SLA Topics  Vivian Cook Key Issues in SLA

The poverty of the stimulus argument and multi-competence

V.J. Cook, University of Essex
Paper given at AILA Congress Thessaloniki, appearing in Second Language Research, 1990, 7, 2, 101-117

[Some figs a bit off and subcripts for Ss an St are erratic. This was the first paper to use the term multicompetence, which, oddly, arose out of a talk on the poverty-of-the stimulus argument given at a symposium on input at AILA 1989. For later developments see the definition of multicompetence, the multicompetence website and the Handbook of Linguistic Multi-competence CUP 2016]

One of the bases for the Universal Grammar (UG) position is the poverty-of-the-stimulus argument: if someone knows an aspect of language that is unlearnable from the environment, then it must be built-in to the mind.  The aim of the present paper is to examine the consequences for this argument of the fact that most human minds actually know more than one language.  It first attempts to clarify the ramifications of the present form of the argument by distinguishing between 'narrow' and 'broad' versions.  It then argues from this foundation that the argument undervalues minds that know more than one language; it goes beyond the current UG position to suggest that linguistic theory needs to see 'multi-competence' as a vital and normal part of language knowledge.

The narrow poverty of the stimulus argument 

A preliminary point is that the poverty-of-the-stimulus argument throughout refers to the central 'core' aspects of grammar that are closely related to UG rather than to the 'peripheral' aspects that are less related; it does not have to account for anything outside this narrow area.  The normal form of the poverty-of-the-stimulus argument has four steps.  To carry out the first step, Step A, means demonstrating that a single native speaker knows a particular piece of linguistic knowledge. 
           STEP A. a native speaker knows X 

This version concerning a single individual will be called here the 'narrow' argument.  A standard example of such knowledge is the principle of structure-dependency (Cook, 1988; Chomsky, 1986; 1988).  Languages that have syntactic movement permit questions such as:
     Is Sam the cat that is black?
but forbid sentences such as:
     *Is Sam is the cat that black?
Movement within the sentence depends, not on the word as such, but on its position in the structure of the sentence.  Step A could be used for any syntactic point that a native speaker knows that is relevant to the 'core' grammar of the language.

The next step is to show that the language evidence available to the speaker could not have led to him or her acquiring the syntactic point in question:

   STEP B. X could not have been acquired from forms of evidence  plausibly available

Such evidence can be positive in the form of language the child actually encounters.  The child indeed hears sentences that are structure-dependent.  The problem is, however, if the child never hears structure-independent sentences, why does the adult immediately reject them the first time he or she hears them?  Positive evidence alone could never lead to this knowledge of what can't be said in language (see Cook, 1988 for amplification of this point).

The other type of evidence is negative - demonstration of what does not occur, either directly by adults correcting the child or indirectly by being absent from what the child hears.  However the child is not corrected for saying *"Is Sam is the cat that black?" simply because he or she never produces sentences that might cause an adult to say "No you shouldn't say *'Is Sam is the cat that black?' you should say 'Is Sam the cat that is black?'"; direct negative evidence does not occur for the appropriate structures. 

Step B involves showing that none of the types of positive and negative evidence are plausible so far as this particular aspect of linguistic competence is concerned; the child couldn't learn structure-dependency from either positive or negative evidence.  The step depends on an occurrence requirement for evidence (Cook, 1988); evidence must not just in principle be available to the child but it must actually occur.  While in principle negative evidence might compensate for the inadequacies of the positive evidence, in practice it is not observed to occur.  However useful correction might be in principle, parents rarely correct their child's syntax in relevant ways. 

This leads to the conclusion that none of these forms of evidence are sufficient:

     STEP C. X is not learnt 

The aspect of language knowledge in question is not learnt.  That is to say, structure dependency is not derivable from actually occurring evidence in the environment.  Though present in the mind, it did not come from language evidence from outside and hence cannot strictly be called 'learnt'.

Consequently this knowledge must be built-in to the human mind. 

