Twenty-five music students taking classes in performance were randomly assigned either to an experimental group who received a course of 15 lessons in the Alexander technique, or to a control. A variety of measures were taken on four occasions: in both high and low stress situations before and after treatment. The experimental group showed improvement relative to the control on the following measures: overall music and technical quality as judged by independent experts blind to subjects' condition assignment, heart rate variance, self-rated anxiety and positive attitude to performance. However, with the exception of heart rate variance, these effects were restricted to performance in the low stress class situations. There were no significant effects on height, peak flow or misuse as judged on the basis of videorecordings of behaviour by independent experts in the Alexander technique.