To evaluate the contribution that perinatal and infant necropsy makes to clinical practice and to see how this might be influenced by the quality of the investigation. Cohort analysis, with data from the all Wales perinatal survey, of perinatal and infant deaths during 1993 of babies born to mothers usually resident in Wales. The clinicopathological classification of death based on clinical details was compared with the classification after necropsy. Similarly, cases in which necropsy yielded new information were identified. The quality of the necropsy was assessed by scoring six aspects of the examination. 400 consecutive deaths at 20 weeks of gestation to 1 year of age. Necropsy rate, effect of necropsy on clinicopathological classification, new information disclosed by necropsy, quality of necropsies, and the link between new information and quality of the necropsy. Necropsy was performed in 232 cases (58%). The clinicopathological classification was altered by necropsy in 29 cases (13%). New information was obtained in 60 cases (26%), and in 42 (18%) it disclosed the cause of death. The quality of necropsy was substantially higher when the main cause of death was detected than when nothing new was found. Necropsy is underused. Clinicians should be more positive about necropsies and realise how much clinically relevant information can be obtained from a good quality examination.