Chomsky's Universal Grammar and Second Language Learning
Online version of ancient paper in Applied Linguistics (1985) - cartoons added. Note: the syntactic model used here is now very dated, though the general ideas of UG are still usable. For my more recent views see More Galilean Challenges for Chomsky (2017) (with a reply by Chomsky!), Multilingual UG as the norm (2009) or Chomsky's Universal Grammar (with Mark Newson 3rd edition 2007)
Vivian Cook Other on-line writings Second Language Acquisition Multi-competence
Having gone underground for a few years, once again Chomsky's ideas of language learning are being discussed.  A recent book, The Language Lottery (Lightfoot 1982), readably outlines the theory; several collections report research into its implications (Tavakolian 1981; Goodluck and Solan 1978) and its theoretical aspects (Hornstein and Lightfoot 1981b; Baker and McCarthy 1981); most importantly a string of books by Chomsky himself has shown the development in his views of language acquisition (Chomsky 1976; Chomsky 1980; Chomsky 1981a). This paper tries to take stock of recent Chomskyan thinking in terms of second language (L2) learning. The first section outlines the theory itself, mostly drawing on Chomsky's own work; though parts may be familiar from earlier versions, such an overview is necessary in order to ensure coherence. The second section considers the implications for L2 learning, particularly important because they appear to contradict some of the cherished assumptions in the field; it should, however, be noted that Chomsky himself has not extended the theory to L2 learning, apart from occasional scattered allusions. While the first part attempts to present a consensus view of the LI theory, the second is much more an individual interpretation of the theory for L2 learning.
1. THE THEORY OF UNIVERSAL GRAMMAR AND LANGUAGE ACQUISITION
A typical way in to the Chomskyan position is through a simple conundrum (Baker 1979): an adult native speaker of a language knows things he could not have learnt from the samples of speech he has heard; since this knowledge is not based on his experience of the world, it must come from some property inside his own mind.
Take the following two sentences, 'Is the programme that is on television any good?' and *'Is the programme that on television is good?' A speaker of English immediately knows that the first sentence is possible and that the second is not; he knows in some sense that the 'is' that is shifted to the beginning of a sentence in a question comes out of the main clause, rather than the subordinate clause.
But how could he have acquired this piece of knowledge about English? Some of the sentences he might have encountered during his life are 'The programme is good', "The programme that is on television is good', 'Is the programme good?', and so on. None of these shows the rule being broken; they give him information about what he can say, not about what he can't say. The rule can be demonstrated to exist only by concocting an ungrammatical sentence that would never occur in real life, *'Is the programme that on television is good?', or by giving a grammatical analysis. But these are the kinds of information that the child learning his first language precisely does not have available to him. If native speakers find the sentence ungrammatical, their judgement must be based on something other than their experience of the world; the remaining possibility is that it is derived from some property of the human mind that they all share.
A second example from English is the well-known pair, 'John is eager to please' and 'John is easy to please', taken from the earlier 'Aspects' model (Chomsky 1965);-on the surface the two sentences seem to have the same structure but, looked at more closely, their underlying structures differ in that in the first John is claimed to please other people, in the second other people are claimed to please John. The sentences of English that the speaker has heard may have included 'Mary is eager', "This is easy', 'Is John eager to please?', and so on, none of which differentiates the two structures. Conceivably an adult might explain the difference to the child, or some feature of the particular situation might make it obvious; such accidental and improbable occurrences cannot explain why children go through the same stages in acquiring 'eager/easy to please' and are successful at about the same age (Cromer 1970). If the child has not learnt the distinction from the input, he must have done so from some property of his own mind. Both examples therefore exploit the same argument, known as 'the poverty of the stimulus', to show that the child knows things about language he could not have learnt from outside, that important aspects of language are not strictly speaking learnable. [Note: Cook 2003 tries to prove this for SLA from an easy/eager experiment].
[cartoon from How children learn language 1971]
The language properties inherent in the human mind make up 'Universal Grammar', which consists, not of particular rules or of a particular grammar, but of a set of general principles that apply to all grammars and that leave certain parameters open; Universal Grammar sets the limits within which human languages can vary. A native speaker of English knows that the sentence 'The train is arriving' is grammatical but *'arrives the tram' and *'arrives' are not; the native speaker of Spanish knows that not only is 'el tren llega' (the train arrives) grammatical, but so also are 'ha llegado un tren' (arrives a train) and 'han llegado' (arrives). One of the parameters that is open in Universal Grammar is the pro-drop parameter which is concerned roughly speaking with the relationship of government between Subjects and Verbs (Chomsky 1981a). English chooses not to have pro-drop; a Subject is required for every sentence and it cannot be inverted with the verb in declarative sentences. Spanish, however, is a pro-drop language in which 'empty' Subjects can occur and inversion can take place, indeed is compulsory in certain circumstances (Green 1976). Hence a particular grammar amounts to a specification of the ways in which it selects from the different possibilities inherent in Universal Grammar. 'The grammar of a language can be regarded as a particular set of values for these parameters, while the overall system of rules, principles, and parameters is UG . . . ' (Chomsky 1982). A partial analogy might be made to the relationship of the European Convention on Human Rights to laws passed in the UK Houses of Parliament; the Convention does not force particular laws on the UK, but it establishes certain principles that the actual laws must conform to; it sets parameters within which the laws can vary.
One way of visualizing Universal Grammar is to see it as part of the brain: 'We may usefully think of the language faculty, the number faculty, and others as "mental organs", analogous to the heart or the visual system or the system of motor coordination and planning' (Chomsky 1980: 39). Consequently, 'learning' is not the right word to describe how language develops. A bulb becomes a flower; some cells become a lung. We do not say that the bulb 'learns' to be a flower or the cells 'learn' to be a lung, although in both cases certain aspects of the environment such as water and nourishment are necessary to the process. Instead we say the bulb and the cells 'grow'. Their growth is the realization of their genetic potential in conjunction with 'triggers' from the environment, the achievement of something that was within them from the start. Why then do we say that the child 'learns' language rather than language 'grows'? Universal Grammar present in the child's mind grows into the adult's knowledge of the language so long as certain environmental 'triggers' are provided; it is not learnt in the same way that, say, riding a bicycle or playing the guitar are learnt: 'a central part of what we call "learning" is actually better understood as the growth of cognitive structures along an internally directed course under the triggering and potentially shaping effect of the environment' (Chomsky 1980: 33). Language acquisition is the growth of the mental organ of language triggered by certain language experiences. Hence the theory of Universal Grammar is frequently referred to as part of biology. Indeed the theory is not dissimilar from ideas current in biology on other issues, for instance the view that 'Embryogenesis may then be seen as the progressive, orderly manifestation of the knowledge which is latent in the egg' (Goodwin 1976: 202).
