Animate objects usually move for intrinsic reasons, while natural inanimate objects move only because of extrinsic physical forces. A model is proposed of how people understand this distinction, involving causal inferences between a moving object's trajectory, context, and animacy. Three experiments are reported, one with adults and two with 3- and 4-year-old children. In two experiments, adults and children viewed computer-animated motion events. Subjects first viewed a natural outdoor scene, containing forces such as gravity, audible wind, and moving water. An object represented by a white dot then moved in the scene "at night," in a trajectory that was either consistent or inconsistent with the extrinsic forces present ("force-consistency"). Subjects judged the object's animacy by choosing between pictures of natural animate and inanimate objects, and also explained the chosen object's motion. Adults generally judged each object's animacy as expected, based on the force-consistency of its trajectory, while children's judgments were more variable. Significant crossover effects showed that the same trajectory could be judged as animate or inanimate, depending on its force-consistency. Both children and adults almost always gave motion explanations that correctly corresponded with their animacy judgments, featuring extrinsic causes for inanimate objects and intrinsic causes for animate objects. Children's explanations showed an increased understanding with age of specific intrinsic and extrinsic causes for motion. In the third experiment, children judged the self-motion capability of familiar objects, and explained why each could start and stop moving. Responses revealed domain-specific causal reasoning about animate objects, inanimate objects, and artifacts: self-motion judgments were almost always correct, and explanations correctly invoked intrinsic or extrinsic causes for motion. The results of all three experiments are consistent with the proposed model, and with previous accounts in which motion paths are causally interpreted as animate or inanimate. They are difficult to reconcile with alternate accounts, which propose that motion paths are directly perceived as animate or inanimate. Methodological implications for the use of computer animation in motion perception studies are also discussed.