Temporal variations of blood parasite density were evaluated in a longitudinal study of young, asymptomatic men in a village with endemic malaria in Mali (West Africa). Our main intention was to challenge the value of a single measure of parasite density for the diagnosis of malaria, and to define the level of endemicity in any given area. Parasitaemia and body temperature were recorded three times a day in the wet season (in 39 subjects on 12 days) and in the dry season (in 41 subjects on 13 days). Two thousand nine hundred and fifty seven blood smears (98.5% of the expected number) were examined for malaria parasites. We often found 100-fold or greater variations in parasite density within a 6-hour period during individual follow-up. All infected subjects had frequent negative smears. Although fever was most likely to occur in subjects with a maximum parasite density exceeding 10000 parasites/mm3 (P = 0.009), there was no clear relationship between the timing of these two events. Examples of individual profiles for parasite density and fever are presented. These variations (probably due to a 'sequestration-release' mechanism, which remains to be elucidated) lead us to expect a substantial impact on measurements of endemicity when only a single sample is taken. In this study, the percentage of infected individuals varied between 28.9% and 57.9% during the dry season and between 27.5% and 70.7% during the wet season. The highest rates were observed at midday, and there were significant differences between days. Thus, high parasite density sometimes associated with fever can no longer be considered as the gold standard in the diagnosis of malaria. Other approaches, such as decision-making processes involving clinical, biological and ecological variables must be developed, especially in highly endemic areas where Plasmodium infection is the rule rather than the exception and the possible causes of fever are numerous.