• No results found

Linguistics and language learning: the Universal

3.5 Universal Grammar and second language acquisition

3.5.4 Empirical evidence

After having illustrated, in the context of second language acquisition, how to apply a Universal Grammar framework, taking the example of one prin-ciple (structure-dependency) and of two parameters (head-direction and strength of Infl), we can now turn to the reassessment of the theoretical positions we outlined in Section 3.5.1.

As we have already mentioned before, which aspects of Universal Grammar might be available and which not, is the subject of much debate.

The various theoretical positions have to attempt to reconcile somewhat contradictory facts about the second language acquisition process:

• Learners do not seem to produce 'wild' grammars, that is, grammars that would not be constrained by Universal Grammar. Does that suggest that at least principles of Universal Grammar are available to them?

• Learners produce grammars that are not necessarily like either their first language or their second language. Does this suggest that parameter set-tings other than those realized in their first or second languages are avail-able to them?

• Some principles and parameters seem to be unproblematic to reset (e.g.

the head parameter), others more difficult, or even impossible (e.g. sub-jacency).Why?

3.5.4.1 Hypothesis 1: no access to Universal Grammar

The view that Universal Grammar is no longer available to second language learners is still very much alive. Proponents of this position argue that there is a 'critical period' for language acquisition during children's early devel-opment^ and that adult second language learners have to resort to other learning mechanisms. The reasons for adopting such a position are several (for a review, see Bley-Vroman, 1989), but perhaps the most convincing one is the commonsense observation that immigrant children generally become native-like speakers of their second language, whereas their parents rarely do. For example, an influential study (Johnson and Newport, 1989) found

a correlation between age of arrival in the USA and native-like judgements on a number of grammatical properties of English. Immigrants who had arrived in the States before the age of seven years performed in a native-like way, and the older learners were on arrival, the more errors they made in the test. The correlation was not equally strong for all grammatical proper-ties investigated, however, and some researchers who have critically evalu-ated their data have argued it does not mean that the adult grammars are not Universal Grammar-constrained (Hawkins, 2001; White, 2003).

Studies adopting this position tend to focus on differences between first and second language acquisition, and on differences in the end result of the acquisition process. For example, in an extensive study of the acquisition of negation in French and German by first and second language learners, Meisel (1997, p. 258) concludes, 'I would like to hypothesize that second language learners, rather than using structure-dependent operations constrained by U G , resort to linear sequencing strategies which apply to surface strings'. Meisel therefore claims that one of the most fundamental principles of Universal Grammar (struc-ture-dependency) is not available to second language learners any more.

It must be said, however, that most studies conducted within a genera-tive framework would argue very strongly that second language gram-mars are Universal Grammar-constrained.

3.5.4.2 Hypothesis 2: full access to Universal Grammar

Full access/no transfer: Flynn (1996) adopts this position. That is, she argues that Universal Grammar continues to underpin SLL, for adults as well as children, and that there is no such thing as a critical period after which Universal Grammar ceases to operate. If it can be shown that learn-ers can acquire principles and/or parameter settings of the second language, which differ from those of their first language, she claims, the best interpre-tation is the continuing operation of Universal Grammar. She goes on to review a range of empirical work with second language learners moving from a language such as Japanese to English (Flynn, 1996, pp. 134-48).

Thus, for example, we have already met her claim that adult Japanese learn-ers of English as a second language can successfully reset the head-direc-tion parameter (i.e. from head-last to head-first). She also claims that similar learners can instantiate principles that do not operate in Japanese, such as the Subjacency principle (which controls ^ - m o v e m e n t in English;

i.e. the way in which we move the w/z-phrase to the beginning of the sen-tence); and can acquire functional categories, supposedly non-existent in Japanese. Flynn concludes her review thus:

It appears that L2 learners do construct grammars of the new TLs [target languages] under the constraints imposed by UG; those principles of UG carefully investigated thus far indicate that those not instantiated or applying vacuously in the LI but operative in the L2, are in fact acquirable by the L2 learner.

We are thus forced to the conclusion that UG constrains L2 acquisition; the essential language faculty involved in LI acquisition is also involved in adult L2 acquisition.

(Flynn, 1996,pp. 150-1) Other researchers who believe that Universal Grammar is still available to second language learners include Thomas (1991), on the basis of work on the acquisition of reflexive bindings and White et al (1992), on the basis of work on w/z-movement as well.

