Second Language Acquisition and Universal Grammar (review)

In 1989, the appearance of Lydia White’s book Universal grammar and second language acquisition was an important moment in the developing field of generative work on second language acquisition (SLA). It brought together linguistic theory and fragmentary evidence from a small number of generative L2 studies available at the time and argued that L2 grammatical knowledge is derived from the interaction between universal grammar (UG) and exposure to samples of the L2, with some influence from the L1. The present volume is a sequel to that work and an equally important landmark. A comparison of the two volumes reveals striking differences between the field as it was then and as it is now. The use that is made of linguistic theory to address issues in SLA has become more sophisticated; the volume of evidence available for testing hypotheses has grown; more is known about the advantages and disadvantages of different data elicitation method; and the role that data play as evidence for hypotheses; and important issues have emerged that are characteristic of SLA itself and distinguish it from cognate fields such as the investigation of mature native grammars or first language acquisition. For example, the relationship between speech production data and competence—a topic to which W devotes Ch. 6—appears much less direct in SLA than it is either for native speakers or in first language acquisition.

The present volume gives an excellent overview of the range of issues that currently interest generative L2 researchers, the way they construct theories about L2-speaker grammars (known to the field as interlanguage grammars—ILGs), and the methods they use to elicit data bearing on hypotheses derived from the theories. The book will be required reading for anyone seriously interested in what we know currently about the initial-state, development, and the steady-state grammars of L2 speakers.

At the same time, the book is not a simple review of existing generative SLA research. It is a sustained argument for a particular theory of SLA with the following premises:

(1) The initial-state L2 grammar consists entirely of the grammatical properties of a speaker’s L1 steady-state grammar, minus language-specific lexical items (known as the full transfer hypothesis);

(2) Learners restructure the grammar when they encounter sentences of the target language which cannot be parsed by the existing grammar (referred to as failure-driven development);

(3) In restructuring, the L2 learner is constrained by UG but has access to all of its resources (known as the full access hypothesis);

(4) There is something akin to an impairment in (adult) L2 grammars at the interface between the output of syntactic operations and the insertion of phonological exponents into the terminal nodes created by those operations. (The assumption here is a ‘separationist’ one, where phonological forms are inserted ‘late’.)

In short, the theory is referred to as ‘full transfer/full access’ (a term originally used in Schwartz & Sprouse 1994, 1996), with the addition of a missing surface inflection hypothesis (proposed in Prévost & White 2000). In arguing the case for this theory, W not only adduces evidence consistent with it, but also evaluates the evidence for and against seven other theories that have emerged in recent years: minimal trees, valueless features, full access without L1 transfer, the initial hypothesis of syntax, the global impairment hypothesis, the local impairment hypothesis, and the no parameter resetting hypothesis.

The book is organized into eight chapters, and the first thing to say is that it is not for the linguistically naive. The reader needs to be familiar already with work in the principles and parameters model of generative grammar as outlined, for example, in Chomsky 1986 and the first three chapters of Chomsky 1995. Although W is wonderfully clear in the way she discusses the linguistic assumptions of particular L2 studies, readers need to feel comfortable with concepts like binding theory, disjoint reference, subjacency, LF, c-command, phi-features, and so on. This is made...