Objectives: (1) Describe for a diverse sample the 24-month incidence of root caries, and (2) test its association with a broad range of clinical, behavioral, financial, and sociodemographic factors. Methods: The Florida Dental Care Study was a cohort study of randomly selected subjects who had at least 1 tooth and were 45 years or older at baseline. In-person interviews and clinical examinations were conducted at baseline and 24 months, with 6-monthly telephone interviews between those times; 723 subjects participated for both examinations. A multinomial logistic regression was done to predict whether the subject was in one of four mutually exclusive groups [new root decay only (NDO); new root filling(s) only (NFO); both new decay and new filling(s) (BOTH), or had neither (NONE)]. Results: Thirty-six percent of subjects had at least 1 new root decayed and/or filled surface (DFS); 17% were in the NDO group, 14% in the NFO group, and 5% in the BOTH group. When limited to participants who had a nonzero increment, the mean (SD) DFS was 2.7 (2.9). Baseline clinical condition (presence of root decay, root filling(s), coronal decay, noncarious root defects, number of teeth present, percent of teeth with at least 4 mm of attachment loss) was predictive of moving from the NONE group into the NDO, NFO, or BOTH groups. The addition of behavioral, financial, and sociodemographic factors improved model fit. For example, regular dental attenders were significantly more likely to move from the NONE group into the NFO group, but regular attendance was not associated with a lower probability of moving from the NONE group into the NDO or BOTH groups. Conclusions: Root caries is a substantive dental health problem in this diverse sample of adults. These analyses demonstrate the utility of disaggregating caries incidence into four mutually exclusive groups for predictive models. Copyright © 2001 S. Karger AG, Basel.