There is an important and active research programme at the laboratory and clinical level to develop indicators of the ultimate local response of tumours to radiation. In order to be of great value clinically indicators should yield a high true positivity and a low false positivity so that revisions of therapeutic strategy will be made only when there is a real need for such a change. The available data in the literature from the clinical studies, when analysed in terms of the true and false positive rates, indicate that the extent of regression at the completion of external beam therapy is not a useful prognostic indicator. This pertains to populations of tumour of a specific histopathological type, tumour size and anatomic site. Studies of laboratory animal tumour models have shown that regression patterns may be useful prognostic indicators for a tumour which is characterized by moderate immunogenicity and where there is close correlation between complete regression and permanent control. In contrast for tumours which are weakly or nonimmunogenic and which regress completely even at low tumour control probabilities, the pattern of regression has not been demonstrated to be of prognostic value.