So, to acquire language, the child needs not only Universal Grammar but also evidence about a particular language; he needs to hear sentences of English to know how to fix the parameter for the order of Verb, Subject, and Object. The evidence he encounters can be positive or negative (Chomsky 1981a). Positive evidence consists of actual sentences of a language; by hearing 'The cow jumped over the moon' or 'Johnny loves cabbages', the child learns that English has Subject-Verb-Object order. Negative evidence falls into two categories, direct and indirect. Direct negative evidence consists of corrections of the child's mistakes by adults: 'You mustn't say "you was", Jimmy, you must say "you were" '. Indirect negative evidence is provided by the non-occurrence of something in the language the child hears; the fact he never hears Subject-Object-Verb order is negative evidence that English is a Subject-Verb-Object language. First language acquisition relies chiefly on positive evidence; the child apparently receives little direct negative evidence in the form of correction of syntax (Brown and Hanlon 1970). The few corrections that occur are largely about dialectal or socially stigmatized forms or socially prescribed politeness formulas, a small fraction of English. The importance of indirect negative evidence is difficult to assess, since it is clearly impossible to specify everything that the child doesn't hear. Its value, however, depends upon the child already having certain expectations about language that are not fulfilled, in other words it presupposes a Universal Grammar in the child's mind. The whole of Universal Grammar does not manifest itself in the child's speech at the same time. Language principles that apply to long or complex sentences are needed only when the child has the capacity actually to produce them; the parameters for SVO order, for example, cannot apply when the child says only one word at a time. Although language is a separate mental organ, its development is influenced by other organs. While the claim that cognitive level and short-term memory capacity limit the type of structure that may be employed has always been part of the theory, the version current in the 1960s was sometimes interpreted as claiming that the child's first sentences are closer to Universal Grammar; McNeill, in a typical remark, said, though with qualifications, 'Early speech is supposedly free of transformations and therefore should be a direct manifestation of children's capacities' (McNeill 1970: 22). In the theory, however, the language principles that are present manifest themselves in accordance with the child's capacity to process information, and other maturational factors; the child cannot reveal all he knows about language, because of his other limitations. A distinction may be drawn between development—the real-time learning of language by children—and acquisition—language learning unaffected by maturation, sometimes called the instantaneous acquisition model (Chomsky 1965; Pinker 1981). Sequence of development reveals more about other cognitive systems than about language acquisition: '. . . suppose it is a fact that children generally acquire the use of simple one-clause structures before compound sentences; there is no reason to assume that this fact must follow from some particular principle of the theory of grammar, as opposed, let us say, to some property of perceptual maturation or the developing short-term memory system' (Hornstein and Lightfoot 1981a: 15). So far as acquisition is concerned, the interim developing grammars of the child are irrelevant. It is nevertheless an open question in terms of development whether the child starts with all the principles of Universal Grammar available, or whether they gradually unfold as part of maturation: evidence to distinguish principles that are present but cannot be revealed from those that are absent is hard to conceive.
If Universal Grammar is present in toto from the beginning, all human languages should conform to the language principles, whether the stable grammars of adults or the temporary grammars of learners: 'A grammar . . . represents a person's linguistic knowledge, whether the person is two years old or twenty-two' (White 1981: 47). On this assumption, while many principles are missing from the interim grammar, the rules that are present do not break them.
Let us now try to specify what the theory is actually dealing with. First of all it is concerned with grammatical competence, the speaker's knowledge of the language, not with what Chomsky calls pragmatic competence—the ability to place 'language in the institutional setting of its use, relating intentions and purposes to the linguistic means at hand' (Chomsky 1980: 225). It contrasts grammatical competence with pragmatic competence rather than communicative competence, since there are many uses of languages other than communication: 'either we must deprive the notion "communication" of all significance, or else we must reject the view that the purpose of language is communication' (Chomsky 1980: 230). Secondly, for those familiar with earlier versions it is plain that rules play a much less central role. Chomsky (1982) shows elegantly how rules are in fact consequences of principles of Universal Grammar and of the way in which particular parameters are set. A grammar consists of a specification of the values of parameters, which may be represented as rules, but these are of secondary importance. Thirdly, a new distinction separates core grammar (those parts of the language that have 'grown' in the child through the interaction of Universal Grammar with the relevant language environment) from peripheral grammar (the parts outside the core). Any particular language such as English contains elements that are derived from its history (the structure of 'the more the merrier', for instance, comes from Old English), that are borrowed from other languages (the pronunciation of 'police' shows it is a late borrowing from French), or that have been added to it by other accidental processes ('the dreaded lurgy' entered British English and became a children's game because of a radio programme); these do not reflect the principles of Universal Grammar in the same way as the core:'... it is reasonable to assume that UG determines a set of core grammars and that what is actually represented in the mind of an individual even under the idealisation to a homogeneous speech community would be a core grammar with a periphery of marked elements and constructions' (Chomsky 1980: 8). While some aspects of an individual's grammar come from Universal Grammar, others are influenced by other factors.
