In the shuttlebox self-stimulation test described by Atrens, rewarding brain stimulation is independently initiated and terminated by rats. It has been hypothesized that gradually accumulating aversiveness of stimulation motivates the rat to terminate the rewarding stimulation train. In aggreement with this view, optimal doses of the known anxiolytics, pentobarbital (5 and 10 mg/kg) diazepam (1 and 2.5 mg/kg), chlordiazepoxide (3 and 5.4 mg/kg) and CL 218,872 (3, 10 and 30 mg/kg) preferentially increased the latency to terminate stimulation (the OFF latency), as compared with the latency to initiate stimulation (the ON latency). Higher doses increased both latencies in a nonselective fashion, suggesting nonspecific performance impairment. The shuttlebox self-stimulation test constitutes a potentially useful measure of experimental approach-avoidance conflict, with the OFF latency indicating anticonflict activity and the ON latency providing a control for performance variables.