Mixed payment systems have become a prominent alternative to paying physicians through fee-for-service and capitation. While theory shows mixed payment systems to be superior, empirically, causal effects on physicians' behavior are not well understood when introducing mixed systems. We systematically analyze the influence of fee-forservice, capitation, and mixed payment systems on physicians' service provision. In a controlled laboratory setting, we implement an exogenous variation of the payment method. Participants, in the role of physicians, in the lab (N=213) choose quantities of medical services affecting patients' health outside the lab. Behavioral data reveal significant overprovision of medical services under fee-for-service and significant underprovision under capitation, though less than predicted when assuming profitmaximization. Introducing mixed payment systems significantly reduces deviations from patient-optimal treatment. Responses to incentive systems can be explained by a behavioral model capturing physician altruism. We find substantial heterogeneity in physician altruism. Our results hold for medical and non-medical students.