Human participants played a computer game in which choices among five alternatives were concurrently reinforced according to dependent random-ratio schedules. "Dependent" indicates that choices to any of the wedges activated the random number generators governing reinforcers on all five alternatives. Two conditions were compared. In the hold condition, once scheduled, a reinforcer -worth a constant five points -remained available until it was collected. In the decay condition, point values decreased with intervening responses, i.e., rapid collection was differentially reinforced. Slopes of matching functions were higher in the decay than hold condition. However inter-subject variability was high in both conditions.