Feedback functions for reinforcement: A paradigmatic experiment

Abstract

A reinforcement schedule states a rule for obtaining reinforcement as a function of behavior actually emitted and perhaps as a function of additional variables. These functions are here called “feedback functions.” Behavior actually emitted is, in turn, a function of obtained reinforcement. This reciprocal interdependency was quantified for an experiment in which pigeons chose among either three or two alternatives. Shifting from three to two alternatives, and vice versa, produced changes in the distribution of responding which were approximately accounted for by equations that combined the feedback functions with the matching law for reinforced responding. These equations predicted, among other things, a violation of the constant-ratio rule of formal choice theory and an absence of simultaneous contrast effects between response alternatives reinforced on variable-interval schedules. Both predictions were approximately confirmed.