To measure the accuracy of incidence rates for invasive cervical cancer derived from cancer registration data, a review of histopathology records was undertaken for all cases of invasive cervical cancer, a 20 per cent random sample of CIN 3 registrations and all cases of cancer of the uterus not otherwise specified (NOS) from Greater Manchester, which were registered by the North Western Regional Cancer Registry between 1988 and 1989. The subjects were 386 cases of invasive cervical cancer, 313 CIN 3 registrations and 69 cases of cancer of the uterus NOS. The main outcome measure was the incidence rate of invasive cervical cancer in Greater Manchester after adjusting for errors in cancer registration. It was found that 375 (97 per cent) of the 386 cases of invasive cervical cancer were verified against the original pathology report but 11 cases (3 per cent) had been registered as invasive cervical cancer in error. Nine cases of invasive cervical cancer were found to have been misclassified as CIN 3, and 10 cases as cancer of the uterus NOS. As the CIN 3 cases were a 20 per cent random sample of all CIN 3 registrations, the best estimate of the number of cases of invasive cervical cancer misclassified as CIN 3 is 45, which makes a total under-registration of 55 cases over the two-year period. After adjusting for under-and over-registration, the incidence rate of invasive cancer of the cervix in Greater Manchester for 1988-1989 increased from 16.5 to 18.1 per 100,000 population.(ABSTRACT TRUNCATED AT 250 WORDS)