Full transfer/full access: this model also believes that second language learners have full access to Universal Grammar principles and parameters, whether or not they are present in the learners' first language (Schwartz and Sprouse, 1994, 1996). But in this view, second language learners are thought to transfer all the parameter-settings from their first language in an initial stage, and subsequently to revise their hypotheses when the second language fails to conform to these first language settings. Learners then develop new hypotheses that are constrained by Universal Grammar. In this view, Universal Grammar is accessed via the first language in a first stage, and directly thereafter when the second language input cannot be accommodated within the first language settings. Studies which support the full transfer or full access hypothesis include Yuan (1998), Slabakova (2000) and Haznedar (2001); for a review of these studies, see White (2003).

Full access/impaired early representations: several researchers also believe that learners can reset parameters to the second language values, but that initially, learners are lacking functional categories altogether. The Minimal Trees approach (Vainikka andYoung-Scholten, 1996b, 1998) has been highly influential and forms the starting point for a number of recent accounts of the development of syntax {see below): only lexical categories are projected initially, which transfer from the first language. Functional cat-egories develop later, but are not transferred from the first language. A similar approach is that of Eubank (1996) and is called the Valueless Features hypothesis. In this view, both lexical and functional categories are transferred early on (with a short stage in which only lexical projections are present), but functional categories lack values such as tense, agreement, etc., and are present as syntactic markers only (i.e. inflections may be lacking, but the syntactic operations linked to these categories will be in place).

These views have much in common with the approaches we will review next (and we will discuss empirical evidence about impaired functional cat-egories in that section), but crucially their belief is that all parameters can be reset.

3.5.4.3 Hypothesis 3: Partial access

No parameter resetting: proponents of this position claim that learners only have access to Universal (jrammar via their first language. They have already accessed the range of principles applying to their first language, and set parameters to the first language values, and this is the basis for their sec-ond language development. Other principles and parameter settings are not available to them, and if the second language possesses parameter settings that are different from those of their first language, they will have to resort to other mechanisms in order to make the second language data fit their internal representations. These mechanisms will be rooted in general prob-lem-solving strategies, rather than being Universal Grammar-based. Bley-Vroman claims:

Thus, the picture of the difference between child language development and foreign language learning as advocated here is the following:

Child language development Adult foreign language learning A. Universal Grammar A. Native language knowledge B. Domain-specific learning procedures B. General problem-solving systems This approach has attempted to account for the phenomena of transfer, and of the differences in the outcome of the learning process in L2 acquisition compared to LI acquisition.

(Bley-Vroman, 1989, p. 51) Schachter is also a supporter of the indirect access hypothesis, which she combines with the notion of a critical period for second language acquisi-tion. In a recent review (Schachter, 1996), she cites a number of studies of adult second language learners, claiming these show failure to acquire prin-ciples which are absent from the learners' first languages, and/or failure to reset particular parameters. For example, she cites her own work with Korean first-language learners of English as a second language, who per-formed randomly in grammaticality judgement tests of w/z-movement. In English, w/z-movement is allowed, but is restricted by the Subjacency prin-ciple (the extracted zo/z-word can move only across certain structural boundaries). In Korean, there is no w/z-movement, so the Subjacency prin-ciple is presumably not operative. If all the prinprin-ciples of Universal

Grammar are still available to the learner, the absence of this particular principle from their first language should not matter, and Subjacency should still be acquirable in English as a second language. Schachter claims that the Korean subjects' failure to recognize ^ - m o v e m e n t problems reflects the non-availability to them of Universal Grammar principles that were not already operative in their first language; that is, that Universal Grammar principles are accessible only as they have taken shape in the first language.

Schachter does accept that Universal Grammar may be available for child second language learners, but argues that there is a critical period (or periods) for the successful acquisition of second language principles and/or parameter settings, if these have not been operative in the learner's first lan-guage. She calls this critical period a Window of Opportunity, and argues that child second language learners pass through different Windows for dif-ferent modules of the target language (Schachter, 1996, p. 188). In support, she cites a study by Lee (1992) that tested Korean-English bilinguals on a particular parameter, the Governing Category parameter (GC), which is set differently in the two languages involved. (As we have seen already, this parameter has to do with the binding of items such as reflexives; the English reflexive must refer to the subject within its own clause, while in Korean it may refer to a more remote subject: Schachter, 1996, p. 178.)

