A two-compartment activity box is described in which the actions of anxiolytic compounds may be readily detected. The box is designed with a partition which divides a dimly-lit black area and a brightly-illuminated white area. Free access to each section is enabled by means of a small opening in the partition. Measures of locomotion and rearing are recorded as well as transitions from black to white and vice versa. Time spent in each section is also measured. In this situation mice demonstrated a clear preference for the dim black section, a finding not inconsistent with the nocturnal nature of this species. By contrast, treatment with diazepam (0.125-5 mg/kg) or triazolam (0.01-0.1 mg/kg) produced a significant increase in activity in the white section. Locomotion and rearing in the black section showed a corresponding decrease. Sulpiride (0.5-20 mg/kg) and tiapride (0.5-40 mg/kg) had similar effects. These actions would appear to be independent of their effects on general activity, as amphetamine (1.25-2.5 mg/kg) increased and haloperidol (0.05-0.2 mg/kg) decreased overall exploratory behaviour in a non-selective manner. These data suggest that "two-compartment activity box" to be a model sensitive to the actions of clinically-active anxiolytics.