Response stability and variability induced in humans by different feedback contingencies

In two experiments, the behavioral effects of different response-feedback contingencies were examined with a task requiring human subjects to repeatedly type three-key sequences on a computer keyboard. In Experiment 1, the subjects first received positive feedback for response variability, followed by no feedback, or vice versa In Experiment 2, the subjects first received positive feedback for response variability, followed by response-independent positive feedback, or vice versa. Response stability and variability were examined using different measures, such as percentage of trials meeting the variability criteria, frequency of use of the different response alternatives, and autocorrelations as an index of response randomness. The subjects' behavior in the first phase in each condition came to reflect the current feedback contingency. Depending on the measure examined, responding after each contingency change was characterized by both response stability and decreases or increases in response variability. The collective results are discussed in the framework of previous animal and human studies on behavioral stability and variability.