Between 1 January 1984 and 31 December 1992, 66 patients with hepatic metastases from colorectal carcinomas underwent liver resection. 40 of these patients had synchronous hepatic metastases, and liver resection was carried out simultaneously with radical resection of the primary tumour; in 26 cases metachronous metastases in the liver were surgically removed. 25 patients had an anatomical resection and the remainder underwent atypical resections. The postoperative mortality rate was 4.5% and the major complication rate was 19.7%. Univariate and subsequently multivariate analyses were used to predict the influence of various clinical, histopathological and surgical variables. The observed 5-year surival rate was 29.6% and the 5-year disease-free survival rate 13.9%. Furthermore, the observed median survival time was 24.7 months and the mean disease-free survival time was 16.7 months. Multivariate analysis showed that stage of primary (pTN) (P = 0.043), tumour grading (P = 0.013) and site of primary (P = 0.007) were factors which independently influenced 5-year disease-free survival whereas stage of primary (pTN) (P = 0.008), tumour grading (P = 0.004) and type of resection (P = 0.035) were identified as having independent influence on 5-year observed survival. We consider liver resection to be an effective form of treatment for patients with resectable liver metastases from colorectal carcinoma, although the overall chances for cure are generally not very promising. It appears that the biological behaviour of the primary tumour, in terms of tumour stage and grading, has the greatest influence on survival.