Three adult baboons were trained using a psychophysical procedure to discriminate between different synthetic vowel sounds [symbol: see text]. Baboons pressed and held a lever down to produce a pulsed train of a single reference vowel that served as the standard stimulus. Animals were trained to release the lever only when this standard vowel sound changed to one of the four remaining comparison vowels. A lever release within 1.5 s of this change in vowel sounds was defined as a correct detection of the change from the standard vowel to one of the comparison vowels, and was reinforced. All baboons readily learned the vowel discriminations and detected vowel changes at the 90-100% correct performance level. Acute IM administration of cocaine prior to test sessions (0.00032-3.2 mg/kg) produced dose-dependent decrements in vowel discriminability. At the same time, cocaine shortened lever release latencies (reaction times) to the vowel stimuli in two of three baboons. The cocaine-induced decrements in vowel discriminability were correlated with the degree to which frequency differences occurred among the different vowels in that lower vowel discriminability scores were found for those vowels with smaller spectral differences from the standard vowel. Further, false alarm rates were not systematically affected by cocaine, indicating that the cocaine-induced decrements in vowel discrimination accuracy occurred in the absence of systematic changes in the reliability of the baboons' discrimination performances.