Correlated Delay of Reinforcement

It is well established that variations in an operant response can be differentiated by arranging appropriate reinforcement conditions. For example, Skinner (1938) and Ferster and Skinner (1957) have selectively reinforced duration, force, and rate of lever pressing in rats. In each case, reinforcement was applied whenever the "correct" response variation occurred and was withheld for all other variations of the response. However, this all-or-none reinforcement is probably not a necessary condition for response differentiation; it may be sufficient that the specified response variation receive a slightly more favorable reward (e.g., larger amount, shorter delay, higher percentage of reward) than do alternative variations of the operant. Under such conditions, if the normal variability in details of the operant ensures some experience with the reward differential, then we would expect an increase in the relative frequency of the specified response variation. The present experiments deal with the selective reinforcement of the temporal characteristics of a behavior chain. Rats were reinforced according to the time they took to traverse a straight runway. For convenience, we shall refer to the time the rat takes to reach the end box as "response time" and the reciprocal of this time as "response speed." Such measures do not imply anything about the form of the response; e.g., a long response time of 5 sec. (say) can be achieved by a large number of behavior sequences other than "running forward at a steady slow pace." In the studies to be reported the delay of