A randomized trial was conducted to determine the effectiveness of a health care plan which uses physicians as gatekeepers to control health services use and charges. New enrollees in United Healthcare (UHC), an independent practice association, were randomly assigned to the standard UHC plan requiring a gatekeeper, or to an alternate plan with equal benefits but without a gatekeeper. Individuals in both plans were similar in demographic characteristics, perceived health status, and other health insurance coverage. The gatekeeper plan had 6 percent lower total charges per enrollee than the plan without a gatekeeper. There were minor differences in hospital use and charges. Ambulatory charges were $21 lower per person per year in the plan with a gatekeeper (95% CI = -39.9, -2.1) and these were due to .3 fewer visits to specialists (95% CI = -0.50, -0.10). We conclude that a health plan which incorporates incentives and penalties for physicians to act as gatekeepers can reduce the cost of ambulatory services by limiting specialist visits.