   STEP D.  X is built-in to the mind

Steps C and D are distinct because logical explanations could exist for things which are known but not learnt other than the innate structure of the mind, for instance, memories of previous existence, morphic resonance, telepathy, or a literal interpretation of the Tower of Babel to take a selection ranging from Plato to Sheldrake.  We will assume however that the proper explanation for linguists to explore is innateness.

The central point to the poverty-of-the-stimulus-argument is that the environment underspecifies the kinds of grammar that people know; language knowledge is too complicated to be learnt.  In one way the argument resembles the theological proof of the existence of God from the complexity or beauty of the world.  In another it is like techniques of molecular phylogeny that reconstruct early species by working backwards from the protein structure of presentday creatures.  Used in another way the argument is held to explain how language learning is possible in a real time span.  Restricting the choices for the learner cuts down on the numbers of grammars that need to be tried: 'the system of UG is so designed that given appropriate evidence only a single candidate language is made available' (Chomsky, 1986, p.83).

The argument can be transferred more or less intact to L2 learning, with the caveat that sources of evidence may also include L1 knowledge.  If an L2 learner knows things which are not in the evidence encountered, the poverty-of-the-stimulus argument applies to L2 learning.  Indeed an L2 learner may know structure-dependency in circumstances when it is not required in the L1.  Otsu and Naoi, cited in White (1989), showed Japanese learners of English possessed structure-dependency though it is not required for syntactic movement in Japanese, even if needed elsewhere, e.g. for LF-movement].  If the syntactic point is not learnable in the first language, there is no reason why it magically becomes learnable in the second, unless there is a crucial difference in the types of evidence available to the learner.  It should be stressed that the argument in L2 is as much a logical argument as it is in the L1.  It does not depend upon data about learners acquiring an L2 any more than the L1 argument depends upon data from children learning an L1, even if research such as that by Crain & Nakayama (1983) in the L1, or by Otsu and Naoi in the L2 provides interesting confirmation.  It is the existence of something in the speaker's knowledge of the language that matters.  Developmental issues are irrelevant to the logical problem of language acquisition.  Indeed one of the major problems with the UG position is deciding what evidence other than the speaker's knowledge can be valid (Cook, 1991a).

The broad poverty-of-the-stimulus argument

Mostly however the basic poverty-of-the-stimulus argument is not presented solely in terms of individuals but is generalised to all human beings.  This will be called the 'broad' version of the argument.  The single native speaker referred to in Step A is tacitly assumed to represent all native speakers.  Though, psychologically speaking, language is learnt by individuals, all native speakers of a language possess more or less the same linguistic competence.  Thus Step A is perhaps more appropriately worded as:
           STEP A' native speakers know X

All native speakers know structure-dependency.  In principle a single person who did not might be sufficient to demolish the claim.  It is still ultimately the individual that matters, despite the broadening of the argument.  The argument never depends on statistics, whether based on large numbers of people or large numbers of languages.  Its topic is individual competence.  Nevertheless it is what human beings have in common that is important.

This has consequences for Step B "X could not have been acquired from forms of evidence plausibly available".  Children encounter many variations of situation and input.  Nevertheless all of them achieve grammatical competence.  Hence, even if it were in principle possible to acquire some piece of linguistic knowledge from some unusual type of evidence, this is not good enough; it has to be shown to happen to all learners.  Again a single learner who knew some aspect of language without having encountered this particular evidence type would be enough to undermine it.  It is not sufficient to show that such evidence has in fact occurred in one or more situations - it is irrelevant whether a generation of linguists have been saying to their children "You must use structure dependency".  Such evidence has to be shown to be, not only logically possible and actually occurring, but also available to all children.  This has been called elsewhere the uniformity requirement (Cook, 1988).  If language acquisition happens everywhere, it cannot be explained by appealing to a local or idiosyncratic property found only in rare or specialised situations.  Many potentially promising features of the environment, such as specialised interaction with parents, modified input, and parental correction, can be ruled out as sources of evidence because they are not true for all learners in all cultures.  As all L1 children acquire the grammar, to be 'plausible' such evidence must be shown to be available to all children not just one.  So Step B therefore modifies the argument by the addition of "to all children".