Let us now see what this means for learning. By using the same language principles, a French child constructs a grammar of French, an English child a grammar of English. The two grammars represent different choices within the guidelines set by Universal Grammar, different applications of the same linguistic principles in response to different environments; 'Experience is necessary to fix the parameters of core grammar' (Chomsky 1981a: 8). But the children also have to learn aspects of language that are peripheral, that do not conform to Universal Grammar. The child's mind 'prefers' to adopt rules based on the handy set of principles with which it is equipped; they are in a sense the easy way out, and need only triggering experience to be learnt. By listening to the language around him, he can decide how to fix the parameter of sentence order as SVO or SOV, for instance. His mind 'prefers' not to adopt peripheral solutions, as they fall outside his pre-programmed instructions; they are more demanding. This may be interpreted through the concept of markedness: the child prefers to learn 'unmarked' knowledge that conforms to Universal Grammar, rather than 'marked' knowledge that is less compatible with it. Core grammar and peripheral grammar are weighted differently in the child's mind. Chomsky sees peripheral learning as systematic and related to core learning; 'there should be further structure to the system outside of core grammar. We might expect that the structure of these further systems relates to the theory of core grammar by such devices as relaxing certain conditions of core grammar, processes of analogy in some sense to be made precise, and so on, though there will presumably be independent structure as well' (Chomsky 1981a: 8). Hence we may expect to find a continuum of markedness from core to periphery. The distinction does not, however, entail that core unmarked grammar is necessarily learnt first. 'We would expect the order of acquisition of structures in language acquisition to reflect the structure of markedness in some respects, but there are many complicating factors; e.g. processes of maturation may be such as to permit certain unmarked structures to be manifested only relatively late in language acquisition, frequency effects may intervene, etc' (Chomsky 1981a: 9).Sequence of development again is an unreliable guide to acquisition.
Though not part of the theory itself, earlier versions were often associated with the notion of language acquisition as hypothesis testing. It is important, however, to define the sense in which hypothesis testing is acceptable. One interpretation has been that the child creates a hypothesis about the grammar more or less at random; he produces sentences according to his hypothesis, and the feedback he receives from the situation tells him whether or not it is correct. In this sense, hypothesis testing has never recovered from the blow administered to it by Martin Braine, who argued that it required negative as well as positive evidence to be successful (Braine 1971; Baker 1979); the child cannot discover if his hypothesis is right or wrong if he is not told when he makes mistakes. But, as we have seen, it is widely accepted that correction is infrequent; the child does not meet enough negative evidence to reject incorrect hypotheses. Nor does he produce enough incorrect sentences to test out hypotheses adequately; while at the earliest stages it could be said that the child's sentences are incorrect, after, say, the age of four, though the child's language lacks many structures, it is by and large grammatical. The lack of negative evidence and of incorrect sentences shows the inadequacy of hypothesis testing through feedback from outside. Another interpretation of hypothesis testing is nevertheless acceptable: To acquire language, a child must devise a hypothesis compatible with presented data—he must select from the store of potential grammars a specific one that is appropriate to the data available to him' (Chomsky 1965: 36). Universal Grammar allows different core grammars in different languages; the child has several initial hypotheses to choose from, several parameters to fix; his internal Universal Grammar severely restricts the range of hypotheses he can entertain, the final choice depending upon evidence from the environment. Hypothesis testing is a possible explanation for language acquisition, in the sense that the child chooses from the limited number of possibilities provided by Universal Grammar in accordance with the evidence he meets.
Let us try to summarize the roles of the environment and of cognition within the theory, two threads that have run through the discussion. First, so far as the environment is concerned, Chomsky has been at pains to point out that all learning involves inherent properties of the child's mind: 'Every "theory of learning" that is even worth considering incorporates an innateness hypothesis' (Chomsky 1976: 13). Even behaviourism attributes to the child an ability to form associations of stimulus and response. All learning theories are therefore interactionist in that they have to take into account both the learner and the situation; to quote Martin Buber in a rather different context, 'Meaning is not in us or in things, but between us and things it can happen'. Theories differ in how they strike the balance between person and situation, Chomsky coming down heavily at the learner end, behaviourism at the situational end. Chomsky's theory assigns a precise role to the environment: negatively it denies that it provides sufficient evidence for the learning of particular aspects of linguistic knowledge without the aid of a powerful in-built grammar; positively it suggests the environment provides positive evidence to help the learner fix the ways in which Universal Grammar applies to the language he is learning. Universal Grammar makes certain things obligatory in any grammar; others it leaves free to vary within pre-set limits; the environment provides evidence about the particular limits that apply in a given case. "Thus, UG presumably determines that a clause S will have a complementiser and propositional content S, which may be tensed or infinitival. But experience determines that in English, infinitival clauses use the form to-Verb and may have the complementiser for rather than that or that buy and play are associated with concepts of the conceptual system as they are' (Chomsky 1981b: 38). One of the implications of the theory, therefore, is a shift in the balance of what is learnt from grammar to lexis. Much grammatical knowledge simply needs fixing though evidence from the environment. What does need learning is how particular lexical items can enter into various structures. 'A large part of "language learning" is a matter of determining, from presented data, the elements of the lexicon and their properties' (Chomsky 1982: 8).
The role of cognition is complex. There are two senses in which cognition is involved; one is the development of overall levels of thinking, the stages of cognitive development familiar from Piaget or Bruner; the other is the systems of information processing involved in handling language, which can be called channel capacity. So far as acquisition is concerned, the mental faculty of language does not need to be related to other faculties of the mind, as for instance Piagetans would claim that language presupposes certain cognitive operations (Sinclair-de-Zwart 1969). So far as development is concerned, language is bound up with other elements of cognitive maturation. Partly this is because development deals with language in use—with pragmatics and performance: 'much of the investigation of early language development is concerned with matters that may not properly belong to the language faculty, . . ., but to other faculties of the mind that interact in an intimate fashion with the language faculty in language use' (Chomsky 1981b: 35). Partly, however, as we have seen, language development interacts with cognition in that certain language principles cannot be deployed until the child has developed the channel capacity to handle them. So, for instance, short-term memory may be vitally important to development, because the length of sentence that can be uttered limits the principles that can be employed. Although language is an independent mental organ, in development it nevertheless needs to draw on other mental organs. Indeed, the same argument applies to physical organs: phonological development may be affected by the myelinization of the nervous system, which gradually allows more complex signals to be transmitted (Lecours 1975); the growth of gyrus granule cells in the hippocampal area of the brain may allow the child to attain a particular cognitive level at the age of about five (Rose 1980). Neither of these physical changes affects acquisition itself, but may have profound effects on language development. Thus certain aspects of cognitive and physical development can influence the order in which a child develops language.
