Vital registration of causes of death in Tanzania is incomplete and many deaths occur outside health care settings. Verbal autopsies (VA) are used to determine the underlying cause of death, and the probable diagnosis helps to estimate reasonably cause-specific mortality. In this paper, we report findings of a verbal autopsy survey which involved eight villages in both low and highlands of Muheza district, north-eastern Tanzania. The survey was conducted following a rapid census, which was done to identify households that had lost one or more members within a period of two years from the date of census. Trained research assistants administered VA questionnaires to parents/close relatives. Two physicians reviewed each report independently and a third opinion was sought where there was discordant report between the two. A total of 9,872 households were surveyed and 134 deaths were recorded. A total of 96 (71.6%) deaths were from lowland villages representing high malaria transmission. Majority (72.4%) of the reported deaths occurred at home whilst 32.1% occurred at heath facility settings. Overall, severe malaria was the leading cause accounting for 34.3% of all deaths. Infants were most affected and accounted for 43.5% of the total deaths. Pulmonary tuberculosis ranked second (8.2%) cause of deaths and was exclusively confined to individuals > or = 15 years. Probable cause of death could not be determined in 13.4% of deaths. In conclusion, majority of deaths in rural north-eastern Tanzania occur at home and the immediate causes are usually unknown or not documented. These findings indicate that the verbal autopsy is a useful tool for detecting leading causes of death at community level in the absence of health facility-based data.