   STEP B' X could not have been acquired from forms of evidence plausibly available to all children

The main motivation for broadening the argument is therefore to include many people.  The narrow argument could be gainsaid by citing a single child or culture in which some type of evidence occurred, say parental correction; Tony at 33 Railways Cuttings, East Cheam, knows structure-dependency because his mother explained it to him.  The broad argument maintains that, since all children acquire competence, whatever evidence is necessary to do so must be available to all of them, essentially cutting the possibilities down to positive evidence.

In practice the transition from the narrow form of the argument, which applies to a single individual, to the broad form, which applies to everyone, has been made without acknowledgement.  Chomsky (1981) starts discussing the poverty-of-the-stimulus argument on page 8 in terms of 'the child' and by page 9 has switched to 'children'.  This is the explanation then for the stylistic inelegance of "him or her" so far in the present paper; for once the problem of gender in written English cannot be resolved by the expedient of making the sentences plural since the singular/plural distinction overlaps the narrow/broad distinction.  The uniformity requirement is seen as a commonplace proof of step B, while in fact it only applies to a broad version of the argument involving more than one learner, i.e. to B'.  The broad argument in a sense includes some aspects of development in that it bases itself on an inspection of some of the actual conditions of learning, however cursory this may be.

Certain problems arise when the broad version of the argument is extended to L2 learning.  The broad L1 argument assumes that all adults attain the same grammatical competence.  The step from the narrow to the broad versions is not too dangerous as one child can fairly represent all children.  But one of the few things agreed about L2 learning is that learners reach very different levels of L2 knowledge.  Some manage a small part of the language, others substantially more, few as much as in the L1.  The final steady-state knowledge of L1 is called Ss; to show the difference of terminal knowledge of L2, this can be distinguished as St.  Ss is effectively the same in all adults - all native speakers share roughly the same grammatical competence.  It is usually believed that St varies from one L2 learner to another, and that it is mostly far from the equivalent Ss in a native speaker, though much of the evidence for this is from areas other than core grammar, as discussed in White (1989).  If the final competence varies, the broad poverty-of-the-stimulus argument is changed.  Step A' native speakers know X cannot be converted into L2 learners know X because all learners do not achieve the same competence.  It might be that only those L2 learners should be considered that reach a state of St equivalent to their L1 Ss, i.e. so-called balanced bilinguals.  But these only represent a small fraction of L2 users and indeed study of advanced learners shows that they still have subtle differences from native speakers (Coppetiers, 1987).  While it is possible to base the L1 argument on the fact that all children acquire the same competence, the force of the broad argument is diminished when it depends upon a small and untypical group of learners.  This obviously does not affect the validity of Step A of the narrow argument a native speaker knows X in L2 learning; it may be possible to find particular individuals who know something in the L2 which they have not learnt from the environment.

Why should L2 learners have different Sts?  One possibility is that access to UG is unavailable or limited, as discussed below, and so L2 learners are forced back onto other ways of learning.  Another is that the type of evidence has more effect on L2 learning than on L1 learning, i.e. Step B' has to be qualified.  In classroom learning the teacher's job is precisely to manipulate language evidence to get a better effect in the students in whatever way they think appropriate.  Step B' in L1 learning depends upon the uniformity requirement that some form of evidence is available to all learners and hence rules out evidence types that are unavailable].  Not so in the L2 classroom, where almost any conceivable type of evidence has been used.  Some learners have been hypnotised, some have studied in their sleep, some have had certain grammatical rules explained to them, some have had the world of meaning reduced to a set of coloured sticks, some have sat in groups and bared their souls, others have sat in language laboratories repeating after the tape.  The idea of plausible evidence having to be available "to all children" is not true of L2 learning.  The argument cannot depend upon the availability of evidence to all learners since learners' Sts and situations vary dramatically.  Because St is variable in L2 learners, the question is, not whether correction actually occurs - clearly it does in some circumstances - but whether it is effective; does the presence of correction in the classroom lead to success in the student? 

