Situationists maintain that psychological evidence (e.g., the well-known Good Samaritan experiment) challenges a key assumption of virtue ethics, namely that virtuous people display cross-situational consistency of behavior. This situationist critique is frequently thought to pose a serious threat to virtue ethics. Virtue ethicists have so far mainly put forward conceptual rather than empirical arguments against situationism. In this paper, we examine the extent to which a plausible empirical argument can be developed against situationism, and in favor of virtue ethics. We note that such an argument will naturally entail empirical simplification, and consequently concessions from both sides of the debate will be required. We argue, however, that with such concessions in place, the debate between situationists and virtue ethicists can be captured by a claim about the way that virtuous behavior may change when situations change. We report the results of an experiment (conducted twice, for replication) showing how social science methods can be introduced to the situationism debate to test such a claim empirically. Our results offer tentative support for eudaemonist and agent-based varieties (but not for target-centered) versions of virtue ethics.