From 1974 to 1984, 442 consecutive patients with carcinoma of the uterine cervix (FIGO IIB: 139, IIIA:10, IIIB:221, IVA:72) were referred for combined intracavitary (IRT) and external radiotherapy (ERT). To improve local control and reduce late rectosigmoid morbidity the treatment strategy was changed from continuous (CRT) to split-course radiotherapy (SCRT) in 1978. Stage by stage the 5-year actuarial estimates of survival, local control, and late morbidity did not differ in relation to strategy. In the patients with tumours larger than 8 cm, the SCRT involved an increased dose in point B, a reduced dose in point A from the IRT, a lower total dose in point A, and a 34 days' prolongation of the total treatment time (TTT). The resulting 5-year actuarial local control rates were significantly lower compared with those after CRT. No difference of late severe morbidity was found except in IVA patients. In the patients with tumours between 4 and 8 cm, the SCRT involved a reduced dose in point A from the IRT, an increased total dose in point A and B, and a 50 days' prolongation of the TTT. In patients with stage IIB, the 5-year actuarial central local control rate was lower (p = 0.06), and the 5-year estimate of late severe morbidity significantly higher after SCRT compared with CRT. It is concluded that the increase of the dose in point B in the SCRT was insufficient to prevent the deleterious effect on local tumour control of either the lower dose from IRT in point A, or the prolonged TTT. The increase of the total dose in the SCRT may explain why the late morbidity was not reduced, and may suggest that the TTT is of no significant importance for the risk of late normal tissue damage.