken of as an English interlanguage, a French interlanguage, etc, as if they belong to a family of English-like languages, French-like languages, etc. But the independent grammars assumption refuses to class the learner's utterances as either A or B but as a third language C; even if the learners are on their way to B, their interlanguage utterances do not belong to B by definition since they have systems of their own.

If interlanguages are indeed languages, the logical conclusion, as Corder (1978) pointed out, is that there are as many languages as there are learner interlanguages, albeit each is spoken minimally by one person for a limited or extended period of time. Potentially the number of interlanguages vastly exceeds the number of actual languages: if all the non-native learners of English have interlanguages of their own, interlanguages outnumber the English language itself by 400 million to 1 (Crystal, 1987).

First language acquisition avoids this issue partly by recognising that the learners are all moving towards a single target - native competence - partly by looking at features common to children's interim states. For example the thrust of Universal Grammar theory in recent years has been to discover the limits on human language. The child's choice is limited by the language faculty of the human mind; the reason that a child can handle language acquisition so efficiently is that there is only a finite number of languages. Interlanguages are possible languages known by the human mind; hence their properties represent Universal Grammar as much as those of conventional languages. Clearly far more sweeping statements can be made about the universals of human language from such a database than from the mere 4000 or so languages that currently exist in the world (Crystal, 1987). It does not matter if the number of interlanguages is indefinite; what we are concerned with is the way in which any interlanguage reflects the human language faculty.

The figures comparing interlanguages to languages above seemed bizarre because they were not of course counting the same thing. Chomsky has often insisted that the social phenomenon of 'language' must be distinguished from the knowledge of language stored in an individual mind, that is, 'grammar'. Individuals do not know languages: they know grammars. I do not have the English language stored in my mind as that is a sociological phenomenon shared by several speakers - langue in the de Saussure sense; I have a grammar of my own that is essentially unique even though it shares much with other speakers of 'English'. Language according to Chomsky is a derived notion - an epiphenomenon: 'The grammar in a person's mind/brain is real ... The language (whatever that may be) is not' (Chomsky, 1982, p.5). The grammar in an individual mind has instantiated UG in a particular way; if this corresponds to the permanent knowledge of a number of adult speakers, this is called a "language", if not an "interlanguage". This clearly fits the independent grammars assumption; L2 learners have grammars of their own which are not languages but have properties of their own. The L2 learner's knowledge is the same as that of a native speaker to the extent that both have a grammatical representation in their minds, for one of which there is an established name as a language such as 'English', the other lacking a name. There is no necessity that a speaker should actually be speaking something that corresponds to one of the world's languages such as French or English, only that the grammar conforms to the possible schemata laid down in Universal Grammar. The term 'interlanguage' has been misleading; the object that SLA research is actually about is intergrammar - the language system stored in an individual mind - not language in the sense either of a collection of sentences or of a social possession.

If the L2 learner has an intergrammar, how can it be established? First language acquisition research has relied heavily on the poverty-of-the-stimulus argument, which establishes the built-in features of the learner's mind from the aspects of grammar that could not have been learnt from the types of evidence available. Cook (1991) distinguished the 'broad' argument that generalises to all L1 children from the 'narrow' argument concerning the acquisition of language by an individual child. The broad argument is difficult to apply to Second Language Acquisition because there is no fixed 'language' steady state, Ss, as the final target. In addition, while generalisations about the importance of positive rather than negative evidence can be readily made for L1 children, the evidence available to all L2 learners is far from uniform and may include forms of evidence such as grammatical correction or grammatical explanation that are virtually unknown in L1 acquisition. Even if the assumption, attacked above, is made that the only possible target is balanced bilingualism (alias monolingual competence in two languages) the force of the broad poverty-of-the-stimulus argument is diminished when it depends upon the small and untypical group of learners that achieve this rather than the vast majority, who do not. To quote Cook (1991), 'the broad argument is difficult to use because of the variability of the final state St and the variety of evidence that the classroom can potentially provide'.

The narrow argument concerning the individual mind can be applied to L2 intergrammars as the possession of individuals as easily as to L1 native competence; it is irrelevant whether the grammar acquired actually belongs to a 'language'. The narrow argument still applies in that, if an individual knows an L2 fact that is unlearnable, it must be built-in. So the UG ideas can be tested as readily through Second Language Acquisition as through first language acquisition.

