Count nouns, adjectives, and perceptual properties in children's novel word interpretations.

Abstract

Three-year-old children were shown a novel exemplar toy and asked to judge test items that differed from the exemplar in shape, coloration, or material substance. In the count noun condition, children judged whether test items had the same novel name as the exemplar. In the adjective condition, children judged whether a test hern could be described by the same novel adjective as the exemplar. The results of 3 experiments indicated that children systematically attend to shape in interpreting novel count nouns, but their interpretation of adjectives is contextually determined. By the age of 6, children have acquired roughly 14,000 words (Templin, 1957). How do they acquire so many words so fast? If one views the child's acquisition of word-referent relations as an instance of unbiased hypothesis testing, then the rate of early word acquisition is difficult to explain. As Quine (1960) pointed out, the use of a word in the context of some scene provides evidence consistent with many hypotheses, only one of which will usually be correct. Chomsky (1986) similarly argued that if language learners were free to form any possible hypotheses about intended meaning from spoken language, then it would be unlikely that they would learn language as rapidly as they do because it would be unlikely that they would test the correct hypotheses early in language learning by happenstance alone. Chomsky argued that the rapid and error-free language learning that we observe in children requires that children be biased to entertain some hypotheses more than others. This idea has motivated much research on children's early word learning. Developmentalists have shown that children's novel word extensions are constrained or biased in certain directions. For example, young children appear biased to interpret count nouns as referring to object categories and not individual objects (Katz, Baker, & MacNamara, 1974) or thematic relations between objects (Markman & Hutchinson, 1984; Waxman & Kosowski, 1990). Children are biased to allow only one label for a single referent (mutual exclusivity, e.g., Markman, 1989; Markman & Wachtel, 1988), and they are biased to attend to shape when extending a novel count noun across novel objects (Landau, Smith, & Jones, 1988). The existence of these biases is sometimes cited as an explanation of rapid word growth; the idea is that word learning proceeds as fast as it does precisely because word learning biases exist.