Brass Tacks in Linguistic Theory

In the normal course of events, children manifest linguistic competence equivalent to that of adults in just a few years. Children can produce and understand novel sentences, they can judge that certain strings of words are true or false, and so on. Yet experience appears to dramatically underdetermine the competence children so rapidly achieve, even given optimistic assumptions about children’s nonlinguistic capacities to extract information and form generalizations on the basis of statistical regularities in the input. These considerations underlie various (more specific) poverty of stimulus arguments for the innate specification of linguistic principles. But in our view, certain features of nativist arguments have not yet been fully appreciated. We focus here on three (related) kinds of poverty of stimulus argument, each of which has been supported by the findings of psycholinguistic investigations of child language. The first argument hinges on the observation that children project beyond their experience in ways that their experience does not suggest. It is untendentious that children project beyond their experience, in the sense of acquiring a state of linguistic competence that they apply to novel constructions. The issue is how children project beyond their experience. That is, do children induce (or abduce) in the fashion of good scientists, on the basis of experience characterized in (more or less) observational terms; or do they project in more idiosyncratic and languagespecific ways? To what degree is human language acquisition ‘‘data driven,’’ and to what degree is it determined by the human genome? Clearly, experience matters. Typical children growing up in Tokyo achieve a state of linguistic competence that differs in some respects from the state achieved by typical children growing up in Topeka. According to the theory of universal grammar (UG), however, the differences between natural human languages—like English and Japanese, which any normal child can learn in the right context—are relatively small as compared with the differences between natural human languages and other logically coherent systems (equally compatible with the experience of human children) for associating signals with meanings. If so, this supports the nativists’ contention that