The metaphor of access to Universal Grammar in L2 learning
V.J. Cook, University of Essex
1994 paper in N. Ellis (ed.) Implicit Learning and Language
More up-to-date UG in SLA paper
Summary
The Universal Grammar theory claims that first language acquisition is based on implicit learning since all that is necessary is the provision of appropriate input to the child in the form of actual sentences of the language. One metaphor sees this input as processed by the Language Acquisition Device to yield the knowledge of language in the native speaker's mind. Such knowledge is nowadays expressed in terms of principles such as binding and parameters such as pro-drop. Second Language Acquisition research has tried to examine whether L2 grammars consist of principles and parameters and whether L2 learning has 'access' to the UG, the evidence being in favour of indirect access via the L1 or partial access. An alternative metaphor sees language acquisition as the initial state of the mind changing into the final state, denying separate status to the LAD and to UG. In this case there are not two 'products' of LAD in Second Language Acquisition but one: a changed state of the mind containing two grammars, called 'multi-competence'. Consequently the mind of the L2 user has to contain two values for a parameter simultaneously.
The metaphor of access to Universal Grammar in L2 learning
The current Universal Grammar (UG) model of language acquisition takes for granted the implicit nature of language learning. The language samples that the learner encounters are not organised to facilitate language learning in any useful way; the interaction with parents involves no relevant correction, feedback, or explanation; the child creates a grammar in the mind uninfluenced by any conscious attention to language learning. Virtually all children acquire a language, regardless of whether they encounter Greek or Chinese, regardless of whether their parents deliberately address them in 'simple' baby-talk or treat them as full conversational partners, and regardless of their personal characteristics. The sole requirement for successful acquisition is that there indeed be adequate samples of language from which the child can create the grammar. The child's mind responds to the data it encounters by creating a grammar of the language, automatically and painlessly.
The metaphor of the Language Acquisition Device
The process of acquisition was originally conceptualised by Chomsky (1964) in terms of a black box Language Acquisition Device (LAD) that takes in input consisting of sentences of the language and produces output in the form of the mental grammar of the language (linguistic competence). This is illustrated in Figure 1.
FIGURE ONE ABOUT HERE
If some aspect of the grammar present in the adult's mind cannot in principle be derived from the input the child encounters, its source can only be the Language Acquisition Device itself: it must be an innate feature of the mind. In an elaborated form, this became known as the poverty-of-the-stimulus argument or Plato's problem (Chomsky, 1986a): some aspects of language knowledge cannot be learnt from the environment.
Since the form of the output grammar had no necessary connection with the form of the LAD, this formulation did not provide an adequate technical apparatus for investigating the innate aspects of language. During the 1980s the development of the theory of syntax known as principles and parameters theory or Government/Binding (GB) theory (Chomsky, 1981; 1986b) started to bridge this gap between the grammar and the LAD. Knowledge of language was now seen as consisting of universal language principles unbroken in any grammar, and of parameters with a limited set of values; since these principles and parameters were found in the output but not in the input, they were part of the structure of the mind. As well, the mind knew the vocabulary of the language, in particular information about how each lexical item might be fitted into the structure of the sentence. In other words the grammar was now described in the same terms as the acquisition device. General introductions to this form of syntax can be found in Cook (1988b) and Haegeman (1991).
Principles and parameters theory
Let us start with some examples of principles to put some flesh on these bones:
- the binding principles constrain the ways in which pronouns can relate to antecedents: how "him" relates to "John" in "John likes him" (John not=him) versus how "him" relates to "himself" in "John likes himself" (John=himself), or, at a more complicated level, the divergent interpretations of "him" and "himself" in "John heard Peter's criticisms of him" (him=John) and "John heard Peter's criticisms of himself" (himself=Peter).
- the principle of subjacency restricts how far forms can be moved in the sentence. "Which book did she say that Peter was reading?" is a grammatical sentence but *"What did she say that Joe told her Bill's guess that Pete liked?" is ungrammatical, because "what" has crossed one 'barrier' too many in moving to the front of the second sentence.
The fact that a principle is not found in some language does not mean that it is not universal. The general principle of structure-dependency for example declares that all syntactic movement within the sentence depends on the sentence structure; English speakers know that "Is Sam the cat that is black?" is grammatical while *"Is Sam is the cat that black?" is not, because the latter violates the requirement that syntactic movement take heed of the structure of the sentence, in this case the division between main clause and subordinate clause. Japanese does not have syntactic movement (although it has other types of movement) so it has no need for structure-dependency for the formation of questions etc. But Japanese does not break the principle of structure-dependency; it is simply absent.
Within these principles there are certain possibilities of variation from one language to another known as parameters; each grammar known by the mind contains a particular set of values for the parameters. The learner acquires the parameter values for the language he or she is exposed to rather than the parameter itself. The language input enables the learner to find the setting for a particular language. A sentence such as "He didn't eat an apple" tells the learner:
- whether Complements precede or follow the heads of phrases - the head parameter: English has a head-initial setting so the Verb Phrase order is Verb + Object "likes beer", the Prepositional Phrase order is Preposition + Noun Phrase "in the morning", and so on; Japanese has a head-final setting so the Verb Phrase is Object + Verb "nihonjin desu" (Japanese am), the Preposition Phrase is Noun Phrase + Postposition "nihon ni" (Japan in), etc.
- whether the subject of the sentence has to be actually present or not - the pro-drop parameter: Spanish is a pro-drop language that may omit subjects as in "Va a casa" ([He]'s going home); English is a non-pro-drop language that does not permit the subject to be left out, e.g. "He's going home" rather than "Going home".
- whether negative elements and adverbs precede or follow the main verb - the opacity parameter; English "John often drinks wine" versus French "Jean boit souvent du vin".
And so on for other parameters.
The resultant knowledge of language consists of the principles as they apply to the language in question and of the settings for each parameter of variation. In addition it contains an extensive lexicon that describes how each and every lexical item fits into the constructions generated by the principles and parameters - whether a verb takes only a subject ("John sneezed"), a subject and a single object ("John lost his bag"), a subject and two objects ("John gave her a present"), a subject and an object consisting of a clause ("John thought he was wrong"), and so on. The concept of 'rule', once crucial to the theory, is now seen as derivative; a rule is an epiphenomenon produced by the interaction of principles, parameter settings, and vocabulary. Take a sentence such as "Peter shaved himself". A rule-based description would enumerate the rules that describe how Verb Phrases such as "shaved himself" consist of a Verb and an Object, how a reflexive such as "himself" refers to the same person as the Noun "Peter", and how the Subject "Peter" comes in front of the Verb Phrase "shaved himself". Each of these is an idiosyncratic rule for a particular construction and for a particular language. The principles and parameters description of "Peter shaved himself" shows how it incorporates the phrase structure principles common to all languages together with the head-initial setting for all phrases in English, how the binding principles true for all language make specific predictions for the co-reference of "Peter" and "himself", and how the Verb "shaved" requires a construction in which there is a Subject and an Object. Principles and parameters go beyond a rule-based description not only in providing general ideas that affect the whole grammar of a language rather than small fragments, but also in establishing a framework that is true for all languages rather than only one.
