A very large number of therapeutic trials of antidepressant drugs have been reported in the scientific literature.Until now, the comparison of one drug with another, or with placebo, has been performed typically by comparing the scores on depression rating scales of the two groups of patients at fixed points of time after the beginning of therapy. It was postulated in 1989 that the curves of the recovery scores followed an exponential curve of the formula y = ae-bx+ c.This hypothesis was tested in a double-blind controlled trial of the antidepressant minaprine, with the use of the scores on the Hamilton Rating Scale for Depression (HAM-D). We found that the correlation coefficient, Pearson's r, between the log of the HAM-D value and the week number of the study was -0.99. This gives a coefficient of determination of 0.98, which makes it clear that the model adequately fits the data. We conclude that the use of the formula gives a method of testing the statistical significance of the difference between treatments as a valuable alternative to traditional tests. We believe that this would give a much more sensitive discrimination between treatments because all of the data points are used to calculate a single parameter-the slope of the curve. (J Clin Psychopharmacol 1996;16:420-424).