The term 'multi-competence' was introduced to describe 'the compound state of a mind with two grammars', contrasted with 'mono-competence', the state of the mind with only one (Cook, 1991).  While the underlying concept has often been taken for granted in bilingualism research, it is by no means widely accepted in Second Language Acquisition research, which tends, as we shall see, to treat L2 competence as if the second language is in a different mind from the first.  A series of papers has looked at the use of multi-competence in different contexts (Cook, 1992; to appear).  The present paper relates multi-competence to the issue of age, mostly within the Universal Grammar theory of Second Language Acquisition.  This means first of all establishing the nature of a multi-competence SLA programme, then seeing how it might accommodate age.

The effects of age on second language learning are far from as clear-cut as popular wisdom, or indeed some linguists, would have us believe.  In their respective reviews of age, Krashen, Long and Scarcella (1979), Cook (1986), and Singleton (1989) come to general, if tentative, solutions, expressed by David Singleton as follows: 'The one interpretation of the evidence which does not appear to run into contradictory data is that in naturalistic situations those whose exposure to a second language begins in childhood in general eventually surpass those whose exposure begins in adulthood, even though the latter usually show some initial advantage over the former.'  Cook (1986) concentrated on the methodological limitations of the research literature, looking particularly askance at the over-reliance on educated immigrants as subjects, on English as the second language, on Indo-European languages as the first language, on the USA as the main country of immigration, on the limited aspects of language tested in the research, and on the learners' date of arrival rather than on their age of start.  Within the overall framework of Second Language Acquisition research the benefits and disadvantages of age still seem far from established.  Certainly it is hard to find an overall Critical Period effect with a particular cut-off age.  In addition it should not be forgotten that age is a factor from birth to death.  The relationship of age to Second Language Acquisition is not just the difference between children and adults but might also involve Second Language Acquisition in old age, for instance Clyne (1977).

Multi-competence and 'failure'

Any cursory reading of the first language acquisition literature comes up with a picture of excitement and wonder at the child's amazing ability to acquire language.  A similar reading of the second language literature soon produces the opposite impression; most researchers seem perplexed and concerned by how bad people are at learning second languages.  All children magically acquire their first language to a high level of knowledge after a few years.  Many second language learners achieve only a minimal L2 competence after long years of struggle and effort.  A main goal for Second Language Acquisition research is seen as explaining this dismal failure.  One historical basis for the concept of interlanguage was Selinker's observation that 95% of L2 learners fail to achieve 'absolute success in a second language' (Selinker, 1972).  The general negative view can be exemplified more recently in Bley-Vroman (1989, p.43) for instance, who discusses 'nine fundamental characteristics of adult foreign language learning' of which 'few are controversial'; the first two are 'lack of success' and 'general failure'.  This 'failure' is used by Schachter (1988) and Bley-Vroman (1989) in particular to argue that Second Language Acquisition differs from first language acquisition in being cut off from Universal Grammar and has to find a substitute channel for acquiring language.  One argument is that knowledge of the L2 is not so complete or so good as knowledge of the L1; Schachter (1988, p.224) claims 'Most proficient ESL speakers do not have fully formed determiner systems, aspectual systems, or tag question systems'.  L2 learners rarely attain the same level of knowledge in their L2 as in their L1.  This is taken to indicate lack of access to UG.  Furthermore, L2 learners regularly get 'fossilised' at some stage rather than progressing inevitably to full native competence.  Schachter's example is the over-use of the present indicative tense by English learners of Spanish.  As no L1 children get stuck in this way, UG is not available for learning the L2.

This no-access argument has been related to the learner's age in experiments by Johnson and Newport (1991).  Their main testing area is the subjacency relation seen in ungrammatical sentences such as:
   *What did Sally watch how Mrs Gomez makes?
versus grammatical sentences such as:
   What did the teacher know that Janet liked?
Subjacency is a highly technical area of principles and parameters theory that was fashionable during the 1980s but became largely subsumed under other areas such as the Empty Category Principle (Cook and Newson, 1995) as the theory developed.  Syntactic movement in human languages is limited by the constraint that the element that is being moved must not cross more than one of certain types of boundaries called 'bounding nodes'.  Sentences like:
   *What did Sally watch how Mrs Gomez makes?
are ruled out because what crosses more than one bounding node.  Johnson and Newport tested Chinese learners of English who would not have needed subjacency for syntactic movement in Chinese.  The results showed that the ability to detect the ungrammaticality of such sentences varies straightforwardly with age: the younger the learners had arrived in the US the better they were at subjacency, ranging from those who had arrived aged 4-7 up to those who had arrived as adults (i.e. not age of start but age of arrival, when they might or mught not have encountered English previously).  One of the causes of L2 failure, even within the UG model, seems to be the age of the learner.