Chomskyan theory claims that, strictly speaking, the mind does not know languages but grammars; 'the notion "language" itself is derivative and relatively unimportant' (Chomsky, 1980, p.126). "The English language" or "the French language" means language as a social phenomenon - a collection of utterances. What the individual mind knows is not a language in this sense, but a grammar with the parameters set to particular values. Language is another epiphenomenon; the psychological reality is the grammar that a speaker knows, not a language. 'The grammar in a person's mind/brain is real; it is one of the real things in the world. The language (whatever that may be) is not' (Chomsky, 1982, p.5). Hence there is no necessity that the speaker should be speaking something that corresponds to one of the actual languages of the world such as French or English, only that the grammar conforms to the possible schemata laid down in Universal Grammar.
Universal Grammar and language acquisition
Reformulating UG as principles and parameters gave a new status to the original LAD suggestion. The form of grammars now had to be expressed in ways that would allow people to acquire them. Rather than LAD being distinct from the output grammars, the description of the grammar was integrated with acquisition. In principle any element in the grammar had to be justified in terms of acquisition. Acquisition of an L1 amounted to instantiating the principles of UG in a particular way and setting a number of parameter 'switches' to the values for the L1. The adaptation of the LAD model into UG theory is often represented in Figure 2, versions being found in, for example, Haegeman (1991, p.15) and White (1989, p.5):
FIGURE 2 ABOUT HERE
Given that the theory was preoccupied with the creation of language knowledge by the child's mind rather than with interaction with the environment, the usual test that was applied was whether something could have been learnt from 'positive' evidence of sentences that the child would actually have heard rather than from 'negative' evidence of parents' correction or of sentences that the child does not hear. Knowledge of language had to be acquirable solely from positive evidence; to rephrase the poverty-of-the-stimulus argument, things that were present in the grammar but not in the positive evidence must be part of the mind itself: language acquisition is implicit in that data trigger values for parameters that are already present in the mind - they set the switches a particular way. Principles such as structure-dependency or binding are not acquirable from the input as they could not be deduced from the observation of sentences. Parameters have to be expressed in such a way that they could be set by the child only from positive evidence; this has led to much discussion of other restrictions that needed to be placed on the child in terms of possible sequences of acquisition, such as the notion that the child always makes the minimal extrapolations from the data rather than the maximal, known as the Subset Principle (Wexler & Manzini, 1987).
Second Language Acquisition and Universal Grammar
During the 1980s, the principles and parameters model became a focus of attention in the field of Second Language Acquisition research, rather more so than in first language acquisition. It seemed that this model could be readily adapted to Second Language Acquisition, providing on the one hand clear-cut syntactic descriptions that would apply to any language, on the other more or less ready-made research questions. Overall, however, this proved less successful than hoped. So far as syntax is concerned, there are two main problems. One is that the theory of principles and parameters syntax is continually evolving; even the syntactic core of the grammar changes from year to year. In a rule-based model the effects of a change are localised to a single rule; in the interactive framework of the principles and parameters model, the slightest change affects the whole grammar. Whichever syntactic basis a researcher chooses is effectively out-of-date come publication day, making results and conclusions hard to interpret. The second problem is the methodology of research: what can count as data for knowledge of a second language? L1 acquisition starts from the single sentences accepted by native speakers on the deliberate assumption that there is a native speaker standard. Whatever the merits of idealisation to a normalised native speaker, this is less convincing in an L2, since there is no clear norm of what a successful L2 learner should look like, other than the monolingual; L2 learners vary extremely in the levels of language they attain while L1 children do not. L2 research can therefore be based with difficulty on the same kind of single sentence evidence used with L1 research. Transcripts of learners' speech are also difficult to use as evidence because of the very claims of the theory that ordinary speech does not contain signs of the phenomena with which it deals - structure-dependency violations, binding interpretation, etc. Experiments with comprehension can only be done with certain constructions, inevitably falling back onto construction-specific rules by virtue of the fact that they cannot deal with the whole range of effects caused by any principle or parameter. Most researchers therefore resort to grammaticality judgments elicited in experiments as their main source of data - a source of evidence that has to be treated with extreme caution as it is unclear how directly it taps the individual's knowledge of language; this is discussed more fully in Cook (1993).
The other major issue has been what to do with UG in Second Language Acquisition research - what are the questions that can be answered? The goals for linguistics listed by Chomsky (Chomsky, 1986a; Chomsky & Lasnik, in press) usually start by establishing the knowledge of language of the native speaker, discovering how this knowledge has been acquired, and seeing how it is used. I have argued elsewhere that this translates directly into the Second Language research goals of establishing knowledge of languages, acquisition of knowledge of languages, and knowledge of use of languages (Cook, 1993). The UG theory of Second Language Acquisition therefore should deal with how one mind has two grammars, how one mind learns two grammars, and how one mind uses two grammars. The research questions should relate to these overall goals.
The metaphor of access to Universal Grammar in Second Language Acquisition
However, the question that mostly concerned researchers in the late 1980s was a variant of the traditional issue of whether L2 learning is the same as L1 acquisition, rephrased as whether UG is 'accessible' in L2 learning (Cook, 1985); two collections have covered this area (Eubank, 1991; Gass & Schachter, 1989) as well as many individual articles. The possibilities are direct access, indirect access, and no access:
- direct access. In the direct access position, UG is as involved in L2 acquisition as in L1 learning. Principles are incorporated into L2 grammars in the same way that they are incorporated into L1 grammars: structure-dependency and binding are found in L2 grammars just as they are in L1 grammars. The values for parameters are set from positive evidence; the sequences for the acquisition of pro-drop or word order settings are the same in L2 learning as in L1 acquisition. In the direct access position, as shown in Figure 3, L2 acquisition adds a second language in parallel to the first:
FIGURE 3 ABOUT HERE
Like first language acquisition, L2 acquisition is an implicit process in the mind: learners are not aware of what they are learning and pay no conscious attention to it. They do not need anything other than positive evidence to set the values for parameters and to instantiate principles.
