Does a prepaid group practice (PGP) deliver less outpatient mental health care than the fee-for-service (FFS) sector when they serve comparable populations with comparable benefits? To examine this issue, we used data from the Rand Health Insurance Study, which randomized families into a prepaid group practice or FFS insurance plans. Participants in a FFS plan with no cost sharing (i.e., free care) are equally likely to visit a mental health specialist in a year, but incur 2.8 times the costs of prepaid participants (p less than .05). This difference is due to fewer visits per user, substitution of psychiatric social workers for psychiatrists and psychologists, and reliance on group rather than individual therapies in the prepaid plan. Because of the experimental design, these differences are due to institutional and incentive differences rather than adverse selection. We found no evidence of appreciable or significant adverse selection into or out of the prepaid group practice. A full evaluation of the desirability of prepaid or fee-for-service care requires data on health outcomes, which are not presented here.