What is the source of this aura of failure that envelopes Second Language Acquisition research?  A person learning to skate does not normally consider themselves a failure for not competing in the Olympic Games; the child who is learning physics at school is not considered a failure for not mastering quantum theory; a child learning a first language is not thought of as a defective adult native speaker but as a child in his or her own right.  In all these other cases the learner's progress is not measured against an external target of perfection but set against standards appropriate to the learner's own nature.  Yet Second Language Acquisition research has somehow accepted that the valid target for the L2 learner is the monolingual.  Most research compares L2 learners implicitly with L1 speakers; the typical research methodology uses grammaticality judgments, obligatory contexts, or elicited imitation, all defined by native usage.  The target the learners are supposedly aiming at, even if few of them achieve, is the competence of the monolingual, not the competence peculiar to L2 speakers.  Failure is defined with reference to monolingual native speakers - the one thing that L2 learners can never be by definition.  The goal of Second Language Acquisition is Selinker's successful 5% - the 'balanced' bilingual or 'ambilingual', as Halliday, McIntosh and Strevens (1964) neatly call it.  Yet such ambilinguals form a small minority of those who have learnt an L2.  Taking them as the paradigm example automatically leads to negative conclusions about L2 learning.  Measured against the yardstick of 100% monolingual competence, the L2 knowledge of a learner ranges from 0% to 100%.  As learners seldom approach the upper limit, L2 learning has been seen in terms of lack of success; few bilinguals succeed in achieving or maintaining an equal balance between L1 and L2 knowledge.

But suppose the L2 learner is acquiring something extra rather than a pale imitation of the L1 - the vocabulary, syntax, and so on, of the L1 plus the vocabulary, syntax, and so on, of the L2.  The knowledge of the L2 increases the L2 user's capabilities beyond those of a monolingual, rather than being confining them to defective L1 knowledge.  Measured against the 100% of a person who knows one language, the balanced bilingual is functioning at 200%; L2 learners of lesser achievement are functioning at levels between 100% and 200%.  Those who start from the balanced bilingual see the L2 learner as a failure for not achieving full L2 competence; even bilinguals, according to Grosjean (1989), 'often assume and amplify the monolingual view and hence criticise their own language competence'.  Those who start from the bilingual view see the learners as a success in their own right in going beyond the initial L1, to whatever degree.

In many ways the Second Language Acquisition literature could be considered discriminatory.  We (and I include myself in these strictures as much as anybody) have been looking down at the pitiful attempts of the L2 user to attain the lofty heights of the monolingual.  Perhaps L2 users should have been looking down from their lofty heights at the impoverished lives of monolinguals straitjacketed by one language, one culture, one way of thinking.  Whatever the L2 user has achieved exceeds the capabilities of monolinguals, rather than falling short.  The pejorative terminology applied to L2 learners resembles a power play by natives who want to keep control of their language, and so deny the rights of non-natives, parallel to the attempts to maintain the position of English across the world at the expense of the Third World (Phillipson, 1992).  In other research areas linguists have learnt to speak in terms of differences between people rather than deficits - race, sex, class, and so on; we should try to recognise the difference of the L2 user as a unique human being rather than as a defective monolingual, just as blacks are not defective whites, women are not defective men, and academics are not defective navvies.

