Dumb mechanisms make smart concepts

There is an ongoing debate on the nature of the processes and knowledge involved in learning language. On one side of the debate, people argue that children learn words through deliberative processes that use propositional conceptual knowledge; on the opposing side, people argue that children learn words through automatic processes and knowledge based on learned associations among perceptual features. In this paper we concentrate on the Animate/Inanimate distinction as evidenced in children's novel noun generalizations. The results of two experiments with 3-yearolds and adults suggest that 1) automatic processing guides children's generalizations of novel nouns and 2) "conceptual" knowledge may be formed as a web of learned correlations.