Many everyday activities involve the use of one action tomodify the effects of another: When driving, shiftinggears modifies the influence of pressing the gas pedal onacceleration; when cooking, the rate of adding aparticular ingredient modifies the influence of stirring onviscosity. Here, we investigate a general ability to learnhow to use actions to control schedules of reinforcement.In Experiment 1, participants quickly discovered theoptimal rate of responding on an action that controlledthe rate of reward contingent on performing a differentaction. In Experiment 2, when the modifying action wasitself rewarded, participants failed to discover the optimalrate. Implications for formal theories of instrumentalbehavior are discussed.