This study reports mortality from neonatal tetanus in Brazil (1979-1987), based on an analysis of death certificates. The distribution of causes of death in time and space was used as an indicator of proportional neonatal mortality from tetanus. Underreporting remains a serious problem, particularly in northern and northeastern Brazil. In most municipalities, only one death was notified during the period under consideration, thus highlighting the importance of considering the theoretical approach of "sentinel events" in the control of this disease. A proportional decrease in mortality from neonatal tetanus was observed in the states of Rondônia, Pará, Sergipe, São Paulo, Paraná, and Rio Grande do Sul. It was not possible to detect an association between immunization coverage with two doses of tetanus toxoid and neonatal tetanus mortality. The reduction in the number of deliveries taking place in the home during the period studied appears to have played an important role in the reduction of neonatal deaths attributable to this cause.