ABSTRACT
Skinner (1956) related how, on a Saturday, he ran short of food pellets for the rats that he was training. To maintain the experiment, he decided to reward the rats once every minute, rather than for each response. Over the weekend, the behavior changed and then stabilized, showing a different but very regular pattern. For any scientist, an orderly pattern of results immediately suggests a fruitful vein of research, and Skinner naturally followed up his findingdiscovering the schedules of reinforcement, which were fully described and empirically investigated by Ferster and Skinner (1957). Reinforcers for behavior
could be made contingent on either the passage of time (interval schedules) or on the numbers of responses emitted {ratio schedules). The times, or the number of responses required, could be fixed or variable. Each combination of these variables produced an invariant pattern on a cumulative recorder, each different in some respects from the others. When fixed numbers of responses were required (Fixed-Ratio, or FR, schedules), there was a pause in responding after each reinforcer, followed by a sharp transition to a high and constant rate that was maintained until the next reinforcer was obtained-the FR break-and-run pattern. When a fixed time had to elapse before a response could be reinforced (Fixed-Interval, or FI, schedules), the pattern was similar, but the transition between pausing and responding was gradual-the FI scallop. When either a variable time had to elapse, or a variable number of responses had to be emitted {Variable-Interval, or VI, schedules or Variable-Ratio, or VR, schedules), animals responded at a high and relatively constant rate with few periods of pausing. But VR schedules engendered generally higher response rates than did equivalent VI schedules. All these patterns were, by and large, independent of the overall times or numbers of responses required, of the species, of the type of response required, of deprivation level, of the type of reinforcer, and of many other possible variables. Here, then, was a set of invariant patterns with very wide generality. Different patterns were shown even in extinction after training on each of the schedules.