- indirect access. In the indirect access position, UG is not available to L2 acquisition except through the L1 grammar. The learner can make use of principles already instantiated in the first language but cannot use those that the L1 has not taken up. For instance, Japanese learners of English know the binding principles from Japanese so they can use them in English; the principle of structure-dependency is not, however, used in syntactic movement in Japanese questions so they should not be able to acquire it in English from knowledge of Japanese. Similarly L2 learners start off with the L1 value for each parameter rather than the value from which the L1 child starts. The same Japanese learners would start from the head-final setting for the head parameter and the pro-drop setting for the pro-drop parameter as these are the values in Japanese.
FIGURE 4 ABOUT HERE
In the indirect access position, positive evidence from outside is still the driving force in this process, which is, furthermore, restricted to those aspects of UG used in the L1. The learning model is still implicit: the learners are acquiring the L2 by hearing examples of language; the fact that this is mediated by an implicit knowledge of the L1 makes no difference. The environment has only a triggering function.
- no access. In the no access position, as illustrated in Figure 5, UG is unavailable to L2 acquisition. The grammar of the L2 is learnt by some other means than UG. It does not incorporate the principles or parameter settings of UG, since these are not derivable from input or from textbooks. The knowledge of the second language must derive from other faculties of the mind than language.
FIGURE 5 ABOUT HERE
Hence in the no access position L2 learning may be, not just an implicit process, but also an explicit one in which features of language are presented to learners and acquired by them through conscious attention, or in which teachers correct particular mistakes in the learners' speech, i.e. in which negative evidence has a role to play. Much of the traditional forms of language teaching are based on a no access view of L2 learning, whether they consist of explanation of grammatical rules, memorisation of vocabulary lists, translation, or any other methods that deliberately treat L2 teaching as 'unnatural' and employ teaching techniques that apply to any school subject rather than being specific to language. Some language teaching at any rate assumes L2 learning is entirely different from L1 learning in that it may be involve explicit reliance on conscious rules rather than implicit use of built-in principles and parameters.
Evidence on access to UG in Second Language Acquisition
Several lines of evidence and argument bear on this question of access to UG. Let us consider them under the two questions of whether L2 grammars conform to principles and parameters theory and whether they show L1 influence.
- are L2 grammars describable in terms of principles and parameters?
If the L2 grammar cannot be described in terms of principles and parameters, it would provide strong evidence for the no access to UG position. If the L2 grammar could, however, be described in principles and parameters terms, this knowledge would be no more derivable from the language input than in L1; hence it must come either from UG itself (the direct access position), or from the L1 (indirect access), even if the final syntactic description of the phenomenon is still unsure. It would seem straightforward to test the first of Chomsky's goals; can the L2 grammar be described in principles and parameters terms? Let us take some of the examples used so far.
First structure-dependency. Otsu and Naoi (1986), cited in White (1989), tested whether Japanese children learning English were aware of structure-dependency despite its absence from Japanese. 11 teenage Japanese schoolgirls who had just been taught English relative clauses were asked to change them into questions; ten of them succeeded without breaking structure-dependency. In other words they were immediately using a principle in the L2 that they had not encountered in the L1. My own current work with the MUGtest (Multi-purpose Universal Grammar test) uses grammaticality judgments to look at the grammars of, inter alia, advanced Japanese learners of English, at present 12 experienced teachers and 15 university students. Six of the sentences in this test involve structure-dependency violations, such as *"Is Sam is the cat that brown?" In terms of correct judgements two groups of English native speaker scored 100% and 99.2% on these sentences. The Japanese teachers scored 97.2%, the students 93.3%, compared to the natives 100%. Although they had never encountered this type of sentence before and although it embodied a principle not utilised in Japanese, advanced learners were able to spot its deviance with a high degree of accuracy. Provisional MUGtest results show that other groups also show a high rejection rate of these structure-dependency violation sentences, namely Chinese speakers 86.6%, and Finnish speakers 100%. At first sight, this provides straightforward evidence for the presence of structure-dependency in the grammars of all these L2 learners whether or not their first language has structure-dependency. Yet the picture is not so clear in other results from the MUGtest. Take grammatical sentences with relative clauses such as "Sam is the cat that is black". In terms of correctness the Japanese teachers scored only 70.8%, the students 75.5%, compared with the 96.6% and 95% scored by two native groups; similarly the grammatical questions with relative clauses such as "Is Sam the dog that is black?", accepted by the native groups at 90% and 91.6%, were only judged correct by 69.4% and 66.6% respectively, compared to a native 83%. So, while Japanese learners indeed spotted structure-dependency violations, they were less certain about both the simple relative clauses and the straightforward questions that the structure-dependency violation sentences seem to build on; their behaviours differ from natives, not on the structure-dependency violations, but on the grammatical sentences. This is also true of the Finns and Chinese tested in the MUGtest.
Secondly, let us look at the principle of subjacency that restricts how far movement may extend in a sentence. Subjacency applies only to movement. Hence it will not be needed in languages that do not have syntactic movement, such as Korean and Japanese. Do L2 learners with L1s without movement indeed know subjacency in the L2? If this were the case, not only would their grammars have principles and parameters but they would also have direct access independent of the L1. Bley-Vroman, Felix, and Ioup (1988) gave grammaticality judgments to 92 Korean advanced learners of English. Native speakers scored 96% correct on subjacency violations, Koreans 82.4%. While the Koreans clearly did less well than the natives, they nevertheless scored considerably above chance. Similarly Johnson & Newport (1991) found 23 Chinese learners of English scored notably below native levels on subjacency sentences, even if their scores were above chance. Again the MUGtest confirms that, with one type of subjacency violation (*"What does he wonder whether Sarah will play?"), the Japanese teachers were 97.2% correct, the students 77.7% - higher than the results for native groups who scored 75.5% and 71.6%; on another type of subjacency (*"What do the politicians know the fact that the voters like?") the teachers were 80.5% correct, the students 35.5%, compared to the natives' 82.2% and 88.3%. This might seem straightforward evidence for the existence of subjacency in the Japanese teachers' grammar, if not the students'. But other test sentences reveal a more complex picture; only 16.6% of the teachers accept grammatical sentences such as "What do the neighbours feel that their children like?" compared to native 84.4% and 85%, only 33.3% accept grammatical sentences such as "Who did he want to play?" compared to native 73.3% and 70%, rather different from the Johnson and Newport (1991) result that most of their Chinese learners could cope with the parallel declarative sentences and simple questions. So once again evidence that could be taken as striking confirmation of the existence of subjacency becomes fuzzy once other aspects of the learners' knowledge are taken into account. Schachter (1989) used a research design that tested whether knowledge of the related constructions correlated with knowledge of subjacency violations. Knowing the related structures did not mean the people tested correctly on subjacency. Schachter argues that the large number of learners (46%) falling into this category compared to those who knew both structures and principle indicates no access to UG. To many, the results of the subjacency research show the presence of subjacency in the L2 learners' grammars, but to a lesser extent than in that of natives. To some extent any use of subjacency by L2 learners would show the presence of UG, as it would have no other possible source; its variability may partly be explained by other aspects of L2 learning, such as the cognitive deficit in L2 processing.
