AIM--To investigate the quality of perinatal and infant necropsies and assess the relation between the quality and value of this investigation in different outcome groups. METHODS--Cohort analysis of 540 deaths during 1993 of babies between 20 weeks' gestation and one year of age born to women usually resident in Wales. Cases were identified from the All Wales Perinatal Survey. Each case was assessed to establish whether the necropsy yielded clinically relevant information. The quality of necropsy was assessed by scoring aspects identified as being part of the necropsy. RESULTS--Necropsy was performed in 335 (62%) cases, and the report was available for assessment in 314 cases. The quality of necropsy was below the minimum standard in 46% (143/314) of cases. The highest quality necropsies were carried out on fetal deaths at 20 to 23 weeks' gestation (12% (10/85) below standard), compared with 65% (87/133) below standard on stillbirths and 68% (21/31) on sudden unexpected infant death. Overall, 42% (131/314) of necropsies were performed in a regional paediatric pathology centre including 88% (76/88) of fetal deaths, 23% (31/133) of stillbirths and 30% (29/96) of infant deaths. The quality score for the necropsy performed in a regional centre failed to achieve the minimum acceptable score in 8% (11/131) of cases compared with 72% (132/182) for those done elsewhere. The cause of death was detected by necropsy in 17% (52/314) of cases. The quality of necropsy was significantly higher when the cause of death was revealed than when nothing new was found. CONCLUSIONS--The overall quality of the perinatal and infant necropsy is poor. This is regrettable as valuable information can be revealed frequently by a good quality necropsy. Adherence to Guidelines for Postmortem Reports recently published by the Royal College of Pathologists should improve the situation.