The discussion of age in L2 learning is indeed dominated by failure or success in native terms.  One example is the question of accent.  Asher and Garcia (1969) had native speakers judge 'accent'; Seliger et al (1975) asked learners to report how long they took to lose their 'foreign accent'; Oyama (1975) measured assessment of learners' foreign accent by linguistic students.  One might point to the looseness and subjectiveness of the concept of accent as compared, say, to a proper phoneme and allophone inventory of the learner's speech.  But what could betray more blatantly that monolingual standards are being applied than using native accent as the yardstick?  Why are L2 users considered deficient if they do not sound like natives?  Are women condemned for not sounding like men?  Why should the ability to have this nebulous accent be so crucial?  For example a test of the teachability of pronunciation showed that 18 hours of pronunciation teaching could make 9 out of 20 Canadians indistinguishable from native speakers of Japanese (Neufeld, 1978).  What's the point?  They are not Japanese and will never for a moment be able to pass for Japanese; they should sound like effective bilinguals. 

The early work on interlanguage drew on the 1960s independent grammars assumption (Cook, 1993) - the belief that the child should be treated as a speaker of a language of its own rather than as a defective speaker of adult language who has inefficiently mastered the rules, perhaps first articulated in McNeill (1966).  Taking this seriously means discovering the learner's own grammar, not measuring how successfully he or she approximates the monolingual grammar, thus avoiding what Bley-Vroman (1983) calls the Comparative Fallacy.  Methodologically the criteria in the age research always seem to be derived from native grammars.  Johnson and Newport (1990) for instance asked whether L2 learners come up to scratch compared to native speakers.  Tut tut they don't.  But might the learners have an alternative grammar rather than a defective grammar?  Research based on the monolingual grammar is confined to making conclusions about similarities and differences; it can never reveal a system that is totally different from the native and show the true nature of multi-competent knowledge unfiltered by the native grammar.

The foundation of Second Language Acquisition research has to be multi-competence - the complex supersystem of language knowledge possessed by those who use more than one language.  The effects of age should be ascertained by looking at the language of L2 learners in their own right, not by seeing how they fall short of natives.  However much the interlanguage researchers protested that they treated the learner's system as independent their very research techniques sooner or later involved the native target.  The question about the relationship of age to Second Language Acquisition should not then be couched in terms of failure; young learners may or may not acquire different knowledge of the L2 than older learners; this is an empirical fact that current research has still not established properly.  What is crucial is to discover their own grammars, however different these may be from natives'.

Age and multi-competence

An entry point to the discussion of age and multi-competence is the development of growth and non-growth models of UG.  The growth version of UG associated with Borer and Wexler (1987) claims that the principles and parameters of UG do not all necessarily come into being at the same time; a biologically-predetermined programme allows them to appear gradually in the child's mind.  The non-growth version suggests that all the principles and parameters are present from the outset; they cannot be seen in the child's speech because of the various processing systems for language the child has not yet evolved, say the short term memory capacity necessary to produce longer sentences.  In terms of L1 acquisition the choice between these alternatives is impossibly hard to decide; how can you tell an invisible principle that is present in the child's mind but never used because of the other restrictions on the child from a principle that is altogether absent from the child's mind, if you do not have a full psychological account of the development of language processing, since in both cases something will be missing from the child's speech? 

A further problem is what is meant by age.  Chronological age is a meaningless variable in itself; like time, we can only perceive it in terms of change: no change, no age.  Hence age might mean change in the physiology of the brain areas, muscles, or auditory equipment involved in speech; older learners may be physically different from younger learners.  Or age might mean mental change - where the person is on the Piagetan sequence of cognitive development, how their short-term memory capacity or structures have developed, and so on.  Or age might be social development - the forms of interaction typical at particular ages.  All of these have indeed been put forward to explain the alleged effects of age and the age changes are in some way verifiable independently of the second language evidence.  But none of them are directly relevant to the UG model of acquisition, since this does not take account of language development, only of the logical problem of language acquisition, and does not see the language faculty as depending in any way on other mental faculties.  Blaming one or other of these for age effects is talking in terms of another model.

We need then to look for other solutions.  Why do the grammars of L2 users apparently vary according to the age at which they started L2 learning, judging from Johnson and Newport's work with subjacency?  One possibility is indeed that UG is not available to the older L2 learners.  But much of the L2 research, such as Bley-Vroman et al (1988), shows that UG is not completely unavailable; scores are always better than chance even if consistently below natives'.  Older L2 learners have poor access to UG rather than no access.  Another possible solution is a variation on the growth model.  Children acquire the L1 in sequence; perhaps subjacency has to come at a particular time in the sequence to be usable, like Konrad Lorenz imprinting his ducks within the critical period.  L2 learners of different ages are at different points in this developmental unfolding of principles and parameters.  Getting subjacency in the L2 out of step with the maturational sequence may be a handicap and stop it being used as efficiently. 

