In order to analyse further the discriminative stimulus properties of stimulant drugs, rats were trained to discriminate 2.0 mg/kg of dl-cathinone in a two-lever operant task. Dose-related generalization was seen to cathinone itself and to a wide range of stimulant drugs including d-amphetamine, cocaine, methylphenidate, pipradrol and cathine, i.e. (+)norpseudoephedrine. The high degree of specificity of the cathinone cue at the specific training dose studied was shown by the fact that the following nonstimulant drugs failed to generalize at all to cathinone, even in large doses--haloperidol, chlordiazepoxide, fenfluramine and fentanyl. The cathinone cue at 2.0 mg/kg was probably of central origin because phydroxyamphetamine (a polar congener of amphetamine) failed to generalize to cathinone at a dose nearly 50 times the ED50 for amphetamine generalization. Phenylethylamine (PEA; alpha-demethylamphetamine) and deuterated phenylethylamine (alpha, alpha, d2-PEA), a long acting derivative of phenylethylamine which is resistant to metabolism by monoamine oxidase, produced at most partial (60%) generalization to cathinone, even in large doses. alpha-Demethylcathinone failed to generalize at all to cathinone at doses up to 10 times the ED50 for cathinone. Thus, the alpha-methyl groups of both amphetamine and cathinone are important in determining their cue properties. The involvement of dopaminergic systems in the cathinone cue was investigated by examining generalization to apomorphine and antagonism by haloperidol. Apomorphine produced at most 29% generalization to cathinone. Haloperidol, at doses up to 0.3 mg/kg, produced at most 50% antagonism of both the cathinone cue and of the ability of amphetamine to substitute for cathinone. It is suggested that the evidence for dopaminergic mediation of the cue properties of cathinone and of other CNS stimulants is somewhat tenuous, whilst endogenous phenylethylamine may play some part in the mediation of the stimulant cue. Haloperidol, alone and in conjunction with amphetamine or cathinone, produced a remarkable tendency for subjects to emit a greater proportion of their total responses on the inoperative rather than the operative lever than was seen after saline or injections of vehicle. This action of a neuroleptic drug suggests, in accordance with Colpaert, Niemegeers and Janssen (1977), that in drug discrimination antagonism studies involving neuroleptics, and perhaps other drugs, quantal (lever selection) rather than quantitative (percentage of responses on the drug lever) indices may be the procedures of choice.