This study considered differences in induction of biological properties between children and preadolescents based on differences in stimuli processing in these two groups. Two experiments tested predictions that young children, but not preadolescents, base their inductive inference on aggregating information from different sources rather than relying on a single source of information. In both experiments, children and preadolescents from three age ranges were presented with an inductive task involving triads of animal pictures: a target baby animal, and two test animals, one a "neighbor" animal who "played with the baby," and an animal who "gave birth to the baby." In experiment 1, linguistic labels were fully crossed with relationship information, whereas in experiment 2 perceptual similarity information was fully crossed with relationship and labeling information. Findings indicated that while 10- and 11-year-olds relied exclusively on inheritance (when target animal shared only inheritance information and not labeling with the mother) across experiments, 4- and 5-year-olds relied on an aggregate of multiple sources of information, and 7- and 8-year-olds fell between these 2 extremes. In addition, while the relative weight of inheritance on inferences increased with age, the weights of other information sources decreased. The results support the hypotheses suggesting that between 8 and 10 years of age, children undergo a developmental shift from a holistic feature-integration induction to knowledge-based induction based on a single most predictive source. (Contains 20 references.) (HTH)