Background: Our objectives were to quantify: 1) the association between incident tooth loss and prior periodontal attachment level; and 2) the contribution to tooth loss made by non-periodontal conditions in increasingly periodontally involved teeth. Methods: The Florida Dental Care Study was a prospective cohort study of persons who at baseline had at least 1 tooth and were 45 years or older. In-person interviews and clinical examinations were conducted at baseline, and at 24 and 48 months, with telephone interviews at 6-month intervals in between. A regression model was used to simultaneously quantify tooth-specific predictors of tooth loss, with person-level factors taken into account. Results: Of the 687 persons who participated for a 48-month clinical examination, 36% lost 1 or more teeth during follow-up, and 5.0% of all teeth were lost. Attachment level up to 2 years before tooth loss was strongly predictive of incident tooth loss, with increases in risk for each millimeter in attachment loss. Certain other tooth-specific conditions (tooth mobility, bulk restoration fracture, decayed surfaces, filled surfaces, tooth type and arch location, root fragment) were strongly and independently associated with increased risk for tooth loss, while others were not (prosthetic crown coverage, cusp fracture, root surface defect). Propensity to choose extraction over other treatment alternatives, as reported by participants at baseline, was also strongly predictive of tooth loss. Conclusions: Increasingly severe attachment level was consistently associated with an increased risk for tooth loss in this sociodemographically diverse sample, with or without other tooth-specific conditions taken into account.