All 72 resections for colorectal liver secondaries during the period 1971-1984 were analysed retrospectively. Liver tumours were single in 35 (49 per cent), unilateral in 55 (76 per cent) and associated with extrahepatic disease in 12 (18 per cent) patients. Operative mortality was 5.6 per cent. With respect to the disease in the liver, the presence or absence of four or more metastases was the predominant prognostic determinant with a 5 year survival rate of 20 per cent in patients with less than four liver tumours, and no 3 year survivor among patients with four or more tumours. When the number of liver tumours was less than four, the prognosis in patients with unilateral disease was not significantly better than in patients with bilateral disease (P = 0.19). No other liver disease variable seemed to play any role in the prognosis. Extrahepatic disease was associated with a poor prognosis and no 5 year survivor. The length of the tumour-free resection margin was the only treatment variable that varied with the outcome: a resection margin of less than 10 mm was followed by a poor survival. Variables that did not influence survival included uni- or bilateral disease, liver tumour volume, tumour size, type of liver resection, Dukes' classification, differentiation of the primary tumour and synchronous or metachronous disease. It is concluded that resection for liver colorectal secondaries is indicated when there are less than four liver tumours, even if bilateral, no extrahepatic disease is present, and a resection margin of at least 10 mm can be obtained. It should not be performed unless all of these requirements are met.