It has been suggested by Philip (1969) that more information could be gleaned from conventional drug trials if it were possible to assess the efficacy of treatment for each patient in the trial. A method for the statistical analysis of individual response to treatment has been put forward by Philip using a modification of Ferguson's nonparametric trend analysis of correlated observations (Ferguson, 1965). Sutherland, Sutherland and Philip (1967) have described the details of a double-blind drug trial in which the efficacy of Prondol, an imipramine derivative, was compared with a standard imipramine preparation. The patients in their trial were rated on the Hamilton scale for depression (Hamilton, 1960) on admission, after two weeks' treatment and after four weeks' treatment. Since the data were not amenable to analysis by parametic methods, Mann-Whitney U tests (Siegel, 1956) were calculated to ascertain whether the change scores between rating occasions were different for the two groups of patients. This cumbersome procedure showed that there was no difference in change scores between the trial drug group and their controls for the 0–2 week and 0–4 week comparisons; there was, however, a significant tendency for the control drug group to show significantly greater change between the two week and four week assessments.