The purpose of this analysis was to evaluate the feasibility of detecting P. gingivalis using selected sites and to indicate increased proportions of this organism in periodontitis patients. In 10 patients suffering from moderate to advanced periodontal disease, separate microbiological samples were taken from the mesial, buccal, distal and oral (lingual or palatal) aspects of every tooth. This yielded a total of 927 microbiological samples, 84 to 102 per patient. Three distinct patterns of distribution and relative proportion of P. gingivalis were recognized. In one group of patients, the organism was not cultured. In a second group, few positive sites with low proportions of P. gingivalis were present. A third group of patients yielded high frequencies and proportions of P. gingivalis. The number of samples necessary to diagnose the presence of P. gingivalis at a 95% confidence level varied considerably between the three groups. In 4 patients, sampling 4 randomly selected sites was sufficient, while in the remaining 3 positive patients, 25 or more samples were required to detect the organism with equal certainty. Seven different protocols for multiple subgingival sampling were studied. When considering the number of samples needed to detect the presence of P. gingivalis and to estimate the highest proportion of this organism, selection of the deepest pocket in each quadrant was the most efficient method of sampling.