Thirdly word order. Do L2 learners show the type of knowledge of word order covered by the principles and parameters theory? An interesting approach to word order was initiated by Clahsen and Muysken (1986). They claimed that L2 learners invent stages of word order that are ruled out by UG. Their evidence for this is the Second Language Acquisition of German. Children learning German as an L1 start by using an SOV order, regarded by many linguists as the underlying order for German despite the fact that German main clauses have Verb Second (i.e. the Verb always comes after an initial Subject, or Adverb); adults learning German as an L2 start by using SVO order. The consequence is that L1 children have to learn to move the Verb to get the SVO construction, L2 learners have to learn to move it to get the SOV construction; the L1 children gradually learn leftward movement of the Verb from an SOV base, the L2 learners rightward movement of the Verb from an SVO base. Clahsen and Muysken (1986) claim that 'by fixing on an initial assumption of SVO order, and then elaborating a series of complicated rules to patch up this hypothesis when confronted with conflicting data, the L2 learners are not only creating a rule system which is far more complicated than the native system, but also one which is not definable in linguistic theory' (Clahsen & Muysken, 1986, p.116). This is denied by duPlessis, Solin, Travis, and White (1987), who use an alternative syntactic analysis to show 'that the interlanguages of L2 learners fall within the range of grammars permitted by UG, rather than being unnatural' (duPlessis et al., 1987, p.74). Hulk (1991) also explored the L2 learning of French (SVO) by Dutch learners (SOV) and showed that these transferred their SOV order to French rather than starting from SVO as the Clahsen and Muysken research would predict. Hence it is not clear that the Clahsen and Muysken claim can stand either factually or theoretically.
To sum up these research areas, L2 learners' grammars seem within the bounds of principles and parameters theory, although there is a certain fuzziness to the results that suggests that this is not the full picture.
- do L2 grammars show L1 influence?
If principles are constant, the only type of L1 influence that arises is the absence of a principle in the L1 and its presence in the L2, as discussed in the last section. The issue is whether the values for a parameter are carried over to the L2. Put another way, can you tell from the knowledge of the second language which first language the speaker knows? The question presupposes that parameters are a valid way of describing L2 grammars; the answer to it distinguishes between the direct access position, which predicts L2 learners start from scratch, and the indirect access view, which predicts that they start from the L1 values.
Let us start with the pro-drop parameter, which has been the subject of a fair amount of L2 research. White (1986) tested two groups of intermediate L2 learners of English with different pro-drop settings in their L1s, namely French, a non-pro-drop language, and Spanish, a pro-drop language. The Spanish/Italian group were worse on the null subject sentences, that is to say, they wrongly accepted English sentences without subjects as grammatical (39% errors) more readily than the French group (19% errors). But this is a matter of percentages rather than an absolute difference: even the Spanish speakers were 61% correct. L2 learners therefore seem to start from their L1 setting rather than from a neutral position. Hilles (1986) found that the speech of Jorge - a 12-year-old Spanish-speaking near-beginner - supports the idea that 'pro-drop is present in early IL, but decreases over time' (Hilles, 1986, p.48). Turning to Spanish, Liceras (1989) tested 32 French and 30 English-speaking L2 learners of Spanish. No effective difference between the two groups of learners emerged; 'resetting the pro-drop parameter from English and French to Spanish is not difficult with respect to null subjects' (Liceras, 1989, p.126). More recent L2 work has dealt with a later explanation of pro-drop called the Morphological Uniformity Principle (Jaeggli & Hyams, 1988). Unfortunately, using the same data from the same learner called Marta, Lakshmanan (1991) and Hilles (1991) come to opposite conclusions about the relevance of the principle; and so it is not as yet possible to evaluate its success. The MUGtest was used with the same Japanese learners, Japanese being a pro-drop language. The teachers wrongly accepted 39% of null subject sentences such as *"Is French", the students 27%, compared to the native groups' scores of 90% and 73.3%. Although these were advanced learners of English, they were still accepting null subject sentences; some English natives are also fairly tolerant of these null subject sentences. All of these therefore suggest that the L1 pro-drop setting makes a difference to the L2 learner's knowledge of a non-pro-drop language like English; the signs from work such as Liceras (1989) are that L1 setting does not make a difference to the learning of a pro-drop language such as Spanish.
A further example is the Governing Category Parameter within the binding principles. The domain within which elements such as "him" or "himself" can be bound to Noun Phrases varies from one language to another; five settings are needed to cover variation between languages. Cook (1990) tested comprehension of English sentences with different binding domains on subjects with three L1s with different parameter settings - Romance languages, Japanese, and Norwegian. The error rates and timings of the groups were not significantly different except in two sentence types; the Japanese had the highest rate overall, the Norwegians the lowest; speakers from all backgrounds, including natives, show a similar order of difficulty, as has already been seen. The effects of L1 setting are not prominent; the overall high error rate for the Japanese and lower error rate for the Romance speakers might be attributed to their setting. Yet the overall similarities in difficulty order for all groups show no effects of setting. An experiment with a similar design is reported in Finer and Broselow (1985) with Korean learners of English. Finer and Broselow conclude that the learners have in a sense 'split the difference' between English and Korean, ending up with an intermediary setting between English and Korean rather than the setting for English or that for Korean. The L1 setting is a starting point for the learner who gradually moves to the L2. As Finer and Broselow (1985) point out, this demonstrates that L2 grammars can be described in principles and parameters terms, i.e. that the learners' grammars are within the possible parameter settings and consequently that the L1 setting is important to L2 learning.
A third area is the head-direction parameter which captures a further broad division between languages over where they locate subordinate clauses within the structure of the sentence. In English, subordinate clauses such as relative clauses typically occur after main clauses (postposed), as in "I liked the meal he cooked", while the order in Japanese is the reverse (preposed), as in "Gohan-o tabete-iru ko-ga naite-imasu" Rice-obj. eating is child-subj. crying is (the child who is eating rice is crying). In itself this difference between English and Japanese may account for the problems that Japanese speakers had with relative clauses in English which we saw earlier. But it also has consequences for the direction of 'anaphora' in the sentence; English has forward anaphora relating "John" and "he" from left to right in:
"John locked the door when he left"
and backward anaphora relating "Kate" and "she" from right to left in:
"When she left, Kate locked the door".
But English speakers find it harder to process sentences in which there is a relationship between a Noun and a missing anaphoric item to the left:
"Helen and George escaped"
[Helen 0 and George escaped] i.e. "Helen (escaped) and George
escaped"
rather than to a missing element to the right:
"Jane lost her temper and left."
