A stepwise approach to determine attachment level changes was utilized to assess the nature of progression of periodontal disease. Following initial screening, 51 subjects with established periodontitis were monitored quarterly for 9 more months. Probing depth (PD) and relative attachment level (RAL) were recorded using an automated, pressure sensitive probe system. To establish intra-examiner error, repeated measurements were performed for all sites at the final visit. An overall standard deviation (SD) for RAL repeated measurements was initially calculated (0.76 mm) using all 6,935 double measurements. Sites were sorted by factors which contribute to the error of attachment level measurements; i.e., pocket depth (shallow, moderate, deep), tooth type (molar, non-molar) and location (buccal, lingual). Data were sorted by the above 12 groups, and SD for repeated measurements was calculated separately for them. The ratio between these SD and the overall SD served as the corrective factor. Each patient's initial threshold (2 SD) was multiplied by these corrective factors thus resulting in 12 thresholds for each subject. Next, linear, exponential and logarithmic regression models were tested for each site, and the regression model showing the highest R value was chosen for that site. AL changes were tested against the patient's threshold for that site. Sites with attachment loss exceeding the threshold were deemed active. Five hundred eighty-one sites (8.3%) exhibited attachment loss exceeding the various thresholds. Of these, linear progression occurred in 195, logarithmic in 224, and exponential in 162 sites. Individual patient's attachment loss ranged from 0.6 to 19.4% of all sites.(ABSTRACT TRUNCATED AT 250 WORDS)