The role of reinforcers in influencing choice was studied by use of a schedule that included a random intermixing of reinforced and explicitly non-reinforced components. The just-reinforced response had a high likelihood of being repeated (win-stay), although there was no differential reinforcement for doing so, whereas responses just followed by explicit non-reinforcement had a very low probability of repetition (lose-stay). Non-parametric indices based on the theory of signal detection were used to derive a choice measure of reinforcement which was independent of alterations in average response rate. Treatments with d-amphetamine (0.2-4.5 mg/kg), chlordiazepoxide (0.25-16 mg/kg) and alpha-flupenthixol (0.03-0.6 mg/kg) showed that changes in the choice measure could be dissociated from changes in the response rate. These findings were supported by extinction and satiation tests.