Seven hundred eight patients aged 5-19 years in a pediatric practice in North Carolina were selected using a random-start, systematic sampling procedure and enrolled in a case control study to determine risks for fluorosis. Subjects were examined by four trained examiners using the Tooth Surface Index of Fluorosis (TSIF). Information on fluoride exposures and other explanatory variables was obtained through parent interviews and mail questionnaires. Fluoride exposure was confirmed by fluoride assay of community drinking water samples. Bivariate associations for the entire sample were tested using MH chi2 statistic. A secondary analysis controlling for fluoride in drinking water was performed using logistic regression for 233 subjects (116 drinking fluoridated water; 117 drinking fluoride-deficient water) who were lifetime residents at the same address. Nearly 78% of subjects had a TSIF score of > 0; 36.3% > 1; and 18.9% > 2. Twenty-two variables found in bivariate analyses (P < 0.15) to be associated with fluorosis were included in multivariate analyses. For subjects drinking fluoride-deficient water, fluorosis (1 or more positive TSIF scores) was associated with dietary fluoride supplement frequency (OR = 6.5) and age of the child when brushing was initiated (OR = 3.0). For subjects drinking fluoridated water, fluorosis was associated with age of child when brushing was initiated (OR = 3.1).