The other possibility is, however, to invoke the nature of multi-competence.  Why should multi-competent speakers of two languages have the same grammar of either language?  My own research has suggested that multi-competence should be treated as a single system; the L1 and L2 systems are symbiotic within the same mind, as is clearly evident for example in the complex phenomenon of code-switching.  The effects of a UG principle are not therefore necessarily the same for this compound multi-competent state as for a monolingual state.  Again the L2 user's grammar must be treated as having its own nature, not as deviant from the monolingual.  Some aspects of UG might indeed only become apparent in people who know more than one language.  The question is then: do L2 learners who start at different ages acquire different multi-competences? 

Age and simultaneous bilingualism

One of the questions that immediately arises with multi-competence is whether the two languages form one overall system in the mind (wholistic multi-competence) or two (separatist multi-competence).  This is a derivative question the answer to which does not affect the notion of multi-competence itself since, at some level, the two languages form part of the same mind by definition, otherwise we would be dealing with people with two heads.  Cook (1992) reviewed evidence that minds with two language are in general different from those with only one - that multi-competence is a different state of mind from monolingualism - concluding that this is true to some extent of L1 and L2 knowledge, metalinguistic awareness, and cognitive processing. 

In several areas of psychology the problem of whether the L2 user has two systems or one might well be considered a non-question.  Anderson (1983) for example considers 'the adult human mind is a unitary construction'.  He refuses to distinguish language from other cognitive systems and, by extension, the first language from the second; they are all part of the same network.  Indeed he often uses classroom L2 learning as an example of his all-purpose theory of learning.  This is discussed further in Cook (1993).  Other network theories such as connectionism appear equally committed to the mind as a single organism, with L2 learning being a matter of increasing the strength of certain connections (Gasser, 1990; Sokolik & Smith, 1992).  More crucially for our current concerns it appears to be a standard axiom in bilingualism research that bilinguals are not two monolinguals in the same head: 'it is clear that a reasonable account of bilingualism cannot be based on a theory which assumes monolingual competence as its frame of reference' (Romaine, 1989).  The overall emphasis in bilingualism studies is on 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.  The 'ultimate attainment' of a second language is not, and could never be, monolingual competence.

It is mainly then within Second Language Acquisition research that the concept of multi-competence needs to be asserted.  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'.  The question is at what level or in what areas the mind separates the two languages into separate systems.  It might be that the core UG areas are inseparable in the two languages, hence the powerful influence of L1 parameter settings on the knowledge of the L2, or it might be that UG cannot be used for an L2 and so the two systems are organised in utterly different ways, as those who support alternatives to UG in L2 learning suggest.  It might be that there is a single lexicon for both languages in which each word is branded with its language.  Cook (1992) examined the evidence for wholistic versus separatist multi-competence in the areas of the lexicon, code-switching, language processing, physical storage of language in the brain, and the relationship between L1 and L2 proficiency, but not syntax, concluding that 'the two systems are more closely linked than had been previously suspected'.

But these types of evidence are largely concerned with adults who have reached their final state of L2 learning.  The division between separatist and wholistic multi-competence may however depend on age, defined in terms both of the mental state at the time at which the people were tested and at the time at which the L2 acquisition started.  Let us take the extreme test-case of the simultaneous acquisition of two languages in young children.  Again, as with bilingualism, one should not perhaps be too pernickety about defining simultaneous acquisition.  The most extreme form is perhaps de Houwer (1990) for instance who records the languages used to her subject in the maternity ward ('English spoken to her by her mother, and Dutch by her father and various members of the nursing staff').  Like the usual arguments against the idealisation to linguistic competence, the balanced bilingual and the child identically exposed to two languages may be fictions that obscure the normal situation.  If Second Language Acquisition is restricted to those who achieved balanced bilingualism, it excludes all but a minute proportion of L2 learners; if we restrict simultaneous acquisition to those children who have equal amounts of both languages from maternity ward onwards, these would be so rare as to be virtually non-existent.  This does not mean of course that such cases are not interesting as the 'purest' evidence and hence the easiest to research.

