This article is a defence of the Full Transfer/Full Access (FT/FA) model. FT/FA hypothesizes that the initial state of L2 acquisition is the final state of L1 acquisition (Full Transfer) and that failure to assign a representation to input data will force subsequent restructurings, drawing from options of UG (Full Access). We illustrate the FT/FA model by reviewing our analysis of the developmental Turkish-German Interlanguage data of Schwartz and Sprouse (1994) and then turn to other data that similarly receive straightforward accounts under FT/FA. We also consider two other competing hypotheses, both of which accept Full Access but not Full Transfer: the Minimal Trees hypothesis (no transfer of functional categories) of Vainikka and Young-Scholten (1994; 1996) and the Weak Transfer hypothesis (no transfer of the values associated with functional categories) of Eubank (1993/94). We provide an example of (extremely robust) L2 acquisition data that highlight the inadequacy of the Minimal Trees hypothesis in regard to stages of Interlanguage subsequent to the L2 initial state. As for Weak Transfer, we show that the morphosyntactic empirical foundations which drive the entire approach are flawed; hence the Weak Transfer hypothesis remains without motivation. Finally, we consider several conceptual issues relating to transfer. These all argue that the FT/FA model provides the most coherent picture of the L2 initial cognitive state. In short, FT/FA embodies the most suitable programme for understanding comparative Interlanguage development.

Psychologically speaking, all linguistic behavior is the overt manifestation of some type of underlying knowledge that is represented in the mind/brain of an individual. Exposure to linguistic data is necessary for growth of the system of knowledge. On the basis of only overt linguistic behavior, how can we ascertain whether the native and nonnative knowledge systems that people have are of distinct or similar types? Is there a (necessary) relationship between type of knowledge and type of linguistic exposure?The hypothesis to be defended is that negative data and explicit data result in a type of knowledge that is not to be equated with linguistic competence. The claim is not that negative and explicit data cannot give rise to knowledge; rather, the specific claim is that only positive data can effect the construction of an interlanguage grammar that is comparable to the knowledge system that characterizes the result of first language acquisition.

No abstract

No abstract

In this paper I argue for the necessity of recognizing the epistemological basis of language (and hence of linguistic theory) for research in and theories of second language acquisition. In particular, I review the arguments for a generative approach to linguistic theory (e.g. Chomsky, 1965, 1975, 1981) and for why language as a system of knowledge must be distinct from other sorts of know ledge (Fodor, 1983), with the purpose of clarifying many misconceptions that seem to have arisen with respect to the work in generative grammar over the last 20 years. After doing this I argue that the null hypothesis for second language acquisition is, as concerns its mental representation of linguistic knowledge, that its epistemological status should be assumed to be the same as that of L 1 until proven otherwise. I then demonstrate how SLA theory (e.g. Krashen, 1981) can be elucidated by subsuming (parts of) L2 under linguistic theory with its firm epistemological basis, and how, in particular, one could empirically test Krashen's theory as well as any other theory of SLA that assumes L 1 and L2 to be epistemologically equivalent. In addition I discuss the need for researchers to consider the special epistemological status of linguistic knowledge before prescribing L2 pedagogy. In sum this is a paper that takes a step back into the philosophical debate concerning the mental status of language in general in order for us to be able to take a step forward in second language research in particular.

This paper reassesses the role of Negative Evidence (NE) in nonnative language acquisition. We argue that the grammar-building process cannot make use of NE to restructure (Interlanguage) grammars - irrespective of logical need. The empirical basis comes from White's (1991a; 1991b) study of French speakers acquiring English, where the 'Verb Movement' parameter and the particular learnability problem of 'unlearning' thematic Verb- movement were the focus. The L2ers start off assuming the L1 value of [+] Verb-movement, thus incorrectly allowing the order S V Adv O in English, and the issue is whether NE can force a switch to the [-] value, whereby S V Adv O should be excluded. While it is indisputable that the L2ers changed their linguistic behaviour as a direct consequence of their exposure to NE, the conclusion we draw is quite distinct from that of White. Based on both the postinstruction data and an argument grounded in formal learnability theory, we show that an inherent contradiction must be imputed to the Interlanguage 'grammar' to account for the results: in addition to no longer permitting S V Adv O, the L2ers also (incorrectly) disallow S V Adv PP; to exclude the latter, the grammar must have 'un learned' base-generating Adverbs to the right of VP but other data dispute this, i.e., S V O/PP Adv is still allowed. Since natural language grammars cannot contain such inherent contradictions, we conclude that a natural language grammar could not be the source of this L2 behaviour. Our explanation is that the L2ers simply extended the *S V Adv O pattern that they were taught. In sum, there is no evidence that NE caused the L2ers to unlearn Verb-movement and hence NE did not restructure the Interlanguage grammar. Implications of this conclusion are discussed in relation to the issues of learnability and 'UG-accessibility' in L2A.

The goal of this article is primarily conceptual in nature in that it attempts to offer a novel way to employ developmental sequence data in order to decide between UG-based (e.g., Felix, 1985;White, 1985) and problem-solving (e.g., Bley-Vroman, 1989;Clahsen & Muysken, 1986) models of (adult) nonnative earnmatical development. It is argued that the two traditional uses of developmental sequences found in the L2A literature are incapable of testing between the two opposing models. It is further argued that a more valid test consists in comparing the developmental sequence (in the acquisition of language X ) of child L2ers to that of adult L2ers, holding their L1 constant. Only here are opposite predictions made: An L2-UG model predicts a single sequence for both adult and child L2ers; a problem-solving approach predicts distinct sequences. L2 developmental sequences from two areas of syntax are then presented to test between the two models. Results support the UG-based model of L2A.

This experimental study compares the acquisition of the English to- and for-dative alternation by L1 English, L1 Japanese, and L1 Korean children. It is well known that there are restrictions on the verbs that can enter into the dative alternation—for example, you can show the results to someone and show someone the results; and you can demonstrate the results to someone but you cannot *demonstrate someone the results. L1 children sometimes overextend the double-object variant to verbs that disallow it. One question we investigate is whether L2 children, like L1 children, overextend the double-object variant. A second question we probe is whether L2 children, like L2 adults, transfer properties of the L1 grammar. Japanese disallows all double-accusative constructions. Korean disallows them with analogues of to-dative verbs; but with analogues of for-dative verbs, Korean productively allows them—more broadly, in fact, than English—if the benefactive verbal morpheme cwu- is added. Results from an oral grammaticality judgment task show (a) that all groups allow illicit to-dative double-object forms and (b) that the Japanese—but not the Koreans—allow illicit for-dative double-object forms. This bifurcation, we argue, stems from the fact that Korean (but not Japanese) has an overt morphological licensor for double objects. We thus find evidence of both (a) overgeneralization, like in L1 acquisition, and (b) L1 influence, like in adult L2 acquisition, in this case from the (syntactic) argument-changing properties of overt morphology.