The levels of salivary mutans streptococci and caries experience were used as predictors for caries incidence in 3 groups of pre-school children from areas with different levels of natural and topical fluoride exposure. Altogether 1022 children, 4-5 years of age at baseline, were examined according to the WHO-criteria and followed for 2 years. The low fluoride group (n = 374) had a low fluoride level in the piped water and no topical fluoride applications; the F-varnish group (n = 442) had low water fluoride but semiannual topical applications of a fluoride varnish; the optimal fluoride group (n = 206) had an optimal level of fluoride in the drinking water and semiannual F-varnish applications. The number of salivary mutans streptococci was estimated and scored at baseline with the Strip mutans chair-side method. The sampling procedure was repeated in 337 children of the low fluoride group 3 weeks after baseline. In comparison with the low fluoride group, caries incidence was 30% and 60% lower in the F-varnish and the optimal fluoride group respectively. The caries predictive ability decreased with increasing fluoride exposure. The sum of sensitivity and specificity decreased from 151% (65% + 86%) in the low fluoride group to 131% (40% + 91%) in the optimal fluoride group. The positive predictive value was highest (62%) in the low fluoride group. Repeated salivary samplings at baseline did not improve the caries predictive power. The results suggest that the overall fluoride exposure should be taken into account when caries risk assessment strategies for preschool children are developed and implemented.