The questions that researchers have usually asked about simultaneous bilingualism are whether the child's two languages develop separately - sometimes called the independent development hypothesis - and whether bilingual acquisition follows the same path as monolingual acquisition.  The evidence that is cited is extremely mixed and goes back many years, usually at least to Ronjat (1913) and Leopold (1939-49); reviews can be found in McLaughlin (1984) and de Houwer (1990).  McLaughlin's review reports, mostly in terms of language processing and memory, that 'research with bilingual children does not support the notion of a dual language system' (McLaughlin, 1984, p.195).  On the other hand de Houwer (1990) finds largely the same literature much more ambivalent when restricted to genuinely simultaneous acquisition; 'there is as little positive evidence for the position that bilingual children develop two separate linguistic systems from the earliest stages of acquisition on as there is for the claim that bilingual children start out with a single linguistic system which is later differentiated or separated into two linguistic systems' (de Houwer, 1990, p.49).

Let us then review briefly the evidence in some of the relevant areas of research with young bilingual children.  While there are many general studies, those directed at any specific area of language seem small in number and far from satisfactory.

Vocabulary

The question that has been most researched is whether children mix the vocabularies of the two languages in the same sentence.  Ronjat (1913) originally claimed there was no lexical confusion between languages.  Burling (1959), however, reported a boy who first used Garo sentence structure with English words, later the reverse.  Leopold (1939) found that 'Words from the two languages did not belong to two speech systems but to one ...'.  Taeschner (1983) studied 2 children learning Italian and German; she found an initial stage at which the children had one lexical system, which separated at the next stage into two lexicons with one syntax.  Tabouret-Keller (1962) found mixed early sentences with unequal proportions from the vocabulary of the two languages.  Swain and Wesche (1975) described lexical mixing between French and English with examples such as Quelle couleur your poupée?; de Houwer (1990) observed a Dutch child learning English and Dutch and found only 4.3% mixed sentences addressed to Dutch speakers, 3.9% to English, and 2.5% and 0.9% that were balanced between the two languages ("Dutlish", as she calls it).  While attacking Taeschner's methodology and analysis, de Houwer herself does not use vocabulary in her conclusions other than to point to the non-existence of an early stage at which the children have a single lexicon for both languages. 

Much of this research is notionally about vocabulary in that it simply categorises the words that occur in the children's speech into L1 or L2.  This provides little information on what the child means by the words it uses - are meanings constant across languages, is a word misused in one language because of its meaning in another language, and so on?  The research does not use linguistic ideas of vocabulary or modern ideas of vocabulary acquisition, say the acquisition of argument structure; it uses words as isolated atoms of language which either occur or don't occur.  It is not a question of two vocabulary systems so much as two word-lists.  This commonly-found separation between the two lexicons contrasts with the extensive research with adults that shows a single shared lexicon, for example Beauvillain and Grainger (1987).  It might be that the two concepts of vocabulary and the two approaches to research are incommensurable, or it might be that there is a genuine difference between children who keep the lexicons separate and adults who have integrated the two lexicons into one.  Without more specific research there is little to show whether the two lexicons, as opposed to the two word-lists, form one system or two.

Pronunciation

Most of the bilingualism research simply ignores pronunciation, for example de Houwer (1990).  Much of the relevant research furthermore does not use systematic ideas of phonology.  Leopold (1947) reports some confusion between the sounds of the two languages and some carryover of phonological processes from one language to the other, as does Ruke-Dravina (1967); Fantini (1985) and Burling (1959) describe children in whom the phonology of one language takes over.  Oksaar (1970), however, studied a child who kept the pronunciation of the two languages separate.  Essentially these case studies represent every possibility from totally merged systems to totally separate.  At best they can provide counter-instances to any absolute claim that all bilingual children necessarily have merged or separate pronunciation systems.  Much phonetic work has indeed shown that adult bilinguals effectively have a single phonological system for Voice Onset Time, for example Nathan (1987).

