In each of three components of a multiple schedule, monkeys were required to emit a different sequence of four responses in a predetermined order on four levers. Sequence completions produced food under a fixed-ratio schedule. Errors produced a brief timeout. One component of the multiple schedule was a repeated-acquisition task where the four-response sequence changed each session (learning). The second component of the multiple schedule was also a repeated-acquisition task, but acquisition was supported through the use of a stimulus fading procedure (faded learning). In a third component of the multiple schedule, the sequence of responses remained the same from session to session (performance). Heroin, methadone, cyclazocine and phencyclidine each produced dose-related decreases in overall response rate. At high doses which produced equivalent rate-decreasing effects, cyclazocine and phencyclidine generally produced greater disruption of accuracy in the learning component than did heroin or methadone. Naltrexone 5.6 microgram/kg shifted to the right by approximately 1/2-log unit the heroin and methadone dose-effect curves, but produced little or no change in the cyclazocine dose-effect curves. At 56 micrograms/kg naltrexone completely antagonized both the rate-decreasing and error-increasing effects of heroin and methadone. The same dose of naltrexone tended to produce greater antagonism of the effects of cyclazocine on accuracy than on rate, which was shifted by only 1/4-log unit. In contrast, naltrexone failed to antagonize the effects of phencyclidine on either rate or accuracy. Thus it would appear that while cyclazocine and phencyclidine produce similar disruptions in the accuracy of a discrimination, the effects of each are differentially sensitive to antagonism by naltrexone.