Three experiments examined how children's domain knowledge and observation of exemplars interact during concept acquisition and how exposure to novel exemplars causes revision of such knowledge. In Experiments 1 (N=126) and 2 (N=64), children aged 4 to 10 years were shown exemplars of fictitious animal categories that were either unrelated to, or consistent with, their prior knowledge in 25% or 75% of presented exemplars. In Experiment 3, children (N=290) saw fictitious animal, artifact, or unfamiliar social categories that were either consistent or inconsistent with their prior knowledge in 20%, 40%, 60%, or 80% of exemplars. In the test, children made judgments about the likely co-occurence of features. In all experiments, prior knowledge and exemplar observation independently influenced children's categorization judgments. Utilization of prior knowledge was consistent across age and domain, but 10-year-olds were more sensitive to observed feature covariation. Training with larger categories increased the impact of observed feature covariation and decreased reliance on prior knowledge.