In naive rats benzodiazepines have clear anxiolytic effects in the elevated plus-maze, but in rats with previous plus-maze experience benzodiazepines are ineffective. This phenomenon does not depend on the drug state on trial 1 or on the inter-trial interval and generalises across mazes of different material; it is dependent on experience of the open arm. The phenomenon of "one-trial tolerance" to the anxiolytic effect of benzodiazepines is not seen in other animal tests; and there is no equivalent phenomenon of "one-trial withdrawal" or of tolerance to anxiogenic effects in the plus-maze. The phenomenon is not seen in unhandled rats, in rats given longer trials in the plus-maze, or in those given an amnesic treatment on trial 1. It is suggested that during the first 5 min in the elevated plus-maze the rats are acquiring a fear of heights and it is this phobic anxiety state that is measured during the second 5-min trial. Thus the nature of the anxiety generated by the maze is different on trials 1 and 2. The results of a factor analysis study confirm this conclusion.