As the purpose of this section has been to present the theory as a whole in order to bring out its implications for L2 learning, it is not the place to evaluate it or to discuss the acquisition research carried out within the theory reported in Tavakolian (1981) and Goodluck and Solan (1978). A criticism that is often voiced is its abstraction from the everyday world. Competence is separated from performance, grammatical competence from pragmatic competence, acquisition from development, core from peripheral grammar, each removing something from actual language use: 'To discover the properties of UG and core grammar we must attempt to abstract away from complicating factors of various sorts, a course that has its hazards but is inescapable in serious inquiry, in linguistics no less than in other domains' (Chomsky 1981b: 39). Some people would dispute whether such abstraction is valid; has the baby been thrown out with the bathwater? Bresnan and Kaplan (1982) claim, for instance, 'there is a scientific responsibility to show that the real does asymptotically approach the ideal under certain circumstances', a responsibility not shouldered by the present theory. Its power depends instead on the argument from the poverty of the stimulus that speakers know things they could not have learnt. Can a single paradoxical argument bear the weight that is put upon it? Chomsky himself insists that the argument of the poverty of the stimulus is not peculiar to linguistics, but part of all sciences concerned with development; it is so obviously true that birds do not 'learn' to have wings that people do not see that the same argument is involved. Moreover, Feyerabend has argued that science proceeds not through theory formulation and testing, but through the presentation of new ways of arguing, which he terms 'propaganda' (Feyerabend 1975). Thus Galileo's theory of relative motion relied not on actual evidence, but on an argument about an artist drawing on a ship in motion. The Chomskyan argument is in the same tradition, the presentation of an argument that is no more unrelated to what it is trying to explain than that advanced by Galileo.
2. UNIVERSAL GRAMMAR AND SECOND LANGUAGE LEARNING
The relevance of the theory to L2 learning depends not so much on the uncertain analogy to LI learning as on the original conundrum of the poverty of the stimulus: how can a speaker of a second language know things he could not have learnt from the language he has encountered? In other words, if the L2 speaker knows that *Is the programme that on television is good?' is ungrammatical, and if the knowledge is not demonstrably derived from experience, it must originate within his own mind. While no research has investigated L2 learners' judgments of such sentences, the other example used above, 'easy/eager to please', has been looked at by d'Anglejean and Tucker (1975) and Cook (1973), who found that L2 learners are indeed able to distinguish the two structures after they have been learning English for a certain period of time. In this instance the knowledge is not likely to be drawn from experience, since on the one hand 'easy/eager to please' has not figured in teaching syllabuses, structural exercises, or pedagogical grammars, and on the other it is improbable that native speakers have demonstrated it to the L2 learner. The answer to the conundrum is once again that the L2 learner's knowledge derives from some property of the mind. However, while the conclusion is the same as in first language acquisition, it may have a different explanation, in as much as the minds of L2 learners or the situations have different properties.
The most obviously different property is that the L2 learner possesses a grammar of a first language incorporating the principles of Universal Grammar and specifying a particular set of values for its parameters. Two possibilities for L2 learning need to be considered: the learner might have access to Universal Grammar either directly or indirectly through the first language. In a way this rephrases the debate about the relationship of first and second language learning—whether L2 learners start from scratch, or depend upon their first language; while the dust has never settled on this dispute, one position is that L2 learning is like LI learning when situational and cognitive factors are ruled out (Cook 1977): the apparent discrepancies are caused either by accidental or necessary differences in the situations, or by non-linguistic differences in the learners' minds, rather than by anything in the language process itself. So far as the principles of Universal Grammar are concerned, the question amounts to asking whether L2 grammars are constrained in the same way as LI grammars. Schmidt showed that a group of L2 learners of English produced only natural surface orders such as 'John sang a song and played the guitar' or 'John plays the. guitar and Mary the piano', rather than unnatural orders such as *'Sang a song and John plays guitar' or *'John the violin and Mary plays the piano', thus obeying the principle that only the second identical noun or verb may be omitted from the sentence (Schmidt 1980). Ritchie found that adult learners of English correctly judged sentences such as "That a boat had sunk that John had built was obvious' grammatical, and sentences such as *'That a boat had sunk was obvious that John had built' ungrammatical (Ritchie 1978), demonstrating that they still had access to a principle of Universal Grammar called the Right Roof Constraint (reformulated slightly differently in the present theory) that elements that are moved in the sentence must not cross certain types of boundary.
The notion of parameter-fixing can formulate the relationship between first and second language learning in a more precise way. To take a specific example, if Universal Grammar is directly accessible to the L2 learner, it should not affect a Spanish learner of English that the two languages have fixed the pro-drop parameter differently; lie simply needs the proper triggers to fix it anew. However, if it is not directly accessible, he can approach English only through the value of the parameter for Spanish. The question of whether L2 learning recapitulates LI learning can be narrowed down to considering whether L2 learners' grammars reflect the principles of Universal Grammar, and whether parameters are still free to be fixed in a second language from triggering evidence.
There is, however, a third possibility: in some sense the L2 learner might be cut off from Universal Grammar after a certain age. Lenneberg's Critical Period Hypothesis, henceforward CPH, set the limits for LI acquisition between the ages of two, before which the child is too immature physically, and twelve, after which the brain is too inflexible (Lenneberg 1967). Although Chomsky wrote of it approvingly in 'Aspects' (Chomsky 1965: 206), the CPH has not been discussed within the current framework. Presumably it would mean that after a particular age, the principles and parameters of Universal Grammar are no longer directly accessible to the learner; the older L2 learner has no option but to work through his L1 or through a non-language faculty. One reason for the lack of discussion of the CPH in the current theory may be that it is concerned with physical or cognitive maturation, that is to say, development; acquisition does not by definition take account of maturational factors. It has always been difficult to reconcile the CPH with successful L2 learning after the critical period. It has usually been salvaged by arguing that the first language acts as a mediator to Universal Grammar; the fact that a person over forty can learn to communicate in a foreign language 'does not trouble our basic assumption on age limitations because we may assume that the cerebral organisation for language learning as such has taken place during childhood, and since natural languages tend to resemble each other in many fundamental aspects, . . . the matrix for language skills is present' (Lenneberg 1967: 176). Chomsky himself has queried the importance of the LI as a mediator: 'While it may be true that "once some language is available, acquisition of others is relatively easy", it nevertheless remains a very serious problem—not significantly different from the problem of explaining first language acquisition—to account for this fact' (Chomsky 1969). In the current theory, mediation would be successful if the values of parameters were the same in the two languages; if the learner has been cut off from Universal Grammar, it is hard to see how he can retrieve the original parameters to fix their values differently in the second language. Evidence for the CPH in relation to L2 learning has been widely discussed by, inter alia, McLaughlin (1978), and Paivio and Begg (1981); the usual conclusion is that it is disconfirmed by evidence that adults and older children are better than younger children at L2 learning when the circumstances are the same, e.g. Asher and Price (1967) and Ekstrand (1976), and that there are not the expected differences between children acquiring their first language and adults acquiring their second when situational and other accidental factors are discounted. Within the present theory, the evidence from Ritchie (1978) and Schmidt (1980) suggests that a strong form of the hypothesis is not tenable, as learners after the critical period demonstrate they have access to at least some of the principles of Universal Grammar.
