The present communication attempts to summarize some of the features of attachment loss which are of interest to the clinician and the statistician analyzing data from clinical trials. These include the measurements employed to detect changes in attachment level, the nature of the destructive disease process and the effects of therapy on the attachment level measurements. Although there are several difficulties associated with the attachment level measurement, at the present time it appears to be the best estimator of periodontal attachment available. The overall standard deviation of this measurement in greater than 46,000 replicate measurements at periodontal sites in 58 subjects was 0.78 mm (range 0.4 to 1.2 mm). In the periodontally healthy subject, the length of the attachment measured around 28 teeth is approximately 700 mm. Therefore, there are approximately 1400 adjacent points along the periodontal attachment where a measurement could be made using a periodontal probe with a 0.5 mm diameter tip. If 6 measurements were recorded per tooth, then approximately 12% of the possible probable points would be evaluated. Recent data indicate that destructive periodontal diseases progress with acute bursts of activity rather than as slowly progressive, continual processes. Such findings suggest new models of attachment loss progression. In one likely model, destructive periodontal diseases would progress by asynchronous bursts of activity at individual sites which occur with greater frequency during a finite period of time in an individual's life. 3 major patterns of attachment loss could be distinguished when frequency distributions of attachment level measurements were constructed for 61 destructive periodontal disease subjects. Pattern I (30 subjects) exhibited a bimodal distribution with localized destruction occurring at less than 34% of sites. Pattern II (14 subjects) exhibited more widespread disease (greater than 33% of sites affected) with a trimodal frequency distribution. Pattern III (17 subjects) exhibited a unimodal distribution in which virtually all sites were affected. The proportions of Fusobacterium nucleatum, Streptococcus intermedius and Eikenella corrodens in subgingival plaque samples were significantly elevated in sites of subjects with patterns II and III (the widespread disease groups). Bacteroides intermedius, Streptococcus uberis and Actinobacillus actinomycetemcomitans were elevated in sampled sites of localized disease subjects (pattern I). The effects of therapy by Widman flap surgery and systemic tetracycline were examined by several statistical analyses.(ABSTRACT TRUNCATED AT 400 WORDS)