Most of our knowledge regarding the natural history of periodontitis in adults has come from clinical studies of dental patients. However, it is known that dental patients usually are not representative of what actually happens in the population. For this type of information, epidemiologic studies are required. The purpose of this paper is to present the trends in attachment loss over a 3-year period in a population of community-dwelling elderly blacks and whites. Specifically, this paper presents attachment loss trends both at the person and site level to address the following issues: 1) whether attachment loss during one period in time makes a person at higher risk for attachment loss at a subsequent period; 2) whether sites with deeper periodontal pockets at baseline are more likely to experience future attachment loss; and 3) whether teeth that experience attachment loss during one time period are more likely to be lost at the next time period. In 1988, the University of North Carolina School of Dentistry initiated the Piedmont 65+ Dental Study, which is a subsample of the parent study, the Piedmont 65+ Study of the Elderly under the direction of investigators from Duke University. The dental study subsample was designed to elicit 800 dentate respondents in the 5-county area who were examined again at 18 and 36 months. Our findings indicated that, for the two 18-month periods investigated, attachment loss during the first period was related positively to the incidence of attachment loss at a subsequent period at the person level, but not at the site level.(ABSTRACT TRUNCATED AT 250 WORDS)