Dose-response profiles were determined for chlordiazepoxide, diazepam and N-desmethyldiazepam in a squirrel monkey punishment (conflict) procedure. The monkeys were trained to lever press under a food-maintained concurrent schedule consisting of an unpunished 6-minute variable interval (VI) schedule, and a 1.5-minute VI schedule, on which responses were punished intermittently (24 response variable ratio) with electric footshocks. The three benzodiazepines effectively increased responding that had been suppressed by punishment; they had inverted U-shaped dose-effect curves. The minimum effective doses for increasing punished responding were: diazepam less than or equal to 0.31 mg/kg p.o.; N-desmethyldiazepam = chlordiazepoxide = 0.62 mg/kg. As a model to assess potential antianxiety activity, this procedure possessed excellent sensitivity and reliability. The following observations were also made. 1) During initial training, as shock intensity was increased and punished responding became suppressed, some monkeys exhibited an increase in unpunished response rates. This may have represented "positive behavioral contrast," but response rate changes were associated with changes in the amount of time the monkeys allocated to each schedule. 2) At certain dose levels, all three compounds exerted antipunishment effects 24 hours after administration. 3) As was reported previously for rats, when the monkeys had no previous drug experience ("drug-naive") they were more sensitive to the depressant effects of the benzodiazepines. With repeated administration, there was a reduction in this sedation and a concomitant increase in the antipunishment effect. This phenomenon was dose- and animal-dependent.