Pigeons were trained to peck a key on a variable-interval 2-min schedule of food reinforcement. Prior to each session, either 2.0 mg/kg methadone (n ϭ 3), 3.0 mg/kg cocaine (n ϭ 4), or 5.6 mg/kg cocaine (n ϭ 2) was administered. When each pigeon's rate of pecking was stable, a range of doses of the training drug and saline were administered prior to 20-min extinction sessions separated by at least four training sessions. Rate of pecking during these extinction tests was generally an increasing function of dose, with the lowest rates obtained following saline and low doses and the highest rates obtained following doses near the training doses. Dose functions from pigeons trained with 5.6 mg/kg cocaine were steeper than those from pigeons trained with 3.0 mg/kg cocaine. Pigeons trained with methadone or 3.0 mg/kg cocaine were then given discrimination training, in which food reinforcement followed drug administration and 20-min extinction sessions followed saline administration. Rates of pecking under these conditions quickly diverged until near-zero rates were obtained following saline and high rates were obtained following drug. Discrimination training steepened dose functions for the training drugs, and the effects of several other substituted drugs depended on the pharmacology of the training drug. The pigeons trained with 5.6 mg/kg cocaine were tested with d-amphetamine, methadone, and morphine prior to discrimination training. d-Amphetamine increased rates dose dependently, and methadone and morphine did not. The results suggest that discriminative control by methadone and cocaine was established without explicit discrimination training.Key words: drug discrimination, state-dependent learning, stimulus control, methadone, cocaine, key peck, pigeonsIn the classic study of stimulus generalization by Guttman and Kalish (1956), different pigeons were trained to peck a key lit from behind with lights of various wavelengths. Pecks were reinforced with food according to a variable-interval (VI) 1-min schedule. When a range of other wavelength stimuli were presented during 30-s extinction periods, inverted U-shaped generalization gradients were obtained, with the highest rate of pecking occurring during the wavelength presented during training (Sϩ). This study introduced methods for the study of generalization in individual subjects, and inspired much research on factors that determine the form of generalization gradients (see Honig & Urcuioli,