Adding types, but not tokens, affects the breadth of property induction

AbstractThe extent to which we generalize a novel property from a sample of familiar instances to novel instances depends on the sample used. In these experiments, we are interested in two sample characteristics: number of types (discrete entities) and number of tokens (copies of the same entity) that share a novel property. Existing studies permit separate and conditional hypotheses about the effects of adding types and tokens, but no study has examined the effects of both variables on generalization stimuli varying in similarity. We find that adding types broadens generalization to similar stimuli, but tightens generalization to dissimilar stimuli. Adding tokens does not affect generalization, but adding repetitions that are framed as types produces some tightening. Implications for models of inductive reasoning are discussed.

Return to previous page