If young children approached the task of word learning with a specific hypothesis about the meaning of novel count nouns, they could make the problem of word learning more tractable. Six experiments were conducted to test children's hypotheses about how labels map to object categories. Findings indicated that (a) 3- and 4-year-olds function with an antithematic bias; (b) children do not reliably extend novel nouns to superordinate exemplars when perceptual similarity is controlled until approximately age 7; and (c) children expect novel nouns to label taxonomic categories at the basic level, even in the presence of a perceptually compelling distracter. Results are interpreted as supporting the principle of categorical scope (R. M. Golinkoff, C. B. Mervis, & K. Hirsh-Pasek, 1994).