Bringing theories of word learning in line with the evidence.

Two issues are at stake in this interchange. One concerns the relation between perceptual, conceptual and linguistic knowledge in early word learning. The other concerns the judicious treatment of evidence. Briefly stated, we designed our experiments (Booth & Waxman, 2002) with one clear goal: to document the pervasive role of conceptual information in naming. Three-year-old children in all conditions witnessed the same novel target objects, labeled with the same novel word (a count noun, e.g. dax). To vary the conceptual status of these named objects, we created brief vignettes that described them either as animate objects (e.g. “...has a mommy and daddy who love it very much”) or as artifacts (e.g. “...was made by an astronaut to do a special job on her spaceship”). We examined children’s extension of these newly learned words. When the objects were described as artifacts, children extended on the basis of shape alone. But when the very same objects were described as animate objects, children extended on the basis of both shape and texture. Moreover, when we placed eyes (a strong perceptual cue to animacy) on the objects, but described them with the artifact vignette, children’s performance was consistent with the artifact, rather than the animate, pattern of word extension. This documents that the conceptual status of an individual permeates early word learning, and does so even in the face of conflicting perceptual cues. It is not surprising that our paper sparked a response from Smith, Jones, Yoshida, and Colunga (2003). This work, like that of others (e.g. Bloom, 2000; Gelman & Markman, 1987; Gelman & Medin, 1993; Keil, 1994; Kemler Nelson, Russell, Duke, & Jones, 2000; Soja, Carey, & Spelke, 1991; Welder & Graham, 2001), challenges the contention that word learning can be accounted for by a “...dumb attentional mechanism” (Smith, Jones, A.E. Booth, S.R. Waxman / Cognition 87 (2003) 215–218 215