Consultation between the clinicians and epidemiologists responsible for the Perinatal Mortality Surveys in Scotland and in the Northern Regional Health Authority in England showed that the classification of perinatal death introduced more than 30 years ago by Sir Dugald Baird still retained its utility, but that unintentional differences in the way cases were being classified had threatened the validity of temporal or geographical comparisons. To overcome this problem an effort has now been made to define the main terms used in this classification more precisely. To preserve continuity, the main structure of the original groupings has been retained; but the opportunity has been taken to adjust certain minor groups in conformity with recent ideas, and also to modify definitions to take into account the greatly improved prognosis for babies of very low birthweight. Otherwise, it is thought that subclassification of the main groups offers a better method of exploring new hypotheses than any radical alteration of the main groups themselves.