In each of different groups of rats trained to discriminate either 8-OH-DPAT, DOI, d-amphetamine, cocaine, chlordiazepoide or ethanol from saline, dizocilpine produced maximum percentages of drug lever (DL) selection that were intermediate between those produced by the training conditions. Dizocilpine also decreased DL selection produced by the training dose in each of the discriminations, except in ethanol-trained rats. In all discriminations, with the exception of ethanol-trained rats, the intermediate levels of DL, selection produced by dizocilpine were associated with increased FRF values (sum of the responses made on either lever before the first reinforcement occurred), increased lever selection latencies, and increased responding on the nonselected lever. At doses that, in general, had effects on response rate similar to those of dizocilpine, intermediate levels of DL selection were produced by BMY 7378 in 8-OH-DPAT-trained rats, by WY 50,324 in DOI-trained rats, by (-)-3-PPP in d-amphetamine- and in cocaine-trained rats, by alpidem in chlordiazepoxide-trained rats, and by PCP in ethanol-trained rats. The intermediate levels of DL selection produced by these latter drugs were not associated with simultaneous increases of FRF values, selection latencies, and responding on the nonselected lever. The results suggest that dizocilpine produces intermediate levels of drug-appropriate responding through the behavioral mechanism of partial generalization only in ethanol-trained rats; in all other discriminations examined here, the effects of dizocilpine appear to involve (1) pharmacological effects that differ from those of the training drug, and (2) behavioral mechanisms that are unrelated to stimulus generalization. The differentiation of partial generalization and other mechanisms whereby intermediate responding can occur in the drug discrimination paradigm requires analyses that are more detailed than those commonly used in drug discrimination research.