Twenty-three patients were studied to assess the reliability of carcinoembryonic antigen (CEA) levels in following the course of colorectal cancer. CEA estimations were made prior to surgery and again postoperatively. The resected specimens were allocated a Dukes' Stage and histological grading (well, moderate or poorly differentiated). In addition, sections were stained for the presence of CEA by an immunoperoxidase method. Of the 23 patients, twelve had either disseminated disease at initial surgery or subsequently developed metastasis/recurrence. Eleven remain disease-free at a minimum follow-up of one year. In all of these the reliability of plasma CEA values in reflecting the disease status has been assessed. No false positive elevations of CEA were found. Three factors emerge as positive predictors of CEA estimation reliability: pre-operative CEA elevation; tumour grading as well differentiated; dark staining for the presence of CEA. These factors identified 15 of the 18 patients (83%) in whom CEA appeared reliable and were not present in any of the five patients where CEA was not reliable. This reliability achieves statistical significance (X2 = 8.5, p less than 0.02). Histological demonstration of CEA may contribute to the reliability placed on plasma CEA estimations and should be considered if serial estimations are to be performed.