Three experiments investigated the proposal that inductive inferences about different properties depend on different measures of similarity. In Experiments 1 and 2, Ss were given the premise that a category of animals has some property and judged the probability that another category of animals also has that property. Ss made the strongest inferences when the kind of property (anatomical or behavioral) corresponded to the kind of similarity between the animal categories (anatomical or behavioral). These results cannot be explained in terms of a single measure of similarity underlying induction. In Experiment 3, Ss rated the similarity of animal pairs with respect to anatomy or behavior. Regression analyses showed that both behavioral and anatomical similarity influenced behavioral inferences, but only anatomical similarity influenced anatomical inferences.