There is therefore no need for SLA research to bemoan the lack of a final state, Ss. In first language development, the independent grammars assumption accepted the child's state at any stage as being a competence of its own; a five-year-old speaks perfectly adequate five-year-oldese, not defective adult speech. The narrow poverty-of-the-stimulus argument can be applied equally to any stage of acquisition; what does the five-year-old, the six-year-old, the child at any age, know that he or she could not have learnt from input at that point in time? It is still an instantaneous logical model of acquisition because learning is not equated with development over time; all that has changed is that the grammar is that of an L1 speaker of another age rather than of an adult. The narrow poverty-of-the-stimulus argument applies, not to a single final competence known by the speaker, but to any grammar known by any speaker at any stage of acquisition: how did this state of knowledge come into being? The same questions of competence and acquisition are relevant whatever the stage of acquisition in L1 or L2; it is an arbitrary limitation to settle on the final stage of native competence as the only one to which the narrow poverty-of-the-stimulus argument applies. Any L2 learner's interlanguage can be tested to see if it conforms to the principles and parameters of UG and whether it is derivable from the input available to that individual. Second language acquisition differs from first language acquisition in principle only in that the output grammar is never a final stage of native competence and rarely approaches equivalence to native competence. Klein and Perdue (1992, p.1) have developed a similar point, though not in an Universal Grammar context: 'The question to be answered is whether a learner variety is based on recognisable organisational principles, how these principles interact, and whether they also apply to fully-fledged languages'.

The apparently irreconcilable gulf between the single L2 state of knowledge implied by SLA research and the manifold flexibility of the bilingualism definition can thus be bridged. What is needed is the straightforward concept of competence applied to the speaker's knowledge of language at a given moment; it is irrelevant whether this moment happens to be final native competence or some other stage; it is what the person knows there and then that matters. The poverty-of-the-stimulus argument, and the rest of the panoply of Universal Grammar theory, can be applied to this state with as much validity as to final steady-state competence. Native competence is beside the point; it is the current state of the mind that matters. It is valid to try to describe the L2 learner's language knowledge at a particular moment of time in terms of UG as an over-riding system that applies to any language at all known by a human mind; it is not valid to describe the learner's language as deviation from the monolingual competence from which it is distinct. The language of the balanced bilingual has no particular status.

But we are still missing a vital ingredient; it is not just that the L2 user's grammar is not a final state but that the knowledge of language of the L2 learner differs from that of a monolingual because it contains two grammars rather than one. The difficulty with the interlanguage concept in SLA research was that the learner's L2 was treated in isolation from his or her L1. SLA research had need of a concept that would embrace the learner's L1 and L2 simultaneously; 'The very fact that bilinguals use various languages in different circumstances suggests that it is their overall linguistic competence that should be compared to that of monolinguals' (Appel and Muysken, 1987, p.3). The L1 in the L2 user's mind was assumed to be of no interest, apart from whatever effects it might have on the L2 knowledge through transfer or interference. Bilingualism studies insists on treating the L2 along with the L1 as part of the whole competence of the speaker. It is the person's total use of language, whether L1 or L2, that is of interest. However, because bilingualism studies do not use the notion of individual mental competence, they have no occasion for a concept for this dual state.

The arguments skated over another vital difference when claiming that the object of study is the individual's knowledge of language at a moment in time. True as this may be, it nevertheless ignores the vital point that an L2 user has, not one grammar, but two. Should we look only at the L2 system at a single moment, avoiding the bilingualism position? Is it valid to study the L2 side of a dual L1+L2 state as if the L1 did not exist? Again the 'limited' L2 grammar would inevitably be compared to its disadvantage with the 'complete' monolingual grammar. The point of L2 acquisition is not to compare three-legged dogs with those that have four. Excluding the L2 user's L1 from the state of knowledge amputates a vital part of the organism and reduces second language acquisition to a shadow of first language acquisition.

Let us therefore introduce a neutral term to fill this gap - "multi-competence" Cook (1991), namely the state of language knowledge of a mind with two grammars. Linguistic competence refers to the L1 native competence; interlanguage to the L2 competence; multi-competence to the total state of language knowledge of a person who uses an L2, including both the L1 competence and the L2 interlanguage. Unlike linguistic competence, multi-competence has no final state; it is whatever it is at the particular moment of observation. Unlike interlanguage, multi-competence is a state of knowledge of two grammars, not just of the L2. Thus this combines both the independent grammars assumption and the lack of a final state. Multi-competence meets the need for the unit recognised by Labov (1970, p.21); 'Research in stable bilingual communities indicates that one natural unit of study may be the linguistic repertoire of each speaker rather than individual languages'. The term 'multi-competence' does not in itself say anything of the nature of this knowledge of languages. For example it is a separate question to be settled by empirical research whether there is a single combined 'wholistic' L1 and L2 system rather than one, as discussed in Cook (1992).

Let us return to the question of what linguistics is about. It is argued in Cook (1993) that Chomsky's three main questions for linguistics - knowledge of language, acquisition of language, and use of language (Chomsky, 1986) - need to be reformulated for SL research in terms of knowledge of languages, i.e. multi-competence. Substituting multi-competence into Chomsky's questions changes their meaning in several ways.