In Lee's study, the Korean learners of English were of different ages; the youngest and oldest subjects had not acquired the English setting for the G C parameter, while the older children had apparently succeeded in doing so. Schachter (1996, p. 187) concludes that these findings show the Window of Opportunity not yet operative for the youngest learners, but available to the older children. As far as adult learners are concerned, she concludes that 'UG . . . fails to shed light on adult L2 acquisition - either in terms of a biological perspective on maturation or in terms of the known linguistic achievements of adult L2 learners'. Instead, she believes, the only principles and parameter settings easily available to the adult second lan-guage learner are those already activated in the course of first lanlan-guage learning.

Impaired functional features: lastly, we will briefly review two approaches which believe that second language grammars are Universal Grammar-constrained, but that not all parameter settings will be available to learners. Second language learners will therefore try to accommodate the second language grammar within the settings they already have. The Modulated structure building hypothesis (Hawkins, 2001) argues that learners start with 'minimal trees' (as described above), that is, lexical pro-jections determined by the first language. Functional propro-jections develop

gradually, with first language functional features transferring on to the sec-ond language, but only when the relevant syntactic representation has been sufficiently elaborated to instantiate the property in question. Hawkins and Chan (1997) argued that functional features cannot be reset in the second language. For example, Cantonese learners of English studied by Hawkins and Chan failed to acquire properties linked with w/z-movement, which does not exist in Cantonese. They argue that learners re-analyse the input to make it fit their first language settings. In an alternative view.

Constructionism 'proposes that the L2er uses a coalition of resources - a U G template (including, for example, a limited set of parameters, a small inventory of null anaphora, universal principles), first-language transfer, primary linguistic data, its mediation in social discourse (input and intake) and instructional bootstrapping - to construct the L2 vocabulary and grammar' (Herschensohn, 2000, p. 220).

What all these accounts crucially have in common is that they believe that second language acquisition is Universal Grammar-constrained, but that access to parametric options is unlike first language acquisition.

As we can see, there is much overlap between the approaches we have briefly reviewed here, and they might be better presented as a continuum.

It is useful, however, to separate them in this way, as they adopt different positions on the issues that are currently at the core of debates in generative second language acquisition. For example, they have different views on the Initial State, on the role of the first language, on the possibility of parameter resetting, on the Steady State (the final stable state), or on the role of non-Universal Grammar constrained mechanisms.

Recent work has generated new and more detailed hypotheses about which particular aspects of Universal Grammar might be transferred from the first language. For example, the availability or not of functional categories at the onset of second language acquisition has been the focus of much debate. Like first language learners, second language learners also go through a phase of using uninflected verbs initially, gradually introducing inflected forms into their grammar (Hyams, 1996; Wexler and Harris, 1996; Ionin and Wexler, 2000). But whereas in first language acquisition, 'the realisation or not of verbal inflection is not a random occurrence in early child language, but it is rather systematically linked to syntactic development' (Herschensohn, 2001), in second language acquisition the evidence is less clear, with some researchers arguing for such links and others not (White, 1996; Prevost and White, 2000; Sorace, 2000;

Franceschina, 2001; Herschensohn, 2001; Myles, in press a). In other words, researchers have been investigating the role played by functional features in second language acquisition, both from the point of view of their

availability (in parallel with discussions in first language acquisition the-ory), but also from the point of view of the transfer or not of first language functional features, and of the possibility of resetting parameters linked with verb morphology. (For fuller discussion of these questions, see Herschensohn, 2000; Hawkins, 2001; Myles, in press a.)

Table 3.1 (taken from White, 2003, p. 270) summarizes some of these issues.

To round off this section, it is fair to say that the argument concerning access to Universal Grammar in SLL is not concluded, and that strong defenders of all these positions can still be found. Often, they seem to be arguing about the best technical interpretation of admittedly indirect and tantalizing evidence, often gathered through grammaticality judgement tests, etc. Research in this area seems to have shifted from the initial ques-tion of the availability versus non-availability of Universal Grammar, towards a more modular view of language and the language faculty, with Universal Grammar itself being modular (Smith andTsimpli, 1995). As a result, the questions that studies in second language acquisition have been addressing are becoming more focused, testing the availability of sub-modules of Universal Grammar rather than Universal Grammar itself.

Table 3.1 L2 acquisition and UG: initial to steady state

UG-impaired UG-constrained Global Local No parameter Full access Full transfer, impairment impairment resetting (without transfer) full access Initial state ? L1 grammar L1 grammar UG L1 grammar

+ inert features

Development Pattern Some L2 No parameter Parameter Parameter matching; properties resetting setting directly resetting separate acquirable; to L2 values (L1 to Ln) constructions features

Li-like L2-like grammar L2-like grammar

{Source: White, 2003, p. 270)

3.6 Evaluation of Universal Grammar-based