In a two-choice drug discrimination procedure, rats were trained to discriminate either 0.056 (low dose) or 0.17 (high dose) mg/kg of the kappa opioid bremazocine from saline. Substitution and antagonism tests were then conducted with a variety of opioids. The opioids U50, 488, ethylketocyclazocine, spiradoline, enadoline, and U69,593 substituted completely (>/=80% bremazocine-appropriate responding) for the bremazocine stimulus in both training groups, a finding consistent with their high-efficacy kappa profile. In contrast, the other opioids examined could be classified into different subsets on the basis of their patterns of substitution and antagonism: (1) (-)-cyclazocine substituted completely for the low-dose bremazocine stimulus and substituted partially (approximately 50% bremazocine-appropriate responding) for, and antagonized partially, the high-dose bremazocine stimulus; (2) butorphanol, nalorphine and nalbuphine substituted partially for, and antagonized partially, the low- and high-dose bremazocine stimuli; (3) levellorphan did not substitute for either training dose of bremazocine and antagonized the bremazocine stimulus at both training doses; (4) (-)-N-allylnormetazocine [(-)-NANM] substituted partially for the low-dose bremazocine stimulus and substituted completely for the high-dose bremazocine stimuli. Despite the high levels of substitution, (-)-NANM antagonized both the low- and high-dose bremazocine stimuli at doses below those required to produce its maximal agonist effects. The substitution patterns exhibited by (-)-NANM were not antagonized by doses of naloxone that antagonized the stimulus effects of (-)-cyclazocine and bremazocine, suggesting that these substitution patterns were non-opioid mediated. Collectively, these results suggest that the different patterns of substitution and antagonism observed with these opioids were due either to differences in their intrinsic efficacy at the kappa receptor or to a non-opioid component of action.