The answer to the conundrum may, however, be caused by differences in the environment; the L2 learner might know things he could not apparently have learnt because the situation supplies him with special types of evidence either not available to the native child or not usable by him. The kinds of evidence that an L2 learner encounters probably depend more on the type of situation than for the native child, particularly whether it is a 'natural' informal situation such as an immigrant using the language for everyday purposes, or an 'artificial' formal situation such as a classroom. In the natural setting one may assume that L2 learners probably meet positive and negative evidence in more or less the same proportions as native children; they also meet language modified by native speakers to their communication needs. One suspects, however, that correction is less likely with foreign adults, since it is more rude to correct an adult than a child. In the classroom setting on the other hand, direct negative evidence sometimes looms larger, since some teachers provide frequent and systematic correction, at least on a surface syntactic level. Older classroom learners may also encounter what can be called explanatory evidence, that is to say, explanations of the grammatical rules of the language. In practice few learners meet adequate explanations of syntactic points such as 'easy/eager to please'; as Chomsky points out, 'it must be recognised that one does not learn the grammatical structure of a second language through "explanation and instruction" beyond the most rudimentary elements, for the simple reason that no one has enough explicit knowledge about this structure to provide explanation and instruction' (Chomsky 1969). Nevertheless, in principle explanatory evidence is available to older L2 learners in formal settings in a way that it is not to the native child, as theories such as cognitive code learning (Carroll 1965) and the Monitor Model (Krashen 1981) have emphasized.
The role of cognition is also different in L2 acquisition, since the learner is not necessarily subject to the same maturational constraints. Let us first consider this in terms of cognitive levels. To study L2 learning in adults is in a sense to study language acquisition divorced from maturation, as Gass and Ard have argued (Gass and Ard 1980). Thus the 'natural' order of L2 development (Krashen 1982) is perhaps closer to acquisition than is LI development; it differs from LI development wherever the LI has, so to speak, been held up for lack of cognitive maturity or channel capacity. Gass and Ard (1980) suggest that children's order of acquisition of relative clauses follows their cognitive development, while the order of acquisition by L2 learners reflects a principle of accessibility; the L2 learner's development tells us more of acquisition than the native child's. While the formula that L2 learning equals acquisition is attractive, it rests upon the assumption that the channel capacity for language use depends upon maturation and does not need to be re-acquired in a second language. However, some aspects of channel capacity are not transferred to a second language, i.e. they are in some way dependent on the first language. In contrast to Gass and Ard's work it has been suggested, for example, that English relative clauses require a certain capacity of speech processing memory in the listener, gained by the native child around the age of seven; the foreign learner does not start from an adult position, but has to reacquire this channel capacity in the new language (Cook 1975). Similarly, some language-related aspects of memory such as rehearsal strategies and short-term memory capacity are substantially transferred to a second language (Cook 1981), others such as the clustering of vocabulary are not (Cook 1977). To say that L2 development is acquisition minus maturation is not the same as saying it is acquisition minus limitations upon channel capacity; L2 development is still affected by cognitive factors. The problem is discovering which cognitive processes need to be re-established in a second language, which can be transferred.
An overall conclusion for L2 learning research is that development is not necessarily reliable evidence for acquisition in a second language: the L2 sequence of development may reflect the re-establishing of 'channel capacity' for using the language, rather than language acquisition per se. Statements about sheer order of acquisition need support from accounts of the interrelationship between development, channel capacity, and other cognitive processes, before they can be considered valid for L2 acquisition or compared to LI acquisition. Hence the discovery of a common acquisition sequence for L2 learners, which has been hailed as 'surely one of the most exciting and significant outcomes of the last decade of second language acquisition research' (Burt and Dulay 1980: 325), must be seen as a first step in the description of development (and incidentally based largely on the presence or absence of a few surface syntactic features, rather than on underlying linguistic principles); it can say little about acquisition until the order has been shown to be the product of language acquisition itself, rather than channel capacity. The fact that L2 learners use simple one-clause sentences before complex sentences, or plural V before possessive V, or interpret 'eager/easy to please' initially as the same structure tells us more about L2 development than about L2 acquisition, without further information.
A related question is the concept of markedness in L2 learning, reviewed in Rutherford (1982). Within the current theory, unmarked aspects of grammar are those that are directly related to Universal Grammar and form the 'core'; marked aspects are less directly related to Universal Grammar and form 'peripheral' grammar; thus markedness reflects the degree to which something is related to Universal Grammar, and consequently the degree to which it is learnable by the child from inbuilt principles. One use of markedness within L2 research has been in connection with the Accessibility Hierarchy (Keenan 1972), which postulates a continuum going from rules that are most accessible and hence most widespread in human languages and most easily learnt, to those that are least accessible, found more rarely- in the world's languages, and learnt with more difficulty. The most often cited example is relative clauses; clauses based on a Subject relationship, e.g. "The man who came in is English', are more accessible than those based on, say, an Object-of-Comparison relationship, e.g. "The boy that I am fitter than is leaving".
Eckmann (1977) argues that a comparison of the target and mother languages predicts that learners should find the most difficulty with those aspects of the L2 that are more marked in terms of accessibility than the L1. Cook (1975) found certain similarities to the Accessibility Hierarchy in L1 and L2 learners of English, as did Gass (1979), although some differences emerged. The acquisition of relative clauses has also been studied within the framework of the present theory by Liceras (1981) in terms of the markedness of 'filters', and by Flynn (1983) in terms of the 'right branching principle'.
