We explore how competition between physicians affects medical service provision. Previous research has shown that, without competition, physicians deviate from patient-optimal treatment under payment systems like capitation and fee-for-service. Although competition might reduce these distortions, physicians usually interact with each other repeatedly over time and only a fraction of patients switches providers at all. Both patterns might prevent competition to work in the desired direction. To analyze the behavioral effects of competition, we develop a theoretical benchmark that is then tested in a controlled laboratory experiment. Experimental conditions vary physician payment and patient characteristics. Real patients benefit from provision decisions made in the experiment. Our results reveal that, in line with the theoretical prediction, introducing competition can reduce overprovision and underprovision, respectively. The observed effects depend on patient characteristics and the payment system, though. Tacit collusion is observed and particularly pronounced with fee-for-service payment, but it appears to be less frequent than in related experimental research on price competition.