Learnability and Cognition: The Acquisition of Argument Structure

In his new book, Leanlability and Cognition: The Acquisitioii of Argunient Structure, Steven Pinker addresses the problem of structural generalization in Language Acquisition: How d o children know when they can go beyond imitation and use words in sentence structures that are unattested for those words? That is, say a child has heard the verb “give” in both of its possible forms: “John gave the ball to Mary”, “John gave Mary the ball”. In addition, say that the child has heard the verbs “bring” and “pull” in only the first form: “John brought the ball to Mary”, “John pulled the wagon to Mary”, how is the child to figure out that “bring” can also appear in the second form, but that “pull” cannot? Pinker’s solution involves both an extension and refinement of his theory of Semantic Bootstrapping (Pinker, 1984); namely, that semantics can help children learn about the syntax of their language. His focus is on a specific aspect of syntax, namely, how children acquire the argument structures of certain classes of verbs, and his solution rests on detailed analyses of the meanings of those verbs. In describing this solution, Pinker presents a detailed theory of verb representation, numerous learning procedures, and some experimental data concerning children’s abilities to make syntactic predictions based on semantics. Pinker’s technique of using recurring examples throughout the book, as he moves from representation to learning to developmental evidence, allows the reader to form a coherent picture of how the entire theory applies. Yet for all its depth and detail, there are some surprising gaps with regards to theory-internal validity and theory-external evidence. Careful scrutiny suggests that the core of his proposed solution, the Narrow-Range level of semantic representation, is extraordinarily difficult to capture in the adult lexicon, and may be unlearnable b y