Children's hypotheses about word meanings: is there a basic level constraint?

We tested the hypothesis that a basic level constraint guides preschool children's hypotheses about word meanings. In the first study, a puppet labeled pictures with novel words. Three-, four-, and five-year-olds then chose referents for the novel word from a set of pictures. A pretest allowed us to select three sets of items for each subject: items that were familiar at the subordinate and basic level (e.g., poodle-dog), items that were familiar at the basic level but not at the subordinate level (e.g., sedan-car), and items that were unfamiliar at both levels (e.g., manta-ray). Contrary to the basic level constraint hypothesis, children tended to interpret novel words (e.g., quadruped) at the basic level only for items where they already knew a basic level name. In a second study, using a simplified procedure, three- and four-year-olds again showed no bias toward basic level interpretations except when translation to familiar basic level words was possible. A third study controlled for the possibility that our use of multisyllabic words might have biased children against basic level interpretations; despite the use of single-syllable nonsense words, children in this study again did not favor basic level interpretations. In the final two studies, adults' ratings of picture similarity and children's judgments in a triad task confirmed that the differences in children's basic level biases could not be accounted for by differences in similarity across the item types. Taken together these results suggest that although other constraints on word meaning may be in place, preschoolers are flexible in interpreting new words at different hierarchical levels.