Numerous factors may contribute to the effectiveness of conventional physiotherapy. In this study we have minimized the influence of such factors by application of scientific method in trial design, concentrating on particular components of therapy. Patients with osteoarthrosis of the knee referred for conventional therapy were randomly allocated to three treatment groups, ice, short-wave diathermy, and untuned short-wave diathermy as a control. It is shown that by three months most patients, whatever the therapy, improved significantly in six out of the seven outcome measures. However, the group receiving ice made better early progress and by three weeks the improvement in the pain score was similar to that attained after three months by the other groups. It is emphasized that trials such as this seek to provide guidelines for the therapist rather than didactic formulae.