The elevated plus-maze test has been in use as a rodent model of anxiety for a decade, and is representative of those tests that are based upon the study of spontaneous behaviour patterns and which have high ecological validity. The origins of the test in studies of the relationship between exploration and fear are reviewed, and attention is drawn to the distinct possibility that variation in the pharmacosensitivity of the procedure may be attributable to often extreme methodological variation between laboratories. In considering further this issue, attention is also drawn to the need to collect data under constant test conditions and to provide the minimum database necessary to reach conclusions regarding the behavioural specificity of drug action. Recent research, which has extended the conventional plus-maze scoring technique to include specific behavioural acts and postures (in particular, those relating to defensive behaviour), is described. The value of such an ethological approach to the plus-maze is then exemplified with original data that demonstrate behaviourally selective, anti-anxiety effects of the GABAA receptor agonist, muscimol (0.125-1.0 mg/kg). It is concluded that, when used appropriately, the elevated plus-maze test can be a very valuable tool in drug screening and in the study of the neurobiology of anxiety and defence. More attention to behaviour and somewhat less emphasis on test simplicity and convenience would seem to be warranted.