OBJECTIVES: To describe the trend and causes of neonatal deaths in a rural district in northern Ghana. METHODS: Descriptive analysis of data collected from the Navrongo Demographic Surveillance System and verbal autopsies conducted on all neonatal deaths from 1995-2002. RESULTS: Of 1118 recorded neonatal deaths 1068 (95.5%) could be analysed. Only 13.2% of deaths occurred at the health facility; 62.7% occurred in the early neonatal period, with prematurity (38%) and birth injuries (19%) as leading causes. Infectious causes (66%) were the major contributors to late neonatal deaths. Infanticide accounted for 4.9% of all neonatal deaths. The cause-specific mortality rate for neonatal tetanus remained under 2.5% throughout the 8-year period. Overall, the neonatal mortality rate declined at an average of 2.5 per 1000 live births per year: Down by nearly 50% from 40.9 (95%C.I. 34.1-46.8) in 1995 to 20.5 (95%C.I.17.3-22.7) in 2002. CONCLUSION: The various health interventions undertaken in this district have had the collateral effect of causing decline in neonatal mortality. Neonatal mortality could be further reduced by preventing and treating neonatal infections, having skilled attendance at delivery and the elimination of infanticide. Data from demographic surveillance sites may be useful in monitoring trends in child mortality.