Many L2 learners also encounter a further type of negative evidence, which I have called explanatory evidence (Cook, 1985); they are given explanations of the language system grammar which they consciously understand.  This is not available in L1 acquisition, so far as is known, and certainly does not meet the uniformity requirement of being available to all learners.  It may be obvious, as Chomsky has pointed out, that teachers do not have the linguistic knowledge to give adequate explanations: 'one does not learn the grammatical structure of a second language through "explanation and instruction" beyond the most elementary rudiments, for the simple reason that no one has enough explicit knowledge about this structure to provide explanation and instruction' (Chomsky, 1972, pp.174-5).  But what if they did?  Why shouldn't a teacher who has studied linguistics explain, say, Binding Theory?  If L2 learning is susceptible to different types of evidence, there is no reason for excluding explanatory evidence simply because it has not actually occurred yet.  At Essex University for instance all modern language students have to take linguistics and a course in the linguistic description of the language, because they are believed to help them speak the language.  Perhaps teaching a full Government/Binding account of a foreign language might be a highly effective language teaching technique.

Let us sum up this section.  The narrow argument still applies to L2 learning in that, if we can show an individual knows an L2 fact that was unlearnable, it must be built-in.  But 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 uniformity requirement cannot be met, nor can the restrictions on plausible forms of evidence.  A workable version of the argument in suitably watered-down form might then be:

Step A'' some non-native speakers know X
Step B'' X could not have been acquired from forms of evidence actually available to those speakers
Step C'' X is not learnt by those speakers
Step D   X is built-in to the mind

One crucial implication is that research into the actual evidence available to specific L2 learners is much more important.  The L1 broad argument depends upon plausible evidence available to all learners and this may be established quite easily; obviously most children do not receive explanatory evidence.  The broad L2 argument needs greater justification from research that correlates the actual evidence available to learners with the state of their Sts.

The multi-competence model of UG

Before developing the relationships of L2 learning to the poverty-of-the-stimulus argument further some other familiar connections between the UG theory and L2 learning need to be looked at.  Perhaps the major issue that has dominated discussion is whether the L2 learner still has access to UG, and if so in what way.  The following diagram illustrates the different possibilities:

     Universal Grammar                 Other mental faculties
             S0..                                             |
             |  :....                                            |
             |      :....                                        |
             |          :....  direct                         | no
             |              :....  access                    | access
             |                  :....                            |
             |                      :....                       \ /
            \ /                         :----------------> L2
           L1 Ss -----------------------------    > St
                         indirect access

         Figure 1 Models of access to UG in L2 learning

First language acquisition starts from the zero state of knowledge S>0 and ends with a final steady state of knowledge Ss.  The terminal state of knowledge in second language acquisition St may be derived either from other mental faculties (a no-access model, i.e. non-availability of UG), or from the UG S>0 (a direct-access model), or from the L1 Ss (an indirect-access model) (Cook, 1988).  Much discussion has taken place in L2 learning research about these three positions, ranging from Ritchie (1978) to Bley-Vroman et al (1988).  The merits of the case in favour of each of these positions is not relevant here.  All of them predicate four distinct entities - a Universal Grammar S0, an instantiation of UG in a grammar of a particular language Ss, non-language faculties of the mind, and a grammar of a second language St.  They differ over the relationships between these four objects rather than over their existence. 

A second general issue is the classic black box metaphor for language acquisition.  This can be redrawn in current parameter-setting terms, as seen below.  L1 evidence goes in to the black box, triggers settings for parameters, and yields an output consisting of Ss, i.e. the grammar of a particular language.  The poverty-of-the-stimulus argument was originally couched in terms of this black box model.  What can be found in the output Ss that is not learnable from the input is ipso facto added by the black box and hence a built-in aspect of the Language Acquisition Device (LAD).  The search for explanatory adequacy that motivated the linguist's choice of description for S>s (Cook, 1974) has now become a requirement that the postulated competence either must be derivable from positive evidence that triggers parameters or must consist of built-in principles of the mind or default parameter-settings.
 L1 input    --->   triggers settings for    --->    S>s (a grammar of
(positive               UG parameters                          a language)
evidence)

Figure 2. The black box model of L1 acquisition

Into the black box goes the linguistic input, out pops an S>s grammar of English - or of Japanese or of whatever language the child encounters - which instantiates the principles of UG and has particular values for each parameter of variation - pro-drop, head parameter, etc. 

