During the period between 1984 and 1999, 309 cases of sudden unexpected death in infancy and early childhood (0-3 years) were investigated at the Institute of Forensic Medicine in Oslo. In 73 cases, an explainable cause of death was found. In this non-sudden infant death syndrome (SIDS) group, 42 cases were due to disease, 14 to accidents, 7 to neglect/abuse and 10 cases were due to homicide. In 43 cases, there were pathological findings at the autopsy or suspect features in the history and/or circumstances, which were, however, insufficient to explain death ("borderline" SIDS). In the remaining 193 cases, nothing of significance was detected ("pure" SIDS). The purpose of the present study was to evaluate the importance of the different diagnostic tools used in diagnosing non-SIDS and borderline SIDS cases. The definition of SIDS requires a negative history as well as a negative autopsy result. Thus, the following variables were analysed: circumstances, medical history and autopsy, which included a gross pathological investigation, histology, neuropathology, microbiology, radiology and toxicology. In diagnosing deaths due to disease, histology, neuropathology and microbiology were the most important diagnostic tools. In contrast, information about the circumstances of death and the gross pathological findings at autopsy most often revealed the cause of death in accidents and cases of neglect/abuse and homicide. Following the drop in SIDS rate in Norway after 1989, the share of pure SIDS in proportion to the total population of sudden unexpected deaths in infancy and early childhood has decreased. The increasing proportion of non-SIDS and borderline SIDS cases presents a challenge to improve the quality of the investigation in cases of sudden death in infancy and early childhood.