A retrospective comparison of cancer incidence data and, where relevant, population data with 16,955 first-time users (patients, relatives and friends) of a national cancer information service (CancerBACUP) during the period April 1995 to March 1996 is presented. The number of events observed was compared with the number of events expected, were the national rates of cancer incidence and population demographics apply. Standardized incidence ratios (SIRs) (observed - expected ratios) were used to indicate any differences. Statistically significant differences (P < 0.001) in the observed and expected sex, age and primary site distribution of patients enquired about were found. Statistically significant differences (P < 0.001) were also identified for the age, employment status, socioeconomic class and geographical location of first-time enquirers (patients, relatives and friends). Enquiries about brain, testis and breast cancers and non-Hodgkin's lymphoma (NHL) were substantially higher than expected; enquiries about bladder, lung, stomach and colorectal cancers were much lower than expected. As the service is provided via a freephone number, it is available to all, and users might be expected to be randomly distributed across the variables listed. The underlying reasons for the differences identified need to be investigated, and the role of information in the care of cancer patients should be formally evaluated.