BACKGROUND: Epidemiological surveillance and research on birth defects require accurate diagnosis and adequate registration. In this regard, the performance of national birth registries is not well described. METHODS: We linked clinical data from all 3,616 cleft cases treated in Norway between 1967 and 1998 with data from the Medical Birth Registry of Norway, and calculated the proportion of clinically verified cases reported to the Registry, stratified by severity. RESULTS: The cleft type most completely ascertained was cleft lip and palate (CLP), of which 94% were reported. Ascertainment was less complete for cleft lip alone (83% recorded), and cleft palate only (CPO) (57% recorded). For each of the three types of clefts, completeness of reporting depended on severity of the cleft. For example, 71% of cases with severe CPO were reported, while only 11% of cases with mild CPO were reported. CONCLUSIONS: Ascertainment was strongly related to cleft type and severity. To the degree that severity of birth defects may be related to their cause, these patterns of registration have implications for surveillance of birth defects as well as the conduct of etiologic studies. The large proportion of cleft palate cases unrecorded at birth suggests that clinical examination of the newborn palate is often inadequate.