A retrospective descriptive epidemiological survey was conducted on those cases registered as sudden infant death syndrome (SIDS) in Taiwan from 1996 to 2005. Relevant information was obtained from the Institute of Forensic Medicine, Ministry of Justice, as well as from the Statistical Yearbooks published by the Department of Statistics, Ministry of Interior and by the Department of Health, Executive Yuan, including all registered SIDS cases with and without forensic autopsy. Data were analyzed and examined by standard statistical tests. From 1996 to 2005, the incidence rate of SIDS in Taiwan was 0.60‰ in average and remained stable. Most SIDS deaths in Taiwan occurred between 1 and 3 months of age and in cool season, especially from December to March. Only 6.03% of SIDS cases were diagnosed after forensic autopsy. Although there was an inclining trend, the rate of forensic autopsy was still low. Besides, there were differences in the incidence rate and autopsy rate of SIDS between administrative regions in Taiwan, and there seemed to be no correlation between the incidence rates and autopsy rates of different regions. Totally 96 SIDS cases were diagnosed after forensic autopsy, most of them died in parent’s or nanny’s places, and the fatal episode apparently occurred during sleep. However, the information of most of the possible risk factors of SIDS (e.g. sleep position and socioeconomic status, etc.) was not available. Autopsy findings in major cases were consistent with those reported in the literature, including congestion and edema in multiple organs, petechiae in the thymus, pericardium and pleura, and focal alveolar hemorrhage. Although some particular pathologic findings, such as mild ischemic neuronal injury and lymphocytic infiltrate in the myocardium, were noted in a few cases who received immunization within 3 days before death, their clinical significance should be further clarified by more studies. In comparison between SIDS cases with and without forensic autopsy, there were some differences in distributions of gender, age and month of death of the deceased. It could be said that to make a diagnosis of SIDS without forensic autopsy may influence the quality of diagnoses and further compromise the reliability of epidemiological researches. It is worthy of our best efforts to increase the rate of forensic autopsy among SIDS cases, to establish specification of procedures for case investigation and inter-institution cooperation, and to complete the local SIDS epidemiological studies.