It is unclear precisely how the Accessibility Hierarchy is to be handled within the present theory; it could be interpreted as a continuum from unmarked core grammar to marked peripheral grammar. The evidence of Eckmann (1977) and Gass (1979) none the less suggests that the L2 learner operates with a concept of markedness based on closeness to the principles of Universal Grammar. White has developed this notion in terms of the relative markedness of different settings of a parameter (White 1983). Her example is the comparative restrictiveness of movement rules in English compared to French (technically S is a bounding category in English, but not French); the English setting for the parameter is less marked. Consequently French learners of English are likely to have particular problems with movement in English, since they are moving from a language with a more marked setting to one with a less marked. It should be pointed out, however, that markedness is used in other senses to the one found here, to refer to grammatical complexity, for example, as in some of the sources cited by Rutherford (1982), or to preferences for particular meanings of words, as in Kellerman (1979). Also the present theory does not assume that markedness is directly reflected in order of development, even if this additional assumption is made by many first and second language researchers. Though there is some plausibility in feeling that 'natural' unmarked forms should be learnt before those that are 'unnatural' and marked, features of channel capacity, etc., distort the sequence. However crucial to acquisition, the actual sequence of development disguises markedness in many ways.
The theory has clear implications for the notion of 'interlanguage' in L2 learning research (Selinker 1972)—the assumption that the L2 learner has a grammar of his own that is systematic in its own terms and that is distinct from both the first and the second languages. We have seen earlier that one interpretation of the theory is that not only the final grammar of competence is governed by Universal Grammar, but also all the interim grammars that the child goes through en route to adult competence (White 1982). Consequently, interlanguage as a human language must fit in with Universal Grammar. The studies by Schmidt (1980) and Ritchie (1978) cited earlier show that interlanguages reflect principles of surface order and of movement, and hence demonstrate their subjection to Universal Grammar.
Interlanguages should be considered no more deviant than ordinary grammars; they too are based on the properties of the human mind. Error Analysis based on the interlanguage hypothesis has therefore two new factors to take into account: one that interlanguages incorporate universal principles, the other that errors may be the result of channel capacity rather than of acquisition per se. One conceptual problem, however, is that L2 learning is seldom complete, in that few learners ever approximate to native competence; all their grammars are interlanguages. Hence the instantaneous acquisition model is difficult to apply, because there is no settled final competence, no 'steady state' grammar.
The concept of L2 learning as hypothesis testing is fundamentally affected by the theory. It has often been suggested that L2 learning is a process in which the learner creates interim guesses about the language which he tries out to see whether they are right or wrong and reformulates them if necessary (Cook 1969); Ellis, for instance, claims 'The principal tenet of IL theory, that the learner constructs for himself a series of hypotheses about the grammar of the language and consciously or unconsciously tests these out in formal or informal learning contexts, has withstood the test of both speculation and considerable empirical research' (Ellis 1982). The arguments against hypothesis testing that were raised earlier are equally true of L2 learning. In the natural environment, or a classroom that simulates a natural environment, the learner encounters only positive evidence, and does not get enough negative evidence to confirm or disconfirm his hypotheses.
'Natural' L2 learning cannot therefore consist of hypothesis testing in the sense in which hypotheses are checked against external feedback; the same argument holds. On the other hand, the learner in a classroom or other artificial setting may receive direct negative evidence or have access to explanatory data; he might therefore receive enough non-primary evidence for hypothesis testing to be feasible. But this leads to an odd paradox: hypothesis testing by feedback has usually been claimed to be the 'natural' informal way of learning a second language, the provision of correction and explanation an 'unnatural' formal way. Hence L2 learning research has to be cautious in its support of hypothesis testing. It is acceptable only in the sense that the learner checks positive evidence against the limited set of hypotheses provided by his Universal Grammar.
The role of Contrastive Analysis is also very different in the theory. Rather than being compared directly, two languages may be compared indirectly through the ways in which they embody the same linguistic principle while fixing parameters differently. Thus English is related to German through the slightly different ways in which it fixes sentence order. Structural comparison is a matter not of actual rules, but of the way in which the rules exploit the same underlying resources. The concept of core and periphery implies that this type of comparison must be supplemented by an account of how the two grammars deviate from core grammar for whatever reason. Thus, while English and French can be found to be similar in terms of the core parameter of sentence order, the account of their relationship also needs to take in the more peripheral rule that auxiliaries precede the Subject in certain types of question. At the core, the theory provides a common measuring stick for two grammars; as we move to the periphery, the stick becomes less appropriate and more attention has to be paid to other factors than Universal Grammar. A further relevant point is the distinction between development and acquisition; classical CA compared the two final steady-state grammars, i.e. was about acquisition. As Zobl has suggested, it may be fruitful to relate CA instead to development; 'it is paramount that the role of prior LI knowledge be conceptualised as a variable which may introduce variation into a developmental sequence' (Zobl 1982).
To sum up, the hypothetical picture of L2 learning that emerges is that the learner contributes a set of language principles and unfixed parameters; the evidence he encounters enables him to fix the parameters into a new grammar. While his first language affects his acquisition, it cannot help him acquire those parts of grammar that vary from one language to another. He also encounters evidence that does not fit Universal Grammar, for which he has to adopt more marked solutions. His environment, though different from the LI child's in some respects and subject to greater variation, does not provide him with any way out of the problem of the poverty of the stimulus. Because of his greater maturity, he does not have the same restrictions as the native child; in terms of cognitive level, but not of channel capacity, his development shows acquisition more closely than first language acquisition. How could this position be shown to be correct? One simple test is to see if the L2 learner knows rules he could not have learnt from the environment and that could not have been mediated through the first language, such as 'eager/easy to please'. Another is to show that interlanguages always reflect Universal Grammar, as Ritchie and Schmidt suggest (Schmidt 1980; Ritchie 1978). A third is to see whether L2 development shows characteristic differences from LI development so far as the absence of maturation is concerned, as argued in Gass and Ard (1980). But far more detailed and wide-ranging research is needed to show that there is real substance to this picture. Even if it is rejected, L2 learning research still has to defend its use of the concepts of hypothesis testing, sequence of development, and interlanguage, which are no longer compatible with the theory of Universal Grammar, either by severing its links with first language acquisition theory or by rethinking its ideas accordingly.
