The effects on memory of drugs of abuse from several pharmacological classes were examined in four adult male squirrel monkeys responding under a delayed matching-to-sample schedule of food presentation. Subjects were required to emit 20 responses on a sample key transilluminated by either a constant white or a flashing blue light. The twentieth response initiated a 3-second delay followed by presentation of two comparison stimuli. If a response was made to the key that matched the sample stimulus (correct match), a single food pellet (97 mg) was delivered. Pentobarbital (0.32-10 mg/kg), diazepam (0.1-5.6 mg/kg), phencyclidine (0.01-0.32 mg/kg) and cocaine (0.1-3.2 mg/kg) dose-dependently reduced accuracy of matching performance towards chance levels. Amphetamine (0.01-1.0 mg/kg) resulted in a small, but statistically significant, reduction in accuracy at a dose of 0.56 mg/kg, while 1.0 mg/kg completely suppressed responding. Analyses indicated that pentobarbital, diazepam and cocaine produced either position or color biases in responding, and in some cases these biases in responding were associated with decreases in accuracy. No such response biases were observed with phencyclidine or D-amphetamine. These results suggest that drug effects on working memory performance can, in some cases, be the result of non-mnemonic processes. Thus, they illustrate the importance of examining behavioral endpoints in addition to task accuracy when interpreting drug effects on working memory in laboratory animals.