In assessing the results of clinical treatment of a disease such as cancer, the radiotherapist is faced with a variety of difficulties. Although he may be convinced that many of the cases treated are cured, he is aware that late recurrences sometimes do arise, and that for this reason a symptom free period of even two or three years subsequent to treatment cannot be regarded as definite proof of a cure. The proportion of patients who remain symptom free for two or three years is not therefore a reliable measure of the proportion permanently cured. There is, however, the further difficulty that, owing to the high average age of patients suffering from cancer, a considerable number of deaths from intercurrent disease is to be expected if the analysis of the results is postponed for more than a few years. Long delay in making the analysis, while it leads to a conservative estimate of the proportion cured, is not only undesirable in itself, but may thus involve discarding a considerable number of patients who died without any recurrence of the disease and who were, in all probability, cured. Finally, there is the inevitable statistical uncertainty in estimating from the treatment results in a relatively small group of cases, the proportion who would be cured if a large number of cases was treated.Some of the difficulties referred to above are inherent in any clinical experiment designed to test a new technique of treatment, and cannot be eliminated by the most elaborate statistical methods.