OBJECTIVE: To determine differential improvements in mortality rates according to age, and to discuss differences in neonatal and post-neonatal mortality rates, using longitudinal data from the past 50 years in three rural Gambian villages. METHODS: All villagers whose date of birth was recorded were followed up until death or October 1997. A Lexis expansion was performed to categorize age: early neonatal, late neonatal, infant, 1-4 years, 5-14 years and 15 or more years. Calendar time was divided into three periods -- prior to 1975, 1975-1984 and after 1985, representing different levels of clinical care. Cox regression and likelihood-ratio test were used to model the hazards ratios. RESULTS: There were 3981 subjects included in the analysis with a total of 59 002 person-years follow-up. There was a dramatic decrease in neonatal (44 to 15), infant (162 to 36) and under-five mortality rates (397-66 per 1000) from pre-1975 to the present. The disproportionate decrease in the mortality rates means that neonatal deaths accounted for a greater percentage of post-natal deaths in later years. There was a marked seasonality in mortality rates with significantly more deaths in the 'hungry' season prior to 1975 (odds = 1.87; 95% CI = 1.62-2.17) and from 1975 to 1984 (odds = 1.84; 95% CI = 1.34-2.53). This seasonality of death has diminished in recent years (odds = 1.23; 95% CI = 0.85-1.76). CONCLUSION: Our analysis indicates a dramatic reduction in mortality rates and an attenuation in the seasonality of death. It is likely that the efficient implementation of basic currently available health measures was one of the key elements in achieving such a major reduction in mortality.