In 1984, a four-year mixed-longitudinal study of oral health was begun in Morogoro Town and Morogoro District, Tanzania. Its main goal was to detect a possible upward trend in prevalence of caries in child cohorts in this developing country. The sample consisted of 722 seven- to nine-year-olds in 1984, 833 seven- to 11-year-olds in 1986, and 989 seven- to 13-year-olds in 1988. Mixed-longitudinal studies provide the opportunity for application of the Age-Period-Cohort analysis. On the basis of external information, either age and/or cohort and/or period effects can thus be isolated. Apart from significant age and period effects, a cohort effect was found, indicating that the prevalence of caries has decreased in these child populations between 1984 and 1988.