Weekly personal NO2 exposures of 246 children aged 3-6 years were measured with Palmes tubes during 13 weeks in winter and spring in 1991. Measurements were made in eight day-care centers in the downtown and suburban areas of Helsinki, Finland. At the same time, inside and outside NO2 concentration of the day-care centers and the ambient air fixed site measurements were conducted. Palmes tubes were found to be applicable for NO2 exposure measurements of preschool children, but rather high sample losses could be expected. The geometric mean of personal NO2 exposure levels of 13-week period was 26.5 micrograms/m3 in the downtown and 17.5 micrograms/m3 in the suburban area. Gas stove and smoking at home increased significantly personal exposure to NO2. The weekly population NO2 exposure correlated rather poorly with the fixed site ambient air NO2 levels (R2 = 0.37), but much better with the NO2 levels inside and outside the day-care centers (R2 = 0.88 and 0.86). In the suburban and downtown groups the between children variations in the NO2 exposures were only 14% and 28% of the total variations, which were dominated by the within child variation. Stationary measurements at the ambient air fixed sites and inside and outside the day-care centers explained the variation in personal exposures of the children well during the spring, but not during the winter. A regression model, where data from outside day-care center measurements, fixed ambient air monitors, residential area and home characteristics (i.e., gas stove, smoking inside at home, type of dwelling) were included, explained 32% of the personal NO2 exposure variation in winter and 67% in spring. In the absence of personal exposure measurements, both stationary measurements and questionnaire information are useful in estimating variations in personal exposures.