The authors have reviewed their experiences in determining the presence of liver metastases in 103 patients by radiocolloid scanning. The sensitivity of liver scanning proved to be quite low if the presence of focal defects in the distribution of the tracer was chosen as the diagnostic criterion. The inclusion of less restrictive criteria such as liver enlargement or irregular distribution of the tracer, resulted in a higher sensitivity without lowering the predictive value of a negative scan. Using the latter diagnostic criterion, sensitivity, specificity and accuracy were in the range of 90%. Abnormal liver scans are common in patients classified at T3-T4 or N+ and their chances of being "true positive" are high. Conversely, abnormal scans are seldom found in patients classified at T1-T2 or N0 and probabilities of "false positive" results are high.