Preference pulses without reinforcers

Preference pulses are thought to represent strong, short-term effects of reinforcers on preference in concurrent schedules. However, the general shape of preference pulses is substantially determined by the distributions of responses-per-visit (visit lengths) for the two choice alternatives. In several series of simulations, we varied the means and standard deviations of distributions describing visits to two concurrently available response alternatives, arranged "reinforcers" according to concurrent variable-interval schedules, and found a range of different preference pulses. Because characteristics of these distributions describe global aspects of behavior, and the simulations assumed no local effects of reinforcement, these preference pulses derive from the visit structure alone. This strongly questions whether preference pulses should continue to be interpreted as representing local effects of reinforcement. We suggest an alternative approach whereby local effects are assessed by subtracting the artifactual part, which derives from visit structure, from the observed preference pulses. This yields "residual" preference pulses. We illustrate this method in application to published data from mixed dependent concurrent schedules, revealing evidence that the delivery of reinforcers had modest lengthening effects on the duration of the current visit, a conclusion that is quantitatively consistent with early research on short-term effects of reinforcement.