[Jane lost her temper and 0 left] i.e. "Jane lost her temper and
(Jane) left"
Japanese speakers are the opposite.
Flynn (1987) reported an extensive investigation of L2 learning within this framework. Learning an L2 with the same setting for the head-direction parameter should be easier than learning an L2 with a different setting. She took 51 adult learners of English, whose L1 was Spanish, like English a postposed language, and 53 whose L1 was Japanese, a preposed language; the learners were divided into three groups by the Michigan Placement Test. Spanish speakers had higher scores than Japanese learners on sentences both with and without anaphora, were far better at both forward and backward anaphora, and found forward and backward anaphors different in difficulty; while Japanese did not. It was not that the Japanese were worse at English, for this difference held, level for level, across the groups; the Advanced Spanish learners for example had 57% correct for no anaphor sentences, the Japanese 24%. The setting for the head-direction parameter of the L1 affected the difficulty of both sentence types for L2 learners, making them comparatively easy for Spanish-speaking learners. Though this research has been fiercely criticised (e.g. Bley-Vroman & Chaudron, 1990), and its connection with the type of principles and parameters model described here is far from straightforward, nevertheless it suggests that L1 values of parameters are initially used in L2 grammars.
Let us turn to the head parameter that governs the choice of head-final phrases, as in Japanese, or head-initial phrases, as in English. The analysis of this parameter has changed over the past few years by decomposing it into separate parameters to take account of a wider range of languages. In Cook (1988a) 409 English secondary school children were taught eight Micro Artificial Languages (MALs) that varied in terms of order of Subject, Verb, and Object, and in terms of whether Prepositions or Adjectives came before or after their heads. Most learners (340) were highly consistent in acquiring word order, and most (321) succeeded in learning the MALs to criterion (12 out of 15). This suggested that their minds must share a word order parameter that is easily reset for position of head and complements after comparatively minimal exposure to a language (36 sentences), i.e. this supported some form of access to UG. The experiment also tested whether learners 'extrapolated' the parameter setting to other phrases they had not encountered; rather than extrapolating the position of the preposition in the Prepositional Phrase from the position of the object in the Verb Phrase as predicted, the children showed a preference for postpositions coming after the NP, regardless of object position. One interpretation of this is that the learners have reacted to the parameter setting for English by opting for its opposite value, support for indirect access.
So, all in all, the research suggests that there is indeed an initial carryover of parameter values from L1 to L2, varying to some extent according to the area of syntax involved, most notably with pro-drop, least with binding. The fairest position on access is not to adopt an either/or approach but to combine the positions into one complex model that can be called the compound position. This is shown in Figure 6.
FIGURE 6 ABOUT HERE
The compound model includes three routes from input to the L2 grammar. The direct access route applies to principles such as subjacency which seem to occur in L2 grammars, regardless of L1 and despite not being learnable from the input. The indirect influence route applies to parameters such as pro-drop where the L1 setting makes a difference to the L2 grammar even at advanced stages. The other influences route takes care both of the possibility that some aspects of L2 grammars are learnable by means such as grammatical explanation and of the cases where L2 seems to incorporate elements that breach the principles and parameters model, say the word order phenomena found in L2 acquisition of German. In the present state of research, some such compound model is required. Further research may indicate which aspects of language belong to which route, or whether in fact one or more of the routes may be eliminated. Clearly the direct access route takes an implicit view of learning as it relies on sources within the mind of which people are largely unaware and uses data from outside simply as a way of triggering settings; it is the same as the L1 version. The indirect influence route is also largely implicit in that differences between the L1 and the L2 in terms of principles and parameters are not usually available to the L2 learner's conscious attention; the evidence for changing parameter setting to the value for the new language is not likely to be explicitly made, certainly until recently, as few natives or language teachers are even aware of it. Nevertheless the widespread provision of 'explanatory evidence' (Cook, 1985) in the classroom does not rule this out as a potentially explicit way of triggering parameter settings. Aspects of language acquired under other influences may undoubtedly be explicit; these may include a diversity of methods of language teaching. For example the traditional method of language teaching through grammatical explanation has often been derided in recent years; yet it has been a conspicuous success with a small minority of elite academically gifted students in European settings. It would never have worked with anybody if it had not had some means of affecting the student's L2 learning.
The metaphor of changing states
Regardless of their position on access, all the figures presented have had a central box labelled 'UG' out of which comes a product labelled 'L1 grammar'; in all but the no access position, the UG box also leads, directly or indirectly, to a product labelled 'L2 grammar'. The central metaphor of the model is still that of the black box with input and output, which may or may not be 'accessible' to Second Language Acquisition: UG is something separate from the grammar that is acquired.
But alongside this black box metaphor runs another one expressed in terms of a changing language faculty. The mind has several different faculties, one of which is language. The initial state of the language faculty, symbolised by S0, can be regarded as UG with all the principles and parameters present but unattached to any language; the final state is when UG has been transformed into one of its possible 'steady states', or SS. 'The language faculty has an initial state, genetically determined; in the normal course of development it passes through a series of states in early childhood, reaching a relatively stable steady state that undergoes little subsequent change, apart from the lexicon ... we call the theory of the state attained its grammar and the theory of the initial state universal grammar' (Chomsky & Lasnik, in press). Language acquisition is seen as the conversion of S0 into SS:
S0 --> SS
In this metaphor the language faculty itself changes with time; there is no separate UG with a product distinct from itself, but a UG that steadily transforms itself. The model should look like that in Figure 7.
FIGURE 7 ABOUT HERE
The grammar is a state of UG, not a product of UG. The language faculty has incorporated the values of the L1 within its built-in elements to become the grammar. To be adapted to Second Language Acquisition research, the states metaphor needs to accommodate the fact that the initial and final states of L2 learning are different from those in L1 acquisition (Cook, 1988b). The initial state for L2 learning is not zero because it already incorporates an L1; it can therefore be called Si. The final state for L2 learning is rarely equivalent to the Ss of a monolingual; another term is needed for this terminal state, St, i.e.:
Si --> St
The original LAD black box metaphor reifies UG as something separate from the grammar; the alternative 'states' metaphor sees the language faculty itself as the grammar. Both metaphoric models have been used alongside each other in the literature as if they were making the same claims. For instance Chomsky describes LAD as 'a procedure that operates on experience acquired in an ideal community and constructs from it, in a determinate way, a state of the language faculty' (Chomsky, 1990); LAD is distinct from UG but produces a product that is a changed state of itself.
