A mathematical model of learning under schedules of interresponse time reinforcement.

Abstract

After a discussion of the behaviours that are found to occur under various common schedules of reinforcement, a mathematical model of learning is proposed, which covers those schedules where reinforcement can be prescribed, as a function of the time between successive responses. There are several possible conceptualisations of the model. The one considered most often is one that uses the basic notions of Stimulus Sampling Theory. The predictions of the model are tested against the known properties of ratio, interval and DR1 schedules of reinforcement. In only one case do the predictions of the model run counter to accepted experimental fact. The model is also tested against specific data from human subjects on ratio, interval, and DRL schedules. The model fits well to the asymptotic interresponse time distributions for these schedules. However the fit of the model to distributions conditional on the previous response being reinforced, or the previous response not being reinforced, range from good to very bad, suggesting that sequential effects are more complex than implied by the model.