Syntax

Syntax is usually handled in the earlier literature of simultaneous bilingualism as an offshoot of the study of vocabulary and pronunciation rather than in its own right.  The question as always is: does the child keep the two systems separate?  Ronjat (1913) found few signs of syntactic interference between the two languages.  The initial stage of the child's word order in Leopold (1939) appeared to belong to neither language but to express semantic meaning.  The sentences produced by the child in Burling (1959) belonged clearly to one language or the other, even when the vocabulary was mixed.  Morphological borrowing from English to Estonian occurred in Vihman (1982).  Swain and Wesche (1975) found some French structures occurring in English, for example They open, the windows?  Volterra and Taeschner (1978) however described a stage when children had two lexicons but a single syntactic system that was neither Italian nor German; the child expressed possession in both languages as Noun + Noun Giulia Buch (Julia's book) or Giulia giamma (Julia's pyjamas).

De Houwer (1990) concentrated on the question of whether the two syntactic systems are separate, finding that the languages (Dutch and English) are kept distinct in gender; bound morphemes such as plural stay in one language; the child mostly uses OV word order in Dutch and VO in English.  At all stages the child keeps the languages separate.  However, apart from word order, as she points out, Dutch and English are fairly similar in syntax.  She also finds 'It is quite striking that the many changes in Kate's two language systems take place at the same time.'

The DUFDE group directed by Jürgen Meisel, reported for example in Meisel (1990a), have looked at young children learning French and German simultaneously.  Meisel (1990b) finds the evidence provided by Volterra and Taeschner (1978) unconvincing as a demonstration of a clear stage of a single syntactic system; the children had distinct word orders in the two languages, similar but not identical to monolingual children.  His conclusion is that 'fusion is not necessarily a characteristic of bilingual language development, but mixing may occur until code-switching is firmly established as a strategy of bilingual pragmatic competence'.

To sum up, the consensus seems to be that, after an initial semantically organised phase, children keep the systems of the two languages distinct, essentially the model presented in Volterra and Taeschner (1978) or Swain and Wesche (1975), except in so far as code-switching and mixing are a normal part of bilingual behaviour, both of the child's and the parents' (Genesee, 1989).  The greatest contrast is with the research into adult bilingualism where mostly at the vocabulary level, and sometimes at the phonological level, a single system seems to be involved.  But, as we have seen, the evidence is in fact shaky; a consistent account treating language rigorously in terms of vocabulary, pronunciation, and syntax does not exist.

So what for multi-competence?  At first sight the implication is that lack of mixing shows two systems; mixing shows one.  But keeping two languages separate does not mean that they do not form part of the same system.  The system I use for hand-writing is quite different from the one I use for word-processing; there is no way in which I mix them; yet noone would claim that I speak two languages simply because I have two different ways of realising written English.  The polylectalists pointed out that all speakers switch continuously from one system to another, else they would not be able to understand children or old age pensioners, Bristol shop assistants or High Court judges.  'Competence is polylectal' (Bailey, 1973).  Any language use involves switching from one style or register to another.  We do not therefore claim that monolingual speakers have many language systems in their minds and marvel how they keep them separate; instead we believe they have a single language system.  Lack of mixing in children no more shows that two systems are involved than does the fact I consistently speak to my cat differently than to the university vice-chancellor.  The research into simultaneous acquisition has usually answered the question whether children differentiate two languages, not the question whether the languages form one system.  Genesee claims for example 'To uphold the unitary system hypothesis one would need to establish that, all things being equal, bilingual children use items from both languages indiscriminately in all contexts' (Genesee, 1989, p.165); this same criterion would prove that monolinguals speak many languages.  Arbitrary mixing would indeed show no system at all rather than one system, which might well distinguish different forms or vocabulary for different uses, as it does in monolinguals.

Let us put this within the context of current UG theory.  The lexical parameterisation hypothesis claims that parameters are part of the lexicon; we do not know the values for the governing category principle in Binding Theory for English, Japanese or Icelandic; we know the setting of the lexical items himself, zibun, or hann (Wexler and Manzini, 1987).  The functional parameterisation hypothesis extends this by claiming that parameters are attached to the lexical entries for functional categories (Ouhalla, 1991).  To take the example of the Agreement Phrase, the English AGRP has set the opacity parameter to opaque and so forbids Verb movement, while the French AGRP has set the value to transparent and so allows Verb movement; hence the normality of Jean embrasse souvent Marie (John kisses often Marie) and the abnormality of *John kisses often Marie; the English verb kiss cannot cross over the boundary of the English AGRP (Pollock, 1989).

