To validate the Swedish Cancer Registry concerning breast tumours in female residents of Malmö 1971-1991, registry-entries were compared with clinical records. Correctness and completeness of entries was determined for invasive breast cancer (BC) and cancer in situ of the breast (CIS). As BC were registered 3,646 tumours. Of these 3,403 (93.3%) were confirmed as BC. Completeness was good with 1.1% of diagnosed tumours missing. Correctness of registered cases of CIS was 93.3%, but completeness only 63.3%. Most missing cases were registered as BC. In bilateral cancer correctness was 78.0% and completeness 77.2%. All aspects of registration improved during the period. Most differences between diagnosed and registered cases were due to coding CIS of intraductal type as BC. This practice changed in 1980 resulting in improved precision. Using registry data for epidemiological research the question of their validity must be addressed. Special consideration should be paid to coding practice at the registry.