The purpose of the first experiment was to investigate whether diazepam could acquire anxiogenic properties by signalling an aversive event. Rats were trained in an operant chamber in the pentylenetetrazol (PTZ) model of anxiety. Thereafter the animals were divided into groups that received classical diazepam conditioning (Group 1), and conditioning of diazepam + tone (Group 2). In the test phase diazepam was injected prior to placement in the operant chamber. Group 2 selected the PTZ‐appropriate lever more often than the other groups, indicating that the tone induced anxiety, and diazepam did not. Tones and shock may therefore be more easily associated than diazepam and shock. The second experiment investigated this. Rats were trained the same way as in the first experiment. Thereafter the experimental group received injections of a small dose of diazepam prior to a second injection of a large dose of diazepam. The hypothesis was that a compensatory anxiogenic conditional response to diazepam's anxiolytic effect should be elicited by the small dose. There were no differences between the groups in lever selection, indicating that a compensatory anxiogenic response was not elicite