Palliative treatment of unresectable colorectal liver metastases is common and often justified with reference to historical data on the natural history of the disease. However, in view of the improved diagnostic accuracy of modern imaging techniques, these previously published series do not provide sufficient guidance to judge the prognostic efficacy of palliative treatment. In the late 1970s we started prospectively to collect data on consecutive patients with colorectal liver metastases according to a standard protocol. We now present data derived from this series on factors that may affect outcome in untreated patients. Between January, 1980, and December, 1990, 1099 consecutive patients were recorded, of whom 566 (51.5%) received no treatment for their hepatic tumour. Excluding 34 early deaths and 48 patients with a second malignant tumour, 484 patients provided the basis for analysis. All patients were followed up to July 1, 1993, or death. At the closing date of the study only 1 untreated patient was still alive. The impact of various factors on survival was analysed by univariate and multivariate analyses. Six independent determinants of survival were identified in the following order: percentage liver volume replaced by tumour (LVRT), grade of malignancy of the primary tumour, presence of extrahepatic disease, mesenteric lymph-node involvement, serum carcino-embryonic antigen, and age. The subsequent combination of the independently significant factors, separately for patients with up to or more than 25% LVRT, yielded a prognostic tree that displayed median survival times of various subgroups of 3.8 to 21.3 months. These findings provide a framework to estimate the survival expectancy of untreated patients, thereby allowing improved assessment of the prognostic significance of palliative therapeutic approaches.