The goal of describing multi-competence involves, not just describing the L1 and the L2 as if they were independent entities in two different people but taking the point of departure as their coexistence in the same mind. Romaine (1989, p.284) points out 'it has not been appreciated how one might take bilingualism as the starting point and subsume monolingualism within it'. What needs to be demonstrated is, not that the two grammars are together, but how they are kept separate, since at some level they form part of the language knowledge in the same mind; 'the problem is one of accounting for the separateness or discreteness of language elsewhere' than in multilingual communities (Romaine, 1989, p.286); Lambert (1990) similarly talks of the bilingual being able to 'gate out' or set aside a whole integrated linguistic system while functioning with a second one, and a moment later, if the situation calls for it, switch the process, activating the previous inactive system and setting aside the previous active one'.

Formulated in UG terms, the central problem of SLA is how the mind manages to hold two grammars simultaneously. Looking at only the L2 grammar might indeed suggest all kinds of limitations; looking at only the L1 side would yield something approximately equivalent to L1 knowledge. Looking at the two together produces something of far greater complexity than the description of L1 linguistic competence and changes our concept of what human knowledge of language can consist of. If language acquisition is the instantiation of UG principles and the setting of values for UG parameters to obtain a grammar of a language, this has happened twice in the same mind. Rather than being fixed in the form of one language, the principles must apply to two; rather than having a single value, each parameter possibly has two at the same time. The description of knowledge of languages has to be flexible enough to accommodate two grammars, or, to be more accurate, any number of grammars ranging from one up to the indefinite number a person could acquire. The single monolingual grammar is not so much the norm as the minimum possible; it should be no more taken as typical of language acquisition than the juggler with only one ball. As Stenson (1990, p.194) puts it, 'Any grammatical theory that purports to account for human linguistic competence must also be able to account for bilingual competence and the associated performance.'

Again, using a concept of multi-competence alters the perspective on language acquisition. There are not two separate processes of first language acquisition and second language acquisition but a single process of language acquisition that results in the gain of this two-fold knowledge. It is not a matter of taking first language acquisition as the norm and then seeing how second language acquisition deviates from it, but of accommodating both first language acquisition and second language acquisition within the same learning process, at some level of abstraction.

As we have seen, the main goal of bilingualism research has been to see how people use more than one language. The point within this framework is not to take the line of Romaine (1989) and others that study should not be competence-based but to insist, as with monolingual linguistic competence, that use of language depends upon underlying competence, here competence in more than one language. Rather than abandoning bilingualism research as an area that cannot be reclaimed from performance, the concept of a flexible non-terminating multi-competence can underpin the study of the use of two or more languages.

The concept of multi-competence yields a distinct object for SLA research that keeps the study as a valid discipline of its own without subordinating it to other disciplines and without making it a simple catalogue of failure. The concept yields a fruitful series of questions that can now start to be answered, particularly the paramount question of how one mind can contain two grammars simultaneously.

Appel, R., & Muysken, P. (1987), Language Contact and Bilingualism, Edward Arnold, London

Birdsong, D. (1992), 'Ultimate attainment in Second Language Acquisition', Language, 68, 4, 706-753

Bley-Vroman, R.W. (1983), 'The comparative fallacy in interlanguage studies: the case of systematicity', Language Learning, 33, 1-17

Bley-Vroman, R.W. (1989), 'The logical problem of second language learning', in Gass, S., & Schachter, J. (eds.), Linguistic Perspectives on Second Language Acquisition, CUP, 41-68

Bley-Vroman, R.W., Felix, S., and Ioup, G.L. (1988), 'The accessibility of Universal Grammar in adult language learning', Second Language Research, 4, 1, 1-32

Bloomfield, L. (1933), Language, Holt, New York

Braine, M. (1963), 'The ontogeny of English phrase structure: the first phase', Language, 39, 1-13

Brown, R. (1973), A First Language: The Early Stages, Allen & Unwin, London

Chomsky, N. (1964), Current Issues in Linguistic Theory, Mouton, The Hague

Chomsky, N. (1965a), Aspects of the Theory of Syntax, MIT Press

Chomsky, N. (1965b), 'Formal discussion: the development of grammar in child language', in Bellugi, U., and Brown, R. (eds.), The Acquisition of Language, Purdue University, Indiana

Chomsky, N. (1976), Reflections on Language, Temple Smith

Chomsky, N. (1982), Some Concepts and Consequences of the Theory of Government and Binding, MIT Press

Chomsky, N. (1986), Knowledge of Language: Its Nature, Origin and Use, New York, Praeger

Cook, V.J. (1973), 'The comparison of language development in native children and foreign adults,' IRAL, 11, 1, 13-28

Cook, V.J. (1985), 'Chomsky's Universal Grammar and second language learning', Applied Linguistics, 6, 1-18

