The aim of this study was to assess agreement and association of gingival bleeding after repeated probing at different time intervals in subjects with gingivitis. Twenty adults participated. Periodontal probing (P approximately 1.27 MPa) was conducted at six sites of every tooth present. Probing was repeated in different quadrants immediately after the first probing (T0), after 1 h (T1), 4 h (T4), and 24 h (T24). A total of 3,459 sites were probed twice. The mean proportion of sites bleeding on probing (BOP) was 0.23 (standard deviation 0.08, range 0.10-0.41). Probing itself had a significant effect on the results of repeated probing. For T0 through T24, respective mean differences of proportions were 0.04, 0.04, 0.01, and -0.03. Ninety-five percent repeatability coefficients of proportions were 0.17-0.18. Estimates of overall kappa were 0.390, 0.234, 0.233, and 0.046 for T0 through T4, respectively. Adjusted two-level binary response models revealed odds ratios (95% confidence interval) for BOP at T0 through T4 of 6.52 (4.34, 9.80), 3.23 (2.19, 4.76), 3.80 (2.63, 5.50), and 2.68 (1.85, 3.89). It was concluded that a certain degree of agreement of site-specific bleeding scores in subjects with plaque-induced gingivitis could be observed only if probing was repeated at once. Adjusted associations between repeat BOP were weak in general, but strongest immediately after first probing. There appears to be a significant effect of probing itself, which may last for more than 1 h, whereas 24-h results are obtained under different conditions.