The argument here is that a choice has to be made between these metaphors, certainly so far as Second Language Acquisition is concerned. For the question of access is meaningless if UG is not in fact separate from the grammar itself: it would be bizarre to talk of the language faculty having, or not having, access to itself; it is the same problem that MSDOS solves by warning "A file may not be copied onto itself". The question is how the L2 knowledge can be incorporated in the same language faculty as the L1 or whether indeed the language faculty itself is unavailable in the L2. Either the language faculty in some sense clones itself in an L2 or the same language faculty has to incorporate, at best all, at worst none, of the knowledge of the second language.
The evidence that L2 grammars may be described in principles and parameters terms can be looked at in this light. This suggests that L2 grammars, or some aspects of them, are indeed stored in the language faculty itself. Structure-dependency, subjacency, and word order principles arguably occur in L2 grammars, so L2 grammars need the same principles. Such phenomena are held to be characteristic of the language faculty rather than of other faculties of the mind. So, either a second language faculty is needed in the mind to store the L2, or they are part of the same language faculty. The differences that occur in the L2 forms of principles and parameters are due, not to lack of access, but to the problem of simultaneously storing two versions alongside each other; for a French/English speaker the way that structure-dependency is used in English has to be stored alongside the way that it is used in French. Some accommodation might be expected between the two forms of knowledge. Hence the knowledge of the L2 user is unlikely to be quite the same as that of a monolingual; it is not the lack of access to UG that is at stake, it is the possession of two versions of UG instantiated in the same mind. This I have termed multi-competence - the compound state of a mind with two grammars (Cook, 1991; 1992).
Let us return to the notion of language as epiphenomenon. What 'language' does the L2 learner speak in the L2? The interlanguage assumption, to which essentially all L2 research pays lip service, claims that the L2 learner's sentences are not, strictly speaking, sentences in the L2 or the L1, but are sentences in a temporary interlanguage that does not correspond to any external language (Nemser, 1971; Selinker, 1972); a French learner of English speaks neither English nor French when attempting to use English; the interlanguage grammar is neither L1 nor L2. The interlanguage assumption has similar implications to the epiphenomenal view of language in that the learner knows a grammar rather than a language. In 1953 Weinreich (1953, p.7) said that 'A structuralist theory of communication which distinguishes between speech and language ... necessarily assumes that "every speech event belongs to a definite language"'. This is precisely what interlanguage denies by refusing to class the learner's utterances as one 'language' or the other. People who speak two languages know grammars, not language. The problem is not how the mind acquires one or more languages but how it acquires one or more grammars.
Hence the kind of result found so often with parameter settings in the L2 should not be surprising: the L2 user behaves, not like a native speaker, nor like an ignoramus, but like somebody half-way between - the principles are there but not so completely, the parameter settings are there but not so firmly, the first language values still influence the learner, and so on. The cause is, not the process of acquisition with its access or lack of access to UG, but the state of knowledge of a mind that knows two languages - an aspect of the knowledge problem rather than the acquisition problem. When the UG of the language faculty becomes instantiated in two forms simultaneously, the L2 form is different from that in a monolingual L1 speaker. Work by Coppetiers (1987) indeed suggests that the L1 grammar of bilinguals is far from uninfluenced by the presence of the L2.
Parameter-switching
A further metaphor attached to the UG model has to be examined, namely the idea of switching parameter values. Typically the parameter is seen as having two possible states or values, say the pro-drop versus the non-pro-drop settings for the pro-drop parameter. The L1 child starts with one or other of the settings or is neutral between them; positive evidence allows him or her to adopt the setting for the circumambient language. So the L1 grammar acquires a single fixed setting. Hence L2 acquisition has to postulate two grammars, two products, as one grammar could not hold two settings for a parameter at the same time; a switch cannot be simultaneously on and off. The wholistic multi-competence view is that the mind indeed possesses both values for a parameter at the same time. The switching metaphor for acquisition turns out to be misleading too. It is not surprising that one finds claims, for example by Finer and Broselow (1985), that L2 learners have values between L1 and L2 for binding; nor is it remarkable that L2 users show that they do not fully have the values of the L2, as with the Japanese teachers and pro-drop; parameters are not permanently in an on or off position.
While this is here put in a context of use of two languages, something similar is needed to account for the knowledge of a single language. Speakers can in fact understand a variety of registers of their first languages, differing in terms of age, sex, region, class, formality, and other factors. To take pro-drop as an example, it is far from unusual to encounter null subject sentences in everyday life; "Can't buy me love" comes from a Beatles song, "Will complement the taste of meat and game" from a_Valpolicella label, "'Ridiculous', she said; 'Never did'; never will'" from a speaker denying she ever says "gonna" (Preston, 1989). It may indeed almost be obligatory to use null subjects in some registers - diary-writing for example, as in "Rush to bathroom", and "Next morning whilst in library queuing, am surprised by sudden case of standing-up colic", taken from the fictional diary kept by Dulcie Domum in the Guardian (Domum, 1992). While some of these examples may reflect performance factors rather than the grammar itself, the residue certainly suggests that English speakers produce and understand null subject sentences in certain registers and situations (Cook, 1990). It may be convenient to treat English as having a single non-pro-drop parameter setting, but the pro-drop alternative solution is lurking in the wings. On similar lines Stowell (1991) looked at the features of English headlines; he claimed that, while their peculiarities are different from the usual parameter settings for English, they are still within the bounds of UG. Even the description of the monolingual speaker may require more than one setting for each parameter rather than an on\off switch. This resembles the theory of polylectal grammars in dialectology (Bailey, 1973), which claims that the mind contains a single polylectal grammar that encompasses all the variants that it has to cope with in listening; language acquisition takes place 'in such a way as to create an underlying grammar which will generate all the variants that he must competently cope with' (Bailey, 1973, p.24): 'competence is polylectal' (Bailey, 1973). A similar approach is also found in the work of Wandruszka (1971, p.8); 'Mehrsprachig sind wir schon in unserer Muttersprache' (We are already multilingual in our mother tongue).
Linguistics has usually preferred variables with two choices, whether binary branching in phrase structure syntax or binary distinctive features in phonology. The pro-drop, opacity, head, and head-direction parameters have two values; the Governing Category Parameter within the binding principles is unusual in having five values. The states metaphor put forward here suggests that there are typically two alternative settings, one of which is chosen in an L2 in response to positive evidence, but the other of which is still present in the system, albeit with greatly diminished strength. This allows the speaker within the L1 to adjust the setting according to the register encountered - pro-drop say in reading a diary, non-pro-drop in other circumstances. Some such flexibility is needed in the L1 to avoid postulating the speaker has many L1 grammars, which are switched among according to the situation. In other words, the picture of the single grammar with fixed parameter settings is impossible to maintain even for the first language; either the speaker needs an indefinite number of grammars that are available according to circumstance, which seems, to say the least, uneconomic; or the speaker uses a single grammar with flexible setting. The states metaphor does not have an indefinite number of positions for a parameter - this would lose the whole advantage of the parameters formulation. Instead it is saying that settings have strengths rather than all-or-none values; each type of language situation increases or decreases the strength of the settings.
