From 1960 to 1987, 1209 patients with colorectal liver metastases were recorded, and followed until 1 January 1990. In 242 cases the diagnosis was based on external imaging, whereas 967 patients had operative confirmation and staging of their liver disease. Three groups of patients were analysed: group 1 involved 921 cases, of whom 902 were deemed non-resectable whereas 19 could not be unequivocally classified. Only 21 patients lived for longer than 3 years, seven survived for 4 years, but there were no 5-year survivors. Group 2 comprised 62 highly selected patients who at laparotomy demonstrated resectable metastatic spread confined to the liver, but this was not treated mainly because of a formerly different therapeutic approach. These patients had a significantly longer median survival time (14.2 versus 6.9 months), but also failed to achieve 5-year survival. The 226 patients forming group 3 underwent hepatic resection with intent to cure. Nine of them had minimal macroscopic disease left, and 34 with all gross tumour removed had positive margins. Survival of patients with these 43 eventually non-radical resections followed an identical course as in group 2 (median survival 13.3 months, maximum 42 months). Of the 183 patients with potentially curative resection ten died after surgery (5.5 per cent). Actuarial 5 and 10-year survival rates in the remaining 173 patients were 40 and 27 per cent with 25 and seven patients alive at respective periods of time. Until 1 January 1990, 64 patients remained free from recurrent disease for up to 24 years. In three patients the tumour status at death was unclear. The other 106 patients developed definite cancer relapse. Nevertheless they demonstrated a prolongation of survival time by a median of 1 year when compared with the 43 non-radically resected patients or the 62 untreated patients with resectable liver-only metastases, and accomplished a maximum survival time of 8 years. Radical excision of colorectal secondaries to the liver therefore offers effective palliation, and in a small number the chance of a cure.