Both of these hypotheses take the view that syntax is invariant; languages differ in their lexicons: 'language acquisition is in essence a matter of determining lexical idiosyncrasies' (Chomsky, 1989, p.44).  Lexical parameterisation takes the view that the lexicon consists of lexical entries for substantive items such as nouns, verbs, reflexive anaphors, etc, giving their meaning, phonological form and syntactic properties.  Children learn these essentially one at a time; when they learn himself they acquire its setting for the governing category parameter.  The functional parameterisation view envisages an extended lexicon which includes entries for more abstract functional categories such as the Complement Phrase and Agreement Phrase with parameter settings.  When children acquire a functional phrase, they start acquiring the parameter settings that go with it.

Lexical parameterisation then fits neatly with the picture of early bilingualism we have been reporting.  Children separate the substantive vocabularies because each lexical entry is separate: English man and French homme are two items in one extended vocabulary, as they are for adult bilinguals.  As such the entries project their own syntactic properties on sentences in which the words occur.  It is no more surprising that children fail to mix vocabularies than that English adults fail to mix the words cat and chewing gum.  There is one multi-competence lexicon with separate entries for each item.

Functional parameterisation raises a further range of issues that require specific research to tackle.  There might be separate entries for functional categories in each language, each with a parameter value assigned to it.  In this case there is a single overall knowledge of language relying on principles and parameters with language separation maintained as different lexical entries.  There might be only one entry possible for each functional category with a parameter attached that has two strengths, switching from one to the other according to which language is involved.  The choice between these alternative can be tested through code-switching.  In the one case the choice is either/or; a functional phrase in code-switching should keep to one language in terms of its parametric values.  This could be interpreted as an extension of the government model of code-switching proposed by Di Sciullo et al (1986) in which governed elements stay in the same language as their governor or as a variant on the Matrix Language Frame Model of Myers-Scotton (1992) in which one language acts as the dominant matrix.  In the other case the choice is more/less.  Functional phrases mix elements within the phrase with a less clear demarcation.

References

Anderson, J.R. (1983a), The Architecture of Cognition, Harvard University Press

Asher, J., & Garcia, G. (1969), 'The optimal age to learn a foreign language', Modern Language Journal, 38, 334-341

Bailey, C-Y. (1973), Variation and Linguistic Theory, Center for Applied Linguistics, Arlington, Virginia

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, Cambridge: Cambridge University Press: 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.

Borer, H., & Wexler, K. (1987), 'The maturation of syntax', in Roeper, T., & Williams, E. (eds.), Parameter Setting, Dordrecht, Reidel, 123-187

Burling, R. (1959), 'Language development of a Garo and English speaking child', Word, 15, 45-68

Chomsky, N. (1989), 'Some notes on economy of derivation and representation', MIT Working Papers in Linguistics, 10, 43-74

Clyne, M.G. (1977), 'Bilingualism of the elderly', Talanya, 4, 45-56

Cook, V.J. (1986), 'Experimental approaches applied to two areas of second language learning research: age and listening-based teaching methods', in V.J. Cook (ed.), Experimental Approaches to Second Language Learning, Pergamon

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

Cook, V.J. (to appear), 'The metaphor of access to Universal Grammar', in Ellis, N. (ed.), Implicit Learning and Language, LEA

Cook, V.J., and Newson, M. (1994), Chomsky's Universal Grammar, second edition, Oxford, Blackwell

de Houwer, A. (1990), The Acquisition of Two Languages from Birth: a Case Study, Cambridge: Cambridge University Press

Di Sciullo, A-M., Muysken, P., & Singh, R. (1986), 'Government and code-mixing', Journal of Linguistics, 22, 1-24

Fantini, A. (1974), Language Acquisition of a Bilingual Child,  Battleboro: The Experiment Press

Gasser, M. (1990), 'Connectionism and universals of Second Language Acquisition', Studies in Second Language Acquisition, 12, 2, 179-200

