To assess the interrater reproducibility of malaria microscopy in epidemiological studies, 711 thick blood films from population-based surveys were randomly selected and reread by 4 experienced microscopists. Sample estimates of the prevalence of P. falciparum infection, geometric mean parasite density and the proportion of samples above various parasite density cut-off levels were almost identical in the routine and quality control readings. Differences were, however, encountered in the sample estimates for gametocyte ratio, proportion of mixed infection and average density index. In all three cases the quality control result was significantly higher than the routine evaluation. On the level of the individual slide there was good interrater agreement for the presence of P. falciparum infections (Kappa index kappa = 0.79) which was even better when parasite densities between 4 and 100/microl were excluded (kappa = 0.94). With respect to the assessment of parasite density, a high level of disagreement was found. While the mean difference between the two readings was not different from 0, the second reading was between 0.12 and 10 times that of the first. However, the level of disagreement significantly fell with increasing parasite densities. Thus malaria microscopy is very reliable for the estimation of parasite ratios and geometric mean parasite densities within and between studies as long as the same methodology is used, but tends to underestimate the gametocyte ratio and proportion of mixed infections. Care must be taken, however, when individual parasite density is related to other explanatory variables, due to the high degree of variability in the parasite enumeration.