The authors conducted three experiments examining the effect of grouping on children's generalization of animal labels. In Experiment 1 (N = 96), first graders (M age = 6 years, 10 months) who had seen a novel animal grouped with similar animals generalized its trained label more broadly than those who had seen it by itself or grouped with dissimilar animals. Generalization of artifact labels was unaffected by condition. In Experiment 2, 4-year-olds (N = 96) also showed this pattern, but 3-year-olds (N = 96) showed no effects. In Experiment 3 (N = 48), 4-year-olds again showed the effect for animal labels in a design that eliminated alternatives to a grouping explanation. At least by age 4, children tend to judge that members of a group of similar animals belong to the same category, and this judgment affects how they interpret a label that is introduced for a member of the group. Explanations for why children tend to make these judgments and the implications for accounts of children's word learning are discussed.