||This study investigated the effects of punishment on one operant as a function of the relative reinforcement density for an alternative, unpunished operant. Thirty male undergraducates were randomly assigned to five groups which differed according to the concurrent VI reinforcement schedules used to maintain punished (R1) and alternative (R2) behavior, respectively. In each of the groups, the density of reinforcement for R1 was either equal to or greater than that for R2. The experimental procedures were constant for all groups and involved repeated measures of the subjects' performance across periods of acquisition training, punishment, and recovery. In acquisition, rate differences in R1 responding were an increasing function of reinforcement density for R1. When noise punishment was superimposed for R1, suppression generally increased as the rate of reinforcement for R2 increased. Furthermore, punishment accelerated R2 responding in most groups. This shift in preference to unpunished behavior was an increasing function of the relative reinforcement rate scheduled for the response. Following the termination of punishment, R1 responding not only recovered, but usually surpassed the replenishment baseline throughout most of the recovery period. This punishment contrast effect was shown by all groups in which the distribution of reinforcement for R1 was greater than that for R2. Additionally, punishment contrast appeared to be an increasing function of the reinforcement rate for punished behavior. These results indicate that reinforcement variables interact with punishment to determine both the amount and permanence of suppression. A comparison of the present findings with those reported in previous experiments suggests that the use of independent, concurrent reinforcement schedules, rather that a single schedule for both punished and alternative behavior, reduced the overall effectiveness of punishment as a response suppressor. Some implications of the results for the efficient use of punishment in behavioral control are briefly discussed.