At the moment a long and treacherous route connects the theory with language teaching. On the negative side it removes some of the justifications for language teaching techniques claimed to be derived from earlier versions of the theory. It has often been suggested that students should be actively encouraged to try out their interim hypotheses in teaching situations so that they can use feedback to determine whether they are right (Cook 1969). Communication games, for example, have been justified on the grounds that they put the learner in a communicative situation where he gets instant feedback. Allwright (1977) argues that 'The success or failure of successive attempts to communicate in such tasks provides automatically G2 and G3 ("cues" and simple knowledge of results) from which the learners can infer the characteristics of the target language.' Such teaching techniques are not supported by the interpretation of hypothesis testing put forward here, even if they are desirable for other reasons. The new version also is not sympathetic towards the primacy of communication in language learning, one of the tenets of communicative methodology. In many ways the current theory seems closer to the humanistic trend in language teaching, with its emphasis on the value of the foreign language to the student's multi-faceted personal growth as described in Stevick (1980). The role attributed to the environment is also very different from that assumed in recent language teaching which has emphasized the importance of the language the students hear, whether in terms of the syllabus, the types of activity carried out in the classroom, or the provision of meaningful input (Krashen 1982); in the theory the environment only provides triggers.
Language teaching too might try exploring the possibility of providing triggering evidence. A while ago Newmark and Reibel argued that the learner should be allowed to apply the language principles in his mind without interference from the teacher (Newmark and Reibel 1969). Chomsky himself wrote that 'we should probably try to create a rich linguistic environment for the intuitive heuristics that the normal human being automatically possesses' (Chomsky 1968). The discovery of some, if not all, of the principles and parameters of Universal Grammar means modifying the Newmark and Reibel argument, since, if we are no longer ignorant, we could provide appropriate triggers for them to function; the learner's task might be expedited by meeting the right evidence at the right moment. As the sequence of L2 development depends partly upon channel capacity, the evidence must show how the parameters are fixed and how they may be encountered at the time when such factors as STM capacity are able to cope. Keith Nelson has described an approach for accelerating first language acquisition based on 'rare event' learning (Nelson 1982); first he assesses whether a child is 'ready' to learn a particular structure, then he provides examples of it over a short period; triggering experience when the time is ripe teaches the child the structure. Similarly the prototype theory of categorization suggests that there are 'best' examples of categories; a robin is a 'better' example of a bird than an ostrich (Rosch 1977). An L2 learner might then be presented with best-example sentences for which he is 'ready', to hasten the operation of Universal Grammar. Perhaps, for instance, the L2 learner of English needs to hear enough sentences early on to fix in his mind the fact that English is an SVO language. Paradoxically a theory firmly based on the inherent powers of the mind can come full circle to the effects of the environment on learning; the existence of a particular parameter of variation triggered by the environment suggests that the timing and nature of the trigger affect its acquisition, other things being equal.
So, to conclude, this paper has tried to explore' the relationship between the theory of Universal Grammar and L2 learning. Feyerabend has suggested that science should simultaneously explore several alternatives, rather than confining itself to a single dominant model at a time: 'pluralism of theories and of metaphysical views is not only important for methodology, it is also an essential part of a humanitarian outlook' (Feyerabend 1975: 52). It has not been argued here that Chomsky's theory of Universal Grammar is uniquely important for L2 learning, but that it is an alternative that applied linguists should evaluate for themselves rather than reject out of hand. A recent characteristic of applied linguistics has been its disassociation from contemporary theoretical linguistics; a bare handful of articles have attempted to relate the Chomskyan position to applied linguistics (Newmeyer 1982; Roca 1979; Sharwood Smith 1982). It would be dangerous if this attitude precluded the applied linguist from suspending his disbelief long enough to investigate what is happening in linguistics, even if after a closer look he decides it is not for him.
1. My thanks are due to the following for their extensive comments on earlier drafts of this paper: D. Arnold, N. Chomsky, F. Eckmann, C. James, D. Lightfoot, F. Newmeyer, I. Roca, T. Roeper, L. White, H. Zobl. Needless to say, the errors that remain are mine.
Allwright, R. 1977. 'Language learning through communication practice.' ELT Documents 77/1.
d'Anglejean, A. and G. R. Tucker. 1975. 'The acquisition of complex English structures by adult learners.' Language Learning 25/2.
Asher, J. T. and B. S. Price. 1967. 'The learning strategy of the total physical response: some age differences.' Child Development 38: 1219-27.
Baker, C. L. 1979. 'Syntactic theory and the projection problem.' Linguistic Inquiry 10/4: 533-81.
Baker, C. L. and J. J. McCarthy (eds.). 1981. The Logical Problem of Language Acquisition. Cambridge, Mass.: MIT Press.
Braine, M. D. S. 1971. 'On two types of models of the internalisation of grammars' in D. Slobin (ed.). The Ontogenesis of Grammar. New York: Academic Press.
Bresnan, D. and R. Kaplan. 1982. 'Introduction: grammars as representations of language' in D. Bresnan (ed.). The Mental Representations of Grammatical Relations. Cambridge, Mass.: MIT Press.
Brown, R. and C. Hanlon. 1970. 'Derivational complexity and the order of acquisition in child speech' in J. R. Hayes (ed.). Cognition and the Development of Language. New York: John Wiley.
Burt, M. K. and H. C. Dulay. 1980. 'On acquisition orders' in S. Felix (ed.). Second Language Development: Trends and Issues. Tubingen: Gunter Narr.
Carroll, J. B. 1965. 'The contributions of psychological theory and educational research to the teaching of foreign languages.' Modern Language Journal 49/5.
Chomsky, N. 1965. Aspects of the Theory of Syntax. Cambridge, Mass.: MIT Press.
Chomsky, N. 1968. 'Noam Chomsky and Stuart Hampshire discuss the study of language.' The Listener 79/2044: 687-91.
Chomsky, N. 1969. 'Linguistics and philosophy' in S. Hook (ed.). Language and Philosophy. New York: New York University Press.
Chomsky, N. 1976. Reflections on Language. London: Temple Smith.
Chomsky, N. 1980. Rules and Representations. Oxford: Basil Blackwell.
Chomsky, N. 1981a. Lectures on Government and Binding. Dordrecht: Foris.
Chomsky, N. 1981b. 'Principles and parameters in syntactic theory' in Hornstein and Lightfoot 1981b.
Chomsky, N. 1982. Some Concepts and Consequences of the Theory of Government and Binding. Cambridge, Mass.: MIT Press.
Cook, V. J. 1969. 'The analogy between first and second language learning.' International Review of Applied Linguistics VII/3.
Cook, V. J. 1973. 'The comparison of language development in native children and foreign adults.' International Review of Applied Linguistics XI/1.