The same black box metaphor can be applied in a straightforward manner to L2 learning:                    
                                     
    L2 input          ---> triggers settings  ---> St (a grammar of
    (many types)          for UG parameters          an L2)    
      Figure 3. The black box model of L2 acquisition

In go L2 data, out pops L2 grammar S>t.  The L2 learning debate has chiefly concerned the role of the L1 in this, i.e. the type of access possible.  Triggering the parameters in an L2 may be exactly the same as triggering them in an L1 in a direct access model, or the learners may have to start from the L1 values in an indirect access model, or may have nothing to do with UG at all in a no-access model.  But, apart from the no-access model, the end product is an L2 grammar that instantiates the principles of UG and sets the parameters appropriately.  Putting the L1 and the L2 metaphors side by side shows two products, Ss and St, coming out of similar processes from two different inputs.  Again there are two discrete entities: Ss and St.

But the L1 and the L2 are after all known by the same person; the two grammars S>s and St are necessarily in one and the same mind.  The mind does not split in two when it gains a second language; bilinguals don't have two heads.  The fact that the black box model yields two products S>s and St makes them appear entirely distinct rather than interrelated.  The mind of the L2 learner simultaneously possesses two finished grammars, two instantiations of UG, competences in the L1 and in the L2. 

The compound state of a mind with two grammars S>s and S>t can be called 'multi-competence', contrasting with the normally accepted 'mono-competence' consisting of only Ss.  Ss and St are not distinct separate products but coexist.  For many people UG exists in the mind in two instantiations, L1 and L2 - multi-competence.  Similar arguments apply to the input.  The input for different L1s is heard by different children.  One child hears English, the other Japanese; both learn the language they hear.  The inputs are separate and discrete.  But in the case of the L2 learner, it is the same person who hears both the L1 and the L2 inputs, not different people; it all forms part of the total linguistic experience of the learner, a compound input.  The two separate L1 and L2 diagrams should then be combined into one, as seen in Figure 4.

Compound input         Black box               Multi-competence
     L1 input        --->   triggers settings      --->  Ss  
                                   for UG parameters
     L2 input        --->                                   --->  St                                  
  Figure 4. The combined black box model of L1 and L2 acquisition

An objection that can be anticipated to the collapsing of the two black box models into one is that L2 learning mostly, but not invariably, takes place after L1 learning.  While correct in terms of chronological development, neither the poverty-of-the-stimulus argument nor the black-box model depend on development through time but are 'instantaneous' logical models; temporal priority between L1 and L2 is beside the point.  Similarly the fact that the L1 input comes before the L2 input or is heard in different situations is neither here nor there: the same mind hears both.  The logical argument as usual pays no account to sequence of input.  Much of the discussion of types of input in L2 learning has dealt with the L2 learner as if the L1 input had never existed and is totally irrelevant rather than being an inevitable part of the L2 learner's experience.

Multi-competence as the norm

So far this paper has amplified the standard UG concepts to take account of L2 learning.  Most researchers might well concede that there is some merit in the argument but see it as a special pleading for a particular sub-type of language learning.  The crucial question is whether the implications of this line of arguing extend outside the case of the L2 learner.  The basic foundation is the decision whether mono-competence in which the mind knows one language should be taken as normal or multi-competence in which the mind knows two.  It is commonly assumed that acquiring one language is the unexceptional norm for a human being.  Acquiring two is tacitly assumed to be something that is peculiar, difficult, an intellectual achievement, a problem - in other words anything but commonplace.  An English person finds it remarkable that someone can use more than one language in their everyday life.  However a person from the Cameroon may use four or five languages in the course of a day, taken from the two official languages French and English, the 4 lingua francas, or the 285 native languages.  On some calculations there are more people in the world like the Cameroonian than like the Englishman; 'there are 3000-5000 languages in the world but only about 150 countries to fit them all into' (Harding & Riley, 1987).  Taking classroom second language learning into account, the European Commission survey into young people in Europe found that 83% of 20-24 year olds had studied a second language (EC Commission, 1987). (See Cook, 1991b for further discussion).  Knowing a second language is a normal part of human existence for most human beings: it may well be unusual to possess only mono-competence. 

