Five different groups of rats were trained in a food-motivated, bar-pressing task to discriminate from the nondrug condition (physiologic saline, i.p.) the effects produced by i.p. injections of ethanol (330, 660, or 990 mg/kg), sodium barbital (80 mg/kg), or sodium phenobarbital (25 mg/kg). The establishment of highly effective discriminations required 20--40 training sessions for all drugs, with the exception that rats trained with 330 mg/kg of ethanol required 80--100 training sessions. After the drug-nondrug discriminations were well established, cross tests revealed that ethanol did not elicit drug-appropriate responding in the groups trained with sodium barbital or sodium phenobarbital. However, both barbiturates elicited drug-approppriate responding to some extent in rats trained with ethanol as the discriminative stimulus. With barbital, the greatest generalization was observed in rats trained with the low dose of ethanol (330 mg/kg). The findings emphasize the need for the use of several training doses and for transfer tests in both directions when the stimulus effects of drugs are compared.