Undergraduates given accurate instructions pressed keys for token points under either a variety of reinforcement schedules (variety training) or under a single schedule. Response rates on a fixed-interval (FI) test schedule then were assessed. Experiment 1 compared variety training inclusive of FI-optimal rates (functional) to training excluding such rates (nonfunctional). Participants provided functional training showed low test rates relative to those provided nonfunctional and single-instruction training, implicating response-rate history as a determinant of subsequent FI performance. Experiment 2 manipulated functional and nonfunctional variety training, and the correspondence of FI test-phase point densities with those of high-, low-, or both high- and low-rate training components. Performances under FI were affected by density correspondence, suggesting discriminative control of response rate by point density. (Contains 4 tables and 5 figures.)