According to Illich & Sanders (1988, p.52), the assumption of mono-competence is typical of contemporary views of language:

 'One obstacle most modern readers face when they want to study the history of "language" is their belief in monolingual man.  From Saussure to Chomsky "homo monolinguis" is posited as the man who uses language - the man who speaks." 

A quotation from Chomsky can show this is a fair characterisation of Chomskyan linguistics:

'We exclude, for example, a speech community of uniform speakers, each of whom speaks a mixture of Russian and French (say, an idealized version of the nineteenth-century Russian aristocracy).  The language of such a speech community would not be "pure" in the relevant sense, because it would not represent a single set of choices among the options permitted by UG but rather would include "contradictory" choices for certain of these options.' (Chomsky, 1986, p.17). 

This quotation occurs in the context of discussion of the idealisation to the homogeneous speech community necessary for studying linguistics.  But the abstraction to a monolingual speaker is of a different kind from assuming that all speakers in a community are effectively the same; it is simplifying the average mind rather than finding an average mind, so to speak.  If most human beings are in fact multi-competent, mono-competence is a misleading representation of the human species rather than a convenient idealisation. 

Both the narrow and broad poverty-of-the-stimulus arguments see the mind as coming to acquire a single grammar, a single setting for each parameter.  The narrow poverty-of-the-stimulus argument is concerned with the individual.  However, if a single person is found, even a nineteenth century Russian aristocrat, who knows two grammars rather than one, the poverty-of-the-stimulus argument can be applied to his or her case.  It is a legitimate question to ask how that person acquired multi-competence. 

The implications for the broad argument are as important.  The norm is for a mind to possess one or more instantiations of UG rather than one, with one or more settings for each parameter.  In terms of input, the broad poverty-of-the-stimulus argument involves the occurrence and uniformity requirements.  Not all children hear input in two languages; therefore it is legitimate to look at the child in the monolingual environment.  But it seems to be normal that a child who gets input in two languages acquires two grammars.  UG theory has dealt with cases of wolf children and others who have failed to achieve normal language by saying that it is not concerned with total deprivation of all language input; clearly some language input is needed to instantiate principles, to instantiate principles, to set parameters and so on.  If children are arbitrarily restricted to one language input this forms a gross form of impoverishment.  The fact that some children receive impoverished input in which they hear only one language is irrelevant to the poverty-of-the-stimulus argument, which does not take such vagrancies into account.  The typical human environment includes input in more than one language; monolinguals are suffering from a form of language deprivation.  Hence the broad poverty-of-the-stimulus argument should be, not how the mind manages to acquire a single grammar, but how it manages to acquire one or more grammars.  In terms of a UG model the problem is not explaining how the mind sets a parameter from evidence; it is how it manages to acquire one or more settings from evidence.  Otherwise, as most human beings know two grammars, the poverty-of-the-stimulus argument would deal with only a minority of human beings.  Mono-competence is as relevant to language learning as the behaviour of those with one leg is to the study of human walking.  The basic claim is that the learning of multi-competence should be seen as the essential issue, mono-competence as the aberration.  The problem of language acquisition is how one mind acquires one or more grammars from input.  Chomsky's idealisation rules out most of the human race and does not cope with the perfectly normal condition of the human mind.

