David Premack, University of Pennsylvania

An opportunity to engage in more probable responses will reinforce a less probable response. Consider a an experiment that reversed the reinforcing effects of two stimuli by varying the probabilities of the responses occasioned by each (Premack, 1962).

A rat’s running in a wheel was controlled by engaging or releasing a brake on the wheel. The rat’s drinking was controlled by moving a drinking tube in or out of an opening in a wall nearby the wheel. As tested during brief periods when both responses were available, running became more probable than drinking when the wheel was locked with water freely available, but drinking became more probable than running when the drinking tube was removed with the wheel freely available.

In each case, the opportunity to engage in the more probable response became an effective reinforcer of the less probable response. When running was more probable than drinking (wheel locked), licking became more likely if it released the brake and allowed the rat to run than if it did not allow the opportunity to run. Conversely, when drinking was more probable than running (drinking tube removed) running became more likely if it produced the drinking tube and allowed the rat to drink than if it did not allow the rat an opportunity to drink.

This experiment implies that reinforcers cannot be defined independently of the responses that they reinforce. In Premack’s experiment, drinking reinforced running when drinking was more probable than running, but running reinforced drinking when the probabilities were reversed. Therefore, reinforcers are relative and not absolute. There important properties are based on the responses for which they provide an opportunity.