&NA;To examine the reliability of clinical features as possible components of a clinical method of prognostic staging for patients with rectal cancer, the results obtained when two pairs of physicians interviewed and examined a consecutive series of 60 patients with unresected rectal cancer were compared. High levels of agreement between physicians were found for most items of the patient's history, on whether the lesion was palpable, on the distance of the tumor from the anus, on the location and circumferential extent of the tumor, and on whether the tumor was fixed or mobile. Although physicians elicited similar physical signs, thier use of terms to describe the rectal lesions varied widely unless agreed‐upon criteria were used. There was 63 per cent agreement on the use of descriptive terms among physicians using criteria and only 13 per cent agreement among physicians not using criteria. These results show that prognostically important clinical features in patients with rectal cancer can be reliably identified and are thus suitable for inclusion in methods of clinical prognostic staging for this disease.