The social interaction test in rats provides a method for detecting anxiolytic activity that does not use food or water deprivation, or electric shock, and therefore obviates difficulties of interpretation that might arise from drug-induced changes in motivation. Since social interaction is measured under more than one test condition any overall increase or decrease in social behaviour can be detected independently from the drug x test condition interaction that characterizes an anxiolytic drug. The Geller-Seifter conflict test was designed with two schedules of reinforcement for the same reasons. Any candidate test for anxiolytic action that examines drug effects under only one experimental condition is open to misinterpretation and may also prove unreliable if the critical experimental factors ( e.g. the level of food deprivation or the shock intensity) are changed. The testing procedure in the social interaction test is relatively time consuming in terms of observer-hours, but no lengthy pretraining of the animals is required. There is no way of fully automating the scoring and therefore it is important that the observers do not know the experimental group of the rats that they are scoring, and that tape recordings are made so that the scores can be checked. It has not so far been fruitful to analyze drug effects on every individual social behaviour, but this method does allow changes in individual behaviours to be detected. By entering the data directly into a computer we are now able to store the frequency and duration of each behaviour as well as the sequence of behaviours. It will then be possible to determine whether a detailed analysis of drug effects on the patterning of social behaviours will prove a useful addition to the social interaction test