Background. Epidemiologic studies of periodontal infection as a risk factor for cardiovascular disease often use clinical periodontal measures as a surrogate for the underlying bacterial exposure of interest. There are currently no methodological studies evaluating which clinical periodontal measures best reflect the levels of subgingival bacterial colonization in population-based settings. We investigated the characteristics of clinical periodontal definitions that were most representative of exposure to bacterial species that are believed to be either markers, or themselves etiologic, of periodontal disease. Methods. 706 men and women aged ≥ 55 years, residing in northern Manhattan were enrolled. Using DNA-DNA checkerboard hybridization in subgingival biofilms, standardized values for Aggregatibacter actinomycetemcomitans, Porphyromonas gingivalis, Treponema denticola and Tannerella forsythia were averaged within mouth and summed to define "bacterial burden". Correlations of bacterial burden with clinical periodontal constructs defined by the severity and extent of attachment loss (AL), pocket depth (PD) and bleeding on probing (BOP) were assessed. Results. Clinical periodontal constructs demonstrating the highest correlations with bacterial burden were: i) percent of sites with BOP (r = 0.62); ii) percent of sites with PD ≥ 3 mm (r = 0.61); and iii) number of sites with BOP (r = 0.59). Increasing PD or AL severity thresholds consistently attenuated correlations, i.e., the correlation of bacterial burden with the percent of sites with PD ≥ 8 mm was only r = 0.16. Conclusions. Clinical exposure definitions of periodontal disease should incorporate relatively shallow pockets to best reflect whole mouth exposure to bacterial burden.