Cook, V. J. 1975. 'Strategies in the comprehension of relative clauses.' Language and Speech 18/3: 204-12.
Cook, V. J. 1977. 'Cognitive processes in second language learning.' International Review of Applied Linguistics' XV/1: 1-20.
Cook, V. J. 1981. 'Some uses for second language learning research' in H. Winitz (ed.). Native Language and Foreign Language Acquisition. Annals of the New York Academy of Sciences 379: 251-8.
Cromer, R. F. 1970. ' "Children are nice to understand": surface structure clues to the recovery of a deep structure.' British Journal of Psychology 61.
Eckmann, F. R. 1977. 'Markedness and the contrastive analysis hypothesis.' Language Learning 27/2: 315-30.
Ekstrand, L. 1976. 'Age and length of residence as variables related to the adjustment of migrant children with special reference to second language learning' in G. Nickel (ed.). Proceedings of the Fourth International Congress of Applied Linguistics. Stuttgart: Hochschulverlag.
Ellis, R. 1982. 'The origins of interlanguage.' Applied Linguistics III/3: 207-23.
Feyerabend, P. 1975. Against Method. London: Verso.
Flynn, S. 1983. 'Similarities and differences between first and second language acquisition: setting the parameters of Universal Grammar' in D. R. Rogers & J. A. Sloboda (eds.). Acquisition of Symbolic Skills. New York: Plenum.
Gass, S. 1979. 'Language transfer and universal grammatical relations.' Language Learning 29/2: 327-44.
Gass, S. and J. Aid. 1980. 'L2 data: their relevance for language universals.' TESOL Quarterly XIV/3: 443-52.
Goodluck, H. and L. Solan. 1978. Papers in the Structure and Development of Child Language. University of Massachusetts Occasional Papers in Linguistics, 4.
Goodwin, B. C. 1976. Analytical Physiology of Cells and Developing Organisms. New York: Academic Press.
Green, J. 1976. 'How free is word order in Spanish?' in M. Harris (ed.). Romance Syntax: Synchronic and Diachronic Perspectives. Salford: University of Salford Press.
Hornstein, N. and D. Lightfoot. 1981a. 'Introduction' in Hornstein and Lightfoot 1981b.
Hornstein, N. and D. Lightfoot. 1981b. Explanations in Linguistics. London: Longman.
Keenan, E. 1972. 'On semantically based grammar.' Linguistic Inquiry, 3.
Kellerman, E. 1979. 'La difficulté, une notion difficile.' Encrages IV: 16-21.
Krashen, S. 1981. Second Language Acquisition and Second Language Learning. Oxford: Pergamon.
Krashen, S. 1982. Principles and Practice in Second Language Acquisition. Oxford: Pergamon.
Lecours, A. R. 1975. 'Myelogenetic correlates of the development of speech and language' in E. H. Lenneberg and E. Lenneberg (eds.). Foundations of Language Development. New York: Academic Press.
Lenneberg, E. H. 1967. Biological Foundations of Language. New York: John Wiley.
Liceras, J. M. 1981. 'Markedness and permeability in interlanguage systems.' Toronto Working Papers in Linguistics V/2.
Lightfoot, D. 1982. The Language Lottery: Toward a Biology of Grammars. Cambridge, Mass.: MIT Press.
McLaughlin, B. 1978. Second Language Acquisition in Childhood. Hillsdale, N.J.: Lawrence Erlbaum Associates.
McNeill, D. 1970. 'The capacity for the ontogenesis of grammar' in D. Slobin (ed.). The Ontogenesis of Grammar. New York: Academic Press.
Nelson, K. 1982. 'Toward a rare-event cognitive comparison theory of syntax acquisition' in P. Dale and D. Ingram (eds.). Child Language: An International Perspective. University Park Press.
Newmark, L. and D. A. Reibel. 1969. 'Necessity and sufficiency in language learning.' International Review of Applied Linguistics VI/3.
Newmeyer, F. 1982. 'On the applicability of transformational generative grammar.' Applied Linguistics III/2: 89-120.
Paivio, A. and I. Begg. 1981. Psychology of Language. New York: Prentice Hall.
Pinker, S. 1981. 'Comments on "Some Issues in the Theory of Language"' in Baker and Kaplan 1981.
Ritchie, W. 1978. 'The right-roof constraint in an adult acquired language' in W. Ritchie (ed.). Second Language Acquisition Research: Issues and Implications. New York: Academic Press.
Roca, I. M. 1979. 'Language learning and the Chomskyan revolution.' Studia Anglia Posnaniensa X: 141-209.
Rosch, E. 1977. 'Human categorisation' in N. Warren (ed.). Studies in Crosscultural Psychology. New York: Academic Press.
Rose, D. 1980. 'Some functional correlates of the maturation of neural systems' in D. Caplan (ed.). Biological Studies of Mental Processes. Cambridge, Mass.: MIT Press.
Rutherford, W. 1982. 'Markedness in second language acquisition.' Language Learning 32: 85-108.
Schmidt, M. 1980. 'Coordinate structure and language universals in interlanguage.' Language Learning 30/2: 397-416.
Selinker, L. 1972. 'Interlanguage.' International Review of Applied Linguistics X/3.
Sinclair-de-Zwart, H. 1969. 'Developmental psycholinguistics' in D. Elkind and R. Flavell (eds.). Studies in Cognitive Development: Essays in Honor of Jean Piaget. New York: Oxford University Press.
Sharwood Smith, M. 1982. 'Learnability and Second Language Acquisition.' Unpublished MS.
Stevick, E. 1980. A Way and Ways. Rowley, Mass.: Newbury House.
Tavakolian, S. L. 1981. (ed.). Language Acquisition and Linguistic Theory. Cambridge, Mass.: MIT Press.
White, S. L. 1981. 'The responsibility of grammatical theory to acquisitional data' in Hornstein and Lightfoot 1981b.
White, L. 1983. 'Markedness and parameter setting: some implications for a theory of adult second language acquisition.' To appear in F. Eckmann, E. Moravcsik, and J. Wirth (eds.). Proceedings of the Twelfth Annual University of Wisconsin-Milwaukee Symposium on Markedness. New York: Plenum.
Zobl, H. 1982. 'A direction for contrastive analysis: the comparative study of developmental sequences.' TESOL Quarterly 16/2: 169-83.