The aim of this study was to compare the utility of two recently identified tumour markers of pancreatic cancer, CA 19-9 and CAR-3, and to ascertain the roles of some factors influencing both antigens. CA 19-9 and CAR-3 were measured in sera of 18 control subjects, 27 patients with pancreatic cancer, 25 with chronic pancreatitis, and 29 with extra-pancreatic diseases. CA 19-9 and CAR-3 were, respectively, found to be increased in 85 per cent and 44 per cent of patients with pancreatic cancer, 28 per cent and 0 per cent with chronic pancreatitis and 72 per cent and 28 per cent with extra-pancreatic diseases. The ROC curves showed that, for any serum value considered, CA 19-9 is more effective than CAR-3 in discriminating between pancreatic cancer and control subjects and chronic pancreatitis. With the combined use of both antigens the results were no better than those given by CA 19-9 alone. Correlations were found between liver function tests and CA 19-9 levels and between cholestasis indices only and CAR-3 values. Our findings show that CAR-3 is not a sufficiently reliable marker of pancreatic cancer, due to its low sensitivity. Nor does it offer any more information than CA 19-9. Both assays are influenced, at least in part, by the extent of the neoplasia. Cholestasis which can greatly influence a serum glycoproteic marker such as CA 19-9, was found also to affect, to a lesser extent, CAR-3, an epitope on the same mucin molecule.