A second language is an extreme case of the kind of switching found in the L1. The stronger setting for the pro-drop parameter in English may indeed be non-pro-drop; but this has to give place to the weaker pro-drop setting in diary situations, etc; and it has to change to pro-drop as soon as the English speaker starts using Italian. While the parameter normally has binary settings, it is not rigidly tied to one setting but can alternate under certain circumstances. This must be distinguished from the process of language development in which the L2 research appears to show L2 learners gradually increasing the strength for a particular setting and never reaching the strength of the native, as the MUGtest example of structure-dependency shows.
Lexical and functional parameters
Finally the question of where the parameters and their settings belong. Here they have been conventionally seen as variables within the principles of syntax. But recent developments in principles and parameters grammar have tended to push the parameters away from the syntax towards the lexicon. Chomsky's succinct 1989 position was that 'there is only one human language apart from the lexicon, and language acquisition is in essence a matter of determining lexical idiosyncrasies' (Chomsky, 1989). Binding has already in a sense been handled in this way; the governing category parameter defines the domain, not for particular languages, but for particular lexical items such as "himself", or "zibun", or "seg"; there is no single overall setting for this parameter but a setting for each relevant item. The learners have to acquire a setting for each anaphor or pronominal they encounter. Hence, the acquisition of the appropriate settings for binding involves the learners acquiring the syntactic specification of lexical items rather than variables in the principles themselves. Perhaps parameters only belong to the lexicon; 'acquisition of a language reduces to selection of substantives from a given store and fixing of values of parameters that apply to functional elements and to properties of the lexicon as a whole' (Chomsky, 1990). Language acquisition therefore combines the two elements of parameter-settings and the lexicon; the rest of language does not need to be acquired as it consists of invariant principles.
Learning a second language does not affect the principles as these do not change from one language to another; hence Japanese learners are as bound by structure-dependency as are English native speakers and spot violations of principles with ease. It does affect the parametric aspects of language; hence Japanese learners are less certain about relative clauses and wh-questions. One of the odd results that has been consistently found in the research in this area is that L2 learners have a response bias towards spotting the ungrammaticality of sentences that break principles rather than spotting the grammaticality of those that conform, supported by our MUGtest results for structure-dependency and related grammatical structures. The reason might be that the Japanese are able to use UG principles for detecting clear violations; sentences that do not involve clear breaches have to be tested against the several parameter-settings for different lexical items that might obtain. The reason why multi-competence is possible is that it represents an enlarged lexicon, with many more lexical items specified in terms of parameters. In one sense this fits with modern ideas of the bilingual lexicon; Cook (1992) surveys various work on the bilingual lexicon, such as Cristoffanini, Kirsner, and Milech (1986) and Beauvillain and Grainger (1987), and concludes 'the evidence is mostly in favour of one interdependent store rather than separate stores'. The lexical parameter view of acquisition fits the view of multi-competence put forward here, although its full implications await the further development of the syntactic theory.
In general this argument suggests that Second Language Acquisition research approached UG from the wrong direction. The question should never be 'does the L2 learner's grammar conform to that of the target language?' but always 'what is the learner's L2 grammar?'. In terms of Keil (1981), we can investigate whether the learner's grammar is subject to 'structural change governed by constraints'. Hence the test for principles is always appropriate as it is testing whether the L2 grammar in fact conforms to the constraints on any human grammar, not just those in the target language. The testing for parameter settings is also appropriate in so far as it sees whether the L2 grammars conform to the bounds set by UG; but it is not appropriate if it tests exclusively for the way in which the L2 target language sets the parameters. This falls into the trap that the concept of interlanguage was intended to avoid - measuring the L2 grammar by the target rather than seeing it as a grammar in its own right. Tests are needed that reveal the parameter-settings of the learner's grammar, not the extent to which the learner's grammar fails to conform to the target. The fuzziness of much of the research mentioned earlier may be due to this inappropriate standard. Testing violations of principles is a valid test of whether L2 grammars conform to UG; testing settings of parameters reveals only whether the learner has adopted a setting that conforms to the L2, rather than revealing the unique nature of the L2 grammar. But, unless the principles and parameters framework and the concept of parameter setting itself are abandoned, the model belongs as firmly as ever within an implicit view of learning. Whatever may change, it is not the assumption that all that is necessary to acquire the core elements of human language is exposure of a mind containing principles and parameters to an input that contains positive evidence for the instantiation of principles and the setting of parameters.
References
Bailey, C-Y. (1973). Variation and Linguistic Theory. Arlington, Virginia: Center for Applied Linguistics.
Beauvillian, C., & Grainger, J. (1987). Accessing interlexical homographs: some limitations of a language-selective access. Journal of Memory and Language. 26, 658-672.
Bley-Vroman, R., & Chaudron, C. (1990). Second language processing of subordinate clauses and anaphora - first language and universal influences: a review of Flynn's research. Language Learning. 40, 245-285.
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. (1964). Current Issues in Linguistic Theory. The Hague: Mouton.
Chomsky, N. (1980). Rules and Representations. Oxford: Blackwell.
Chomsky, N. (1981). Lectures on Government and Binding. Dordrecht: Foris.
Chomsky, N. (1982). Some Concepts and Consequences of the Theory of Government and Binding. Cambridge, Massachusetts: MIT Press
Chomsky, N. (1986a). Knowledge of Language: Its Nature, Origin and Use. New York: Praeger.
Chomsky, N. (1986b). Barriers. Cambridge, Massachusetts: M.I.T. Press.
Chomsky, N. (1989). Some notes on economy of derivation and representation. MIT Working Papers in Linguistics. 10, 43-74.
Chomsky, N. (1990). Language and Mind. In D.H. Mellor (Ed.), Ways of Communicating. Cambridge: Cambridge University Press. 56-80
Chomsky, N., & Lasnik, H. (in press). Principles and parameters theory. In J. Jacobs, A. von Stechow, W. Sternefeld, W & T. Vennemann, (Eds.). Syntax: An International Handbook of Contemporary Research. Berlin: de Gruyter.
Clahsen, H., & Muysken, P. (1986). The availability of universal grammar to adult and child learners - a study of the acquisition of German word order. Second Language Research. 2(2), 93-119.
Cook, V.J. (1985). Chomsky's Universal Grammar and second language learning. Applied Linguistics. 6, 1-18.
Cook, V.J. (1988a). Language learners' extrapolation of word order in phrases of Micro-Artificial Languages. Language Learning. 38(4), 497-529.
Cook, V.J. (1988b). Chomsky's Universal Grammar: An Introduction. Oxford: Blackwell.