Cook, V.J. (1991), 'The poverty-of-the-stimulus argument and multi-competence', Second Language Research, 7, 2, 103-117

Cook, V.J. (1992), 'Evidence for multi-competence', Language Learning, 42, 4, 557-591

Cook, V.J. (1993), Linguistics and Second Language Acquisition, Basingstoke, Macmillan

Corder, S.P. (1971), 'Idiosyncratic errors and Error Analysis', IRAL, 9, 2, 147-159. Reprinted in Richards (1974)

Corder, S.P. (1978), 'Language-learner language', in Richards, J.C. (ed.) Understanding Second and Foreign Language Learning: Issues and Approaches, Newbury House, Rowley, Mass., 71-93

Davies, A., Criper, C., & Howatt, A. (eds.) (1984), Interlanguage, Edinburgh University Press

de Saussure, F. (trans. 1959), Course in General Linguistics, Peter Owen, London

Dulay, H.C., & Burt, M.K. (1973), 'Should we teach children syntax?', Language Learning, 23/2, 245-258

Flynn, S. (1987), A Parameter-setting Model of L2 Acquisition, Reidel, Dordrecht

Gass, S. (1979), 'Language transfer and universal grammatical relations', Language Learning, 29, 327-344

Halliday, M.A.K., McIntosh, A., and Strevens, P. (1964), The Linguistic Sciences and Language Teaching, Longman, London

Haugen, E. (1953), The Norwegian Language in America, University of Pennsylvania, Philadelphia

Hoffman, C. (1991), An Introduction to Bilingualism, Longman, London

Johnson, J.S., & Newport, E.L. (1991), 'Critical period effects on universal properties of languages: the status of subjacency in the acquisition of a second language', Cognition, 39, 215-258

Klein, W. (1986), Second Language Acquisition, CUP

Klein, W., & Perdue, C. (1992), Utterance Structure: Developing Grammars Again, John Benjamins, Amsterdam

Klima, E.S., & Bellugi, U. (1966), 'Syntactic regularities in the speech of children', in Lyons, J., & Wales, R.J. (eds.), Psycholinguistics Papers, Edinburgh U.P.

Labov, W. (1970), The Study of Non-Standard English, National Council of Teachers of English, Champaign, Illinois

Lambert, W.E. (1990), 'Persistent issues in bilingualism', in Harley, B., Allen, P., Cummins, J., & Swain, M., (eds.) (1990), The Development of Second Language Proficiency, CUP

Larsen-Freeman, D., & Long, M.H. (1991), An Introduction to Second Language Acquisition Research, Longman, Harlow

Mackey, W.F. (1970), 'The description of bilingualism', in J. Fishman (ed.), Readings in the Sociology of Language, The Hague, Mouton, 554-584

Meisel, J.M., Clahsen, H., & Pienemann, M. (1981), 'On determining developmental stages in natural second language acquisition', Studies in Second Language Acquisition, 3, 2, 109-135

Nemser, W. (1971), 'Approximative systems of foreign language learners', International Review of Applied Linguistics, 9, 115-123. Reprinted in Richards (1974).

Phillipson, R. (1992), Linguistic Imperialism, OUP

Richards, J.C. (ed.) (1974), Error Analysis, Longman

Romaine, S. (1989), Bilingualism, Blackwell, Oxford

Schumann, J. (1978), The Pidginisation Process: A Model for Second Language Acquisition, Rowley MA, Newbury House

Selinker, L. (1972), 'Interlanguage', IRAL, 10/3. Reprinted in Richards (1974)

Selinker, L. (1992), Rediscovering Interlanguage, Longman, London

Skutnabb-Kangas, T. (1981), Bilingualism or not: The Education of Minorities, Multilingual Matters, Cleveland

Spolsky, B. (1989), Conditions for Second Language Learning, OUP

Sridhar, K.K., & Sridhar, S.N. (1986), 'Bridging the paradigm gap; second language acquisition theory and indigenised varieties of English', World Englishes, 5, 1, 3-14

Stenson, N. (1990), 'Phrase structure congruence, government, and Irish-English code-switching', in Hendrick, R. (ed.), Syntax and Semantics Vol 23 The Syntax of the Modern Celtic Languages', San Diego, Academic, 169-199

Towell, R., & Hawkins, R. (1994), Approaches to Second Language Acquisition, Multilingual Matters

Weinreich, U. (1953), Languages in Contact, The Hague, Mouton

White, L. (1985), 'The acquisition of parameterised grammar: subjacency in second language acquisition', Second Language Research, 1, 1, 1-17

White, L. (1986), 'Implications of parametric variation for adult second language acquisition: an investigation of the pro-drop parameter', in Cook, V.J. (ed.) (1986), Experimental Approaches to Second Language Acquisition, Oxford, Pergamon

White, L. (1989), Universal Grammar and Second Language Acquisition, John Benjamins