A Model of Changeover Behavior in Concurrent Schedules of Reinforcement

Animals in the wild have access to a variety of food sources, and they must allocate their time between them in an efficient manner. This requires deciding when to depart from the source currently being exploited to travel to a different one. A general model of this decision process can be derived from three assumptions well supported by the learning literature. The first one is the so-called matching law, which says that the proportion of time allocated to any one behavior equals the proportion of total value obtained from that option. The second one is that the dwell times (i.e., the length of time spent at a food source before departing) follow an exponential distribution; that is, the probability of departing at any moment remains constant across time. The last assumption is that the sum of the leaving rates (i.e., the number of departures from a food source per unit of time) depends only on the total food rate produced by all the available sources. This thesis provides an overview of a first version of this model and its assessment using two large datasets from published experiments. It also describes an experiment with rats that was designed to refine the model based on the limitations noted during the assessment. Rats were placed in operant chambers where they could press two levers to produce food at rates chosen in a pseudorandom fashion at the beginning of each daily session. The results were consistent with the assumptions of the model, and suggested ways in which its equations could be improved. The updated version of the model accurately described the data while also suggesting ways to incorporate other factors known to affect how animals make decisions. Ways in which the model could be applied to other situations, both at the basic and applied level, are discussed.