Between July 2000 and June 2001, we used weekly active case detection (ACD) of clinical malaria episodes in 618 children aged < 5 years to describe the epidemiology of malaria in Ifakara, southern Tanzania. Plasmodium falciparum-positive blood slides prepared from children with axillary temperature 37.5 degrees C were used to define clinical malaria and a rolling cross-sectional survey documented the prevalences of parasitaemia and anaemia. A random subsample of children was visited daily for 1 month at the end of the study to assess the effect of more frequent visits on estimated incidence rates. Only 50 (8%) children had 1 or more episodes of clinical malaria during the year, an overall incidence of 0.275 episodes/100 child-weeks-at-risk, with no age dependence. The maximum parasite prevalence of 25% was reached in children aged 4 years. The incidence of illness was significantly lower in children visited daily than in those visited weekly, suggesting a marked effect of frequent visits on estimated incidence rates. We conclude that the age pattern of malaria detected through ACD is a more robust epidemiological indicator than absolute incidence rate estimates and that, in contrast to the surrounding area, Ifakara town is subject to only moderate perennial malaria transmission.