On the insufficiency of evidence for a domain-general account of word learning

Abstract

One question lies at the heart of our interchange with Bloom and Markson: is word learning the result of domain-speci®c or domain-general abilities? Undoubtedly, we cover a good deal of common ground. We agree (a) that discovering which aspects of language acquisition are speci®c to language, and which are shared more generally with other cognitive tasks, requires careful attention, (b) that the answer to this question may depend upon which aspect of language is under consideration, and (c) that the acquisition of novel words and facts share some important components, including establishing a mapping to a designated individual, and retaining this mapping over time. Yet there is also a clear difference between our positions. Markson and Bloom (1997) argued forcefully for a domain-general account of word learning. We countered, highlighting some crucial distinctions between what it takes to learn a word as compared to a fact. We concluded that there is, as yet, insuf®cient logical or empirical support for a domain-general account of word learning. To recap, Markson and Bloom (1997) demonstrated that children were resoundingly successful at fast-mapping either a novel word (e.g. `This is a koba') or fact (e.g. `My uncle gave this to me'), and retaining these mappings over time. They therefore concluded that fast-mapping is not speci®c to word learning. We agree entirely. However, Markson and Bloom went beyond this evidence, claiming to have uncovered a1⁄4evidence against a dedicated system for word-learning1⁄4o. It is this latter claim ± the claim that became the title of their original paper ± to which we take strong exception. The real problem is that although Bloom and Markson's data pertain to one S.R. Waxman, A.E. Booth / Cognition 78 (2001) 277±279 277