There is a need for novel anxiolytics, which are effective, but do not cause sedation, tolerance, and rebound anxiety on discontinuation. To investigate a procedure that can be used to assess these characteristics preclinically, rats were initially trained to press a lever at a high rate to obtain food. Once trained, periods of punishment were introduced in which electric shocks were superimposed. The intensity of these electric shocks was increased every 90 s from very low (0.01 mA) to sufficiently high to stop most subjects responding (0.4 mA), so that a complete rate/intensity function was obtained during each punishment period. The benzodiazepine, chlordiazepoxide, and two novel subtype-selective gamma-aminobutyric acid-A agonists, TP003 and TPA023, significantly increased response rates mildly suppressed by intermediate levels of electric shock without any effect on unpunished response rate. Two clinically anxiogenic agents, yohimbine and flumazenil, reduced the rate of punished responding. Aripiprazole and amphetamine reduced both punished and unpunished responding. Repeated treatment with diazepam 2.5 mg/kg daily for 15 days, initially markedly reduced unpunished response rates, but also increased punished response rates, an effect which became greater with repeated treatment. Abrupt cessation of diazepam treatment produced a reduction in punished responding. Diazepam (5 mg/kg daily) produced a greater reduction in unpunished responding, a smaller increase in punished responding, and a larger and longer lasting reduction in punished rates on withdrawal. In conclusion, the procedure detected anxiolytic and anxiogenic effects of drugs, and the sedative side effects, development of tolerance, and rebound-anxiety on discontinuation of a benzodiazepine. This procedure should have utility in the characterization of novel treatments of anxiety.