in operant conditioning, a procedure in which the likelihood that a response will be reinforced depends on the response exceeding (or being less than) a value based on a distribution from previous responses. Usually, the distribution is based on some set of the most recent responses, and it is updated with each response. For example, the current response might be eligible for reinforcement if its peak force falls above the 90th percentile of the distribution of forces from the previous 50 responses. The most recent force then replaces the earliest one in the distribution so that the distribution remains based on 50 entries.