Are explanations of different kinds (formal, mechanistic, teleological) judged differently depending on their contextual utility, defined as the extent to which they support the kinds of inferences required for a given task? We report three studies demonstrating that the perceived "goodness" of an explanation depends on the evaluator's current task: Explanations receive a relative boost when they support task-relevant inferences, even when all three explanation types are warranted. For example, mechanistic explanations receive higher ratings when participants anticipate making further inferences on the basis of proximate causes than when they anticipate making further inferences on the basis of category membership or functions. These findings shed light on the functions of explanation and support pragmatic and pluralist approaches to explanation.