Cook, V.J. (1990). Observational evidence and the UG theory of language acquisition. In I. Roca (Ed.), Logical Issues in Language Acquisition. Dordrecht: Foris. 33-46.
Cook, V.J. (1991). The poverty-of-the-stimulus argument and multi-competence. Second Language Research. 7(2), 103-117.
Cook, V.J. (1992). Evidence for multi-competence. Language Learning. 42(4), 557-591.
Cook, V.J. (1993). Linguistics and Second Language Acquisition. Basingstoke: Macmillan.
Coppetiers, R. (1987). Competence differences between native and near-native speakers. Language. 63(3), 545-573.
Cristoffanini, P., Kirsner, K., & Milech, D. (1986). Bilingual lexical representation; the status of Spanish-English cognates. Quarterly Journal of Experimental Psychology. 38A, 367-393.
Domum, D. (1992). Bad Housekeeping. The Guardian. May 9th.
duPlessis, J., Solin, D., Travis, L., & White, L. (1987). UG or not UG, that is the question: a reply to Clahsen and Muysken. Second Language Research. 3(1), 56-75.
Eubank, L. (ed.) (1991). Point Counterpoint: Universal Grammar in the Second Language. Amsterdam: Benjamins. 411-418.
Finer, D., & Broselow, E. (1985). Second language acquisition of reflexive binding. North-East Linguistic Society. 16.
Flynn, S. (1987). A Parameter-setting Model of L2 Acquisition. Dordrecht: Reidel.
Gass, S., & Schachter, J. (Eds.) (1989). Linguistic Perspectives on Second Language Acquisition. Cambridge: Cambridge University Press.
Haegeman, L. (1991). Introduction to Government and Binding Theory. Oxford: Blackwell.
Hilles, S. (1986). Interlanguage and the pro-drop parameter. Second Language Research. 2(1), 33-52.
Hilles, S. (1991). Access to Universal Grammar in Second Language Acquisition. In L. Eubank (Ed.). Point Counterpoint: Universal Grammar in the Second Language. Amsterdam: Benjamins. 305-338.
Hulk, A. (1991). Parameter setting and the acquisition of word order in L2 French. Second Language Research. 7(1), 1-34.
Jaeggli, O.A., & Hyams, N.M. (to appear). Morphological uniformity and the setting of the null subject parameter. North East Linguistics Society. 18
Johnson, J.S., & Newport, E.L. (1991). Critical period effects on universal properties of languages: the status of subjacency in the acquisition of a second language. Cognition. 39, 215-258.
Keil, F.C. (1981). Constraints on knowledge and linguistic development. Psychological Review. 88(3).
Lakshmanan, U. (1991). Morphological Uniformity and null subjects in child second language acquisition. In L. Eubank (ed.), Point Counterpoint: Universal Grammar in the Second Language. Amsterdam: Benjamins. 389-410.
Liceras, J.M. (1989). On some properties of the "pro-drop" parameter: looking for missing subjects in non-native Spanish. In S. Gass, & J. Schachter (Eds.). Linguistics Perspectives on Second Language Acquisition. Cambridge: Cambridge University Press. 109-133.
Nemser, W. 1971). Approximative systems of foreign language learners. International Review of Applied Linguistics. IX, 115-123. Reprinted in Richards (1974).
Otsu, Y., & Naoi, K. (1986). Structure-dependence in L2 acquisition. Paper presented at JACET, Keio University, Tokyo, September. Cited in White (1989).
Preston, D.R. (1989). Sociolinguistics and Second Language Acquisition. Oxford: Blackwell.
Richards, J.C. (ed.) (1974). Error Analysis. Harlow: Longman.
Schachter, J. (1989). Testing a proposed universal. In S. Gass & J. Schachter (Eds.). Linguistic Perspectives on Second Language Acquisition. Cambridge: Cambridge University Press. 73-88.
Selinker, L. (1972). Interlanguage. International Review of Applied Linguistics. X(3). Reprinted in Richards (1974).
Stowell, T. (1991). Empty heads in abbreviated English. Paper given at Generative Linguists of the Old World, Leiden, March.
Wandruszka, M. (1971). Interlinguistik: Umrisse einer neuen Sprachwissenschaft. Piper & Co: Munich
Weinreich, U. (1953). Languages in Contact. The Hague: Mouton.
Wexler, K., & Manzini, M.R. (1987). Parameters and learnability. In T. Roeper & E. Williams (Eds.), Parameters and Linguistic Theory. Dordrecht: Reidel.
White, L. (1986). Implications of parametric variation for adult second language acquisition: an investigation of the pro-drop parameter. In V.J. Cook (Ed.), Experimental Approaches to Second Language Acquisition. Oxford: Pergamon, 55-72.
White, L. (1989). Universal Grammar and Second Language Acquisition. Amsterdam: Benjamins.
_________________
° Language ° Output
Input ---> ° Acquisition ° ---> (a grammar of
(language data) ° Device ° a language)
°_______________°
Figure 1 The LAD model of L1 acquisition
___________________
° principles °
Input ---> ° UG ° ---> L1 grammar (consisting
° parameters ° of principles, parameter
°_________________° settings, vocabulary)
Figure 2 Universal Grammar model of language acquisition
__________________
L1 Input ---> ° principles ° ---> L1 grammar (principles,
° UG ° parameter settings, vocabulary)
° parameters °
L2 Input ---> °_________________° ---> L2 grammar (principles,
parameter settings, vocabulary)
The direct access position
________________
L1 input ---> ° principles°
°UG ° --> L1 grammar --|
L2 input --| ° parameters° | L2 grammar (L1
| °______________° |--> principles, L1
| | parameter
| | settings,
|------------------------------------| vocabulary)
The indirect access position
_________________
L1 input ----> ° principles °
°UG ° ---> L1 grammar (principles,
L2 input --| ° parameters ° parameters, vocabulary)
| °_______________° :
| _________________ :
| ° other mental ° :
|-> ° processes ° ---> L2 grammar (no UG principles,
°_______________° no parameter settings)
The no-access position
_________________
° ° ---------> L1 grammar (principles,
L1 Input ----> ° principles ° parameter settings,
° ° vocabulary)
°UG ° | indirect UG
° parameters ° direct | influence
L2 Input -|--> ° ° UG access V
| °_______________° ---------> L2 grammar (principles,
| parameter settings,
| _________________ |----> vocabulary)
| ° other mental ° |
| ° ° | other
|--> ° processes ° ----| influences
°_______________°
The compound model of UG in L2 acquisition
______________________________________
° principles (+ form in L1) °
° °
Input ---> ° UG parameters (+ settings in L1) °
° °
° vocabulary (+ L1 lexical items°
°____________________________________°
Alternative 'states' model of language acquisition