The progression of proximal alveolar bone loss (ABD index) in an unselected material comprising 406 individuals was analysed by a longitudinal investigation over a period of ten years. In order to minimize the number of measurements and the drop-out of measurements of the difference in alveolar bone height between the two examinations a partial recording system was constructed. A partial index including five sites (12 m, 11 m, 33 d, 31 d, 41 m) gave high correlation to a total recording (r = 0.96). The best alternative in periapical radiographs to measuring the alveolar bone height in relation to the root length was to measure it in relation to the tooth length. This second method does not necessitate identification of the cemento-enamel junction. The mean alveolar bone difference was 5.5 per cent of the mean root length, which corresponds to an average bone loss of 0.09 mm per year. Eighteen variables were analysed as predictors in stepwise multiple regression analyses. The dependent variable was the "ABD index". Four predictors reached significance at 1 per cent level in the multivariate analysis, the alveolar bone loss 1970 ("ABL index 1970"), "Age", "Number of lost teeth" and "Russell's Periodontal Index" ("PI"). The coefficient of determination (R2) was 0.40. In a selected part of the material, consisting of individuals with at least 20 remaining teeth, the stepwise multiple regression analysis was repeated with the same 18 predictors. On this occasion two predictors reached significance at 1 per cent level, "PI" and "Smoking", and these two factors showed an interaction.