Stereoisomers of the N-methyl-D-aspartate antagonist dizocilpine (MK-801) were studied to determine whether behavioral effects on complex operants depend on reinforcement loss accompanying behavioral disruption. Rats earned food pellets if the run of consecutive left-lever presses preceding a trial-terminating right-lever press approximated a target of 12. A percentile schedule reinforced any run closer to the target than two-thirds of the runs on the most recent 24 trials. Once the sequence was learned, half the subjects were shifted to a procedure that yoked reinforcement for each length run to the probability that length generated pellets during asymptotic percentile performance. Although these two procedures generate similar control run and reinforcement distributions, disrupting behavior reduced reinforcement probability far more under the yoked than the percentile procedure. Despite this difference in drug-induced reinforcement loss, both enantiomers produced similar dose-related decreases in run length and response rate under both procedures, with the (-) isomer approximately one log unit less potent than the (+) isomer. The absence of differential effects under these procedures diminishes the likelihood that reinforcement loss contributes to dizocilpine's effects, indirectly bolstering claims that dizocilpine directly affects learning.