Genesee, F. (1989), 'Early bilingual development: one language or two?', Journal of Child Language, 16, 161-179

Goodluck, H. (1986), 'Language acquisition and linguistic theory', in P. Fletcher & M. Garman (eds.), Language Acquisition: Studies in First Language Development, CUP, 49-68

Grosjean, F. (1989), 'Neurolinguists, beware!  The bilingual is not two monolinguals in one person', Brain and Language, 36, 3-15

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

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, 215-58

Krashen, S., Long, M., & Scarcella, R. (1979), 'Accounting for child-adult differences in second language rate and attainment', TESOL Quarterly, 13, 573-582

Leopold, W.F. (1939-1949), Speech Development of a Bilingual Child: A Linguist's Record, Volume I Vocabulary Growth in the First Two Years, Volume II Sound Learning in the First Two Years, Volume III Grammar and General Problem in the First Two Years, Volume IV Diary from Age Two, Evanston, Northwestern University Press

McLaughlin, B. (1984), Second Language Acquisition in Childhood: Volume I Pre-school Children, Erlbaum, New Jersey

McNeill, D. (1966), 'Developmental psycholinguistics' in Smith, F., & Miller, G. (eds.), The Genesis of Language, Cambridge, Mass, MIT

Meisel, J.M. (ed.) (1990a), Two First Languages: Early Grammatical Development in Bilingual Children, Dordrecht, Foris

Meisel, J.M. (1990b), 'Early differentiation of languages in bilingual children', in Hyltenstam, K., & Obler, L.K. (eds.), Bilingualism across the Lifespan, CUP, 13-40

Myers-Scotton, C. (1992), 'Constructing the frame in intrasentential code-switching', Multilingua, 11, 101-27

Nathan, G.S. (1987), 'On second-language acquisition of voiced stops', Journal of Phonetics, 15, 313-322

Neufeld, G. (1978), 'On the acquisition of prosodic and articulatory features in adult language learning', Canadian Modern Language Review, 34, 163-174

Oksaar, E. (1970), 'Zum Spacherwerb des Kindes in Zweisprachiger Umgebung', Folia Linguistica, 4, 330-358

Ouhalla, J. (1991), Functional Categories and Parametric Variation, Routledge, London

Oyama, S. (1976), 'A sensitive period for the acquisition of of a non-native phonological system', Journal of Psycholinguistic Research, 5, 261-285

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

Pollock, J-Y. (1989), 'Verb movement, universal grammar, and the structure of IP', Linguistic Inquiry, 20, 365-424

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

Ronjat, J. (1913), Le développement du langage observé chez un enfant bilingue, Paris, Champion

Ruke-Dravina, V. (1967), Mehrsprachigkeit im Vorschulalter, Gleerup, Lund

Schachter, J. (1988), 'Second language acquisition and its relationship to Universal Grammar', Applied Linguistics, 9, 3, 219-235

Seliger, H., Krashen, S., & Ladefoged, P. (1975), 'Maturational constraints in the acquisition of second languages', Language Sciences, 38, 20-22

Singleton, D. (1989), Language Acquisition: The Age Factor, Multilingual Matters, Clevedon

Sokolik, M.E., & Smith, M.E. (1992), 'Assignment of gender to French nouns in primary and secondary language: a connectionist model', Second Language Research, 8, 1, 39-58

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

Swain, M., & Wesche, M. (1975), 'Linguistic interaction: case story of a bilingual child', Language Sciences, 37, 17-22

Tabouret-Keller, A. (1962), 'L'acquisition du langage parlé chez un petit enfant en milieu bilingue', Problemes de Psycho-linguistiques, 8, 205-219

Taeschner, T. (1983), The Sun is Feminine, Springer, Berlin

Vihman, M.M. (1982), 'The acquisition of morphology by a bilingual child', Applied Psycholinguistics, 3, 141-160

Volterra, V., & Taeschner, T. (1978), 'The acquisition and development of language by bilingual children', Journal of Child Language, 5, 2, 311-326

Wexler, K., & Manzini, M.R. (1987), 'Parameters and learnability', in Roeper, T., & Williams, E. (eds.), Parameters and Linguistic Theory, Reidel, Dordrecht