Further assessment of a model of changeover behavior: Implications for the matching law

Choice in concurrent schedules is organized in visits to each alternative, and the duration of these visits is exponentially distributed. A model of changeover behavior based on this fact successfully described changeover behavior in two large data sets from published experiments, but some limitations were apparent in this analysis, seemingly reflecting an effect of the passage of time on the data across the lengthy experiments. This article describes an experiment that exposed rats to a dynamic concurrent variable-interval procedure designed to address these limitations. One of 35 possible combinations of overall and relative reinforcement rates was chosen pseudorandomly at the beginning of each session without signaling the specific combination in effect. By allowing the sensitivity parameter in the generalized matching law to be a function of the overall reinforcement rate, the model provided a satisfactory description of the results. This modification includes free parameters that presumably reflect the effects of how discriminable the alternatives are and how costly it is to switch between them, increasing the scope of the matching law. The updated model holds promise as the foundation for a general theory of performance in concurrent schedules of reinforcement.