Let us come back to the idea of negative evidence.  The classic argument is that the acquisition of competence depends solely on positive evidence, that is to say on the provision of actual sentences of the language; hearing "John ate the apple" is enough to trigger the head parameter (Chomsky, 1988).  Or, in slightly more sophisticated versions, negative evidence must always be supported by positive evidence (Saleemi, 1988).  One possibility is that the acquisition of multi-competence requires the use of negative evidence and that there may be certain aspects of the L2, say binding, which are only accessible via this route.  Given our formulation, negative evidence needs to be looked at within the total multi-competence framework.  If multi-competence is the norm and negative evidence can arguably be applied to the St element of it, then negative evidence can be seen as part of the whole argument rather than immediately ruled out of court.  The fact that children don't get it may again be accidental; if they did, they might benefit from it.  The evidence types discussed under B' must now include those that are known to be beneficial to L2 learning as well as those that occur in L1 learning.  It is the total architecture that counts.

Beyond the present discussion lies the intriguing question of how the different parts of multi-competence are related.  So far it has been assumed that there are indeed two or more discrete grammars Ss and St, an assumption that can be termed 'separatist multi-competence'.But the two grammars may form an inter-related system.  Switching from speaking the L1 to speaking the L2 may not be taking one grammar off-line and replacing it with another, but weighting the parts of one overall system, i.e. 'wholistic multi-competence'.  Some evidence for the interrelationship of the separate elements of multi-competence is the availability of L1 settings while using the L2 in codeswitching and comprehension (Grosjean, 1989), the difference of St from Ss even in advanced L2 speakers (Coppetiers, 1987), and the effects of the L2 on the processing and evaluation of the L1 (Flege et al 1987); Nathan, 1987); these are discussed in Cook (in progress).

This is not the place to develop the concept of multi-competence further.  Some of the problems left unsolved are: deciding at which point of L2 development an S>t qualifies to be called an L2 competence at all; relating multi-competence to the use of language; evaluating how multi-competence relates to the question of access to UG in L2 learners; and deciding how multi-competence relates to simultaneous rather than consecutive L2 learning.  The purpose here was to draw attention to the need to incorporate multi-competence into the argumentation of UG.  There may be many valid objections to the specific points put forward here.  But the general point remains: the existence of multi-competence needs to be accommodated more fully within the learnability arguments.  It is unclear why this normal state of the human mind has been put to one side in such arguments, except of course for the cultural conventions of countries and academic communities that see homo monolinguis as normal, homo multilinguis as a peculiar oddity.

References

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

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

Chomsky, N. (1988), Language and Problems of Knowledge: The Managua Lectures, MIT Press

Commission of the European Community (1987), Young Europeans in 1987, EC Commission

Cook, V.J. (1974), 'Is explanatory adequacy adequate?', Linguistics, 133 online version

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

Cook, V.J. (1988) Chomsky's Universal Grammar: An Introduction,    Blackwell

Cook, V.J. (1991a), 'Observational evidence and the UG theory of language acquisition', in I. Roca (ed.), Logical Issues in Language Acquisition, Foris, to appear

Cook, V.J. (1991b), Second Language Learning and Language Teaching, Edward Arnold, to appear

Cook, V.J. (in progress), 'Evidence for multi-competence' online version

Coppetiers, R. (1987), 'Competence differences between native and near-native speakers', Language, 63, 544-573

Crain, S.C., & Nakayama (1983), 'Structure dependence in grammar formation', Language, 63, 3, 522-543

Flege, J.E. (1987), 'The production of "new" and "similar" phones in a foreign language: evidence for the effect of equivalence classification', Journal of Phonetics, 15, 47-65

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

Harding, E., & Riley, P. (1986), The Bilingual Family: A Handbook for Parents, C.U.P.

Illich, I., & Sanders, B. (1988), The Alphabetization of the Popular Mind, Marion Boyars, GB

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

Ritchie, W. (1978), 'The right roof constraint in an adult acquired language', in W. Ritchie (ed.), Second Language Acquisition: Issues and Implications, New York, Academic Press

Saleemi, A. (1988), Learnability and Parameter Fixation, Ph.D., University of Essex

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

A further more recent experiment-based paper on the poverty of the stimulus argument is here