Background: The objective of this investigation is to use non-invasive, state-of-the-art, diagnostic techniques to measure periodontal disease progression and model periodontal disease activity over time. In this investigation, digital subtraction radiography and an electronic controlled force periodontal probe capable of attachment level measurement were used to measure bone loss and attachment loss, respectively. The use of these nearly continuous measures of attachment and bone loss allowed detection of small amounts of disease activity and provided data to be used in modeling of the disease process over time. Methods: Forty-four patients were studied for 18 months. Examinations used clinical attachment level measures at 1-month intervals and quantitative radiology at 6-month intervals. The sites were analyzed by regression for statistically significant changes. These data were used to determine sites of periodontal disease activity for testing various models of periodontal disease progression. Results: Overall 22.8% of sites lost attachment, 5.4% gained, and 71.7% demonstrated no statistically significant change. The mean time to lose 1 mm of attachment was 8.4 ± 0.6 months. In the first model tested a step-wise discriminant analysis was used to determine whether or not baseline measurements of plaque (PI), gingival inflammation (GI), attachment loss, and probing depth (PD) could be used to derive a satisfactory model for disease progression. Although the overall model was statistically significant with PI, PD, and GI contributing to the model (Wilks' lambda = 0.859, F = 5.71, P <0.0012), its predictive power was relatively weak. A considerably stronger significant model resulted when the rate of attachment loss over the first 6 months, baseline PI, and baseline GI were included (Wilks' lambda = 0.712, F = 14.17, P <0.00001). A significant model also resulted when bone loss during the first 6 months and baseline probing depth were included (Wilks' lambda = 0.438, F = 61.48, P <0.00001). When the last model was applied to each site, the sensitivity in predicting disease progression was 80.0% and the specificity in ruling out progressive disease was 93.9%. Conclusions: This study indicates that clinically significant progression of attachment loss in posterior tooth sites occurs as a frequent event in adult periodontitis. The modeling data also suggest that short-term (6 month) measures of periodontal disease progression greatly improve the ability to model attachment loss over a longer period in untreated periodontitis patients. J Periodontol 2000;71:1583-1590.