Previous studies have shown that, in rodents, chlordiazepoxide and other benzodiazepines can interfere with learning in passive avoidance or conditioned suppression procedures. The most consistent effects are observed when the drugs are administered before the acquisition trial and subjects are re-tested in the non-drugged state. It is not clear, however, whether this effect on learning is associated with the behavioural depressant actions of these drugs. In the present study mice were injected with chloridiazepoxide, diazepam, zopiclone, or CGS 9896 and locomotor activity measured in a two-compartment box. The animals were then enclosed in one of the compartments and received a series of footshocks. On a second trial, 24 h after the first, the mice were returned to the box without injection and locomotion and time spent in each compartment were measured. During trial 1 chlordiazepoxide, diazepam, and zopiclone produced dose-related decreases in locomotor activity. The same doses disrupted fear conditioning. CGS 9896 also interfered with the conditioning of fear but did not reduce exploratory activity during the first trial at any of a wide range of doses, showing that learning can be affected without direct behavioural depressant activity. In a further experiment, chlordiazepoxide and CGS 9896 disrupted fear conditioning when injected before trial 1 but not when injected immediately after this trial. Mice drugged with chlordiazepoxide or CGS 9896 before both trials 1 and 2 also showed disrupted conditioning, demonstrating that the drug effects cannot be interpreted in terms of state dependent learning.