There is evidence that faecal egg counts of Schistosoma mansoni vary considerably from day to day, which results in poor sensitivity of single stool readings. Intra-specimen variation of S. mansoni egg counts may also be considerable, but has previously been considered as the less important component. We quantified the relative contribution of these two sources of variation among 96 schoolchildren from an area in Cĵte d'Ivoire highly endemic for S. mansoni. Stool specimens were collected over 5 consecutive days, and 5 egg-counts were made in each specimen by the Kato-Katz technique. The point prevalence of the first sample was 42.7% and the cumulative prevalence after the maximum sampling effort was 88.5%. Using generalized linear mixed models we found that the presence of S. mansoni eggs in a stool sample varied much more between days than within specimens, indicating that stool sample examination over multiple days is required for accurate prevalence estimates. However, using the same approach, we found that among infected children intra-specimen variation in egg counts was 4.3 times higher than day-to-day variation. After praziquantel administration, day-to-day variation was more important than before, since most infections were very light and thus likely to be missed altogether by stool examination on a single day. We conclude that diagnostic sensitivity in high transmission areas is maximized by making several stool readings on several days, but examining 1 stool specimen several times can make reasonable estimates of infection intensity.