The purpose of this study was to test the diagnostic discrimination of bleeding on probing in a group of 71 patients who have been treated for periodontitis. They have been maintained over 5 years receiving prophylaxis every 3 months and clinical evaluation once a year. 1,746 teeth were used and extracted and retreated teeth were not included. The presence or absence of bleeding on probing was recorded for each tooth over the five yearly examinations. After recording the site variations between the 1st and the 5th year maintenance values, the highest value of attachment loss per tooth was selected. Teeth were considered diseased when they presented at least one site with 4 mm of probing pocket depth at the baseline (year 1) and 2 mm of attachment loss after 5 years (Criterion 1), or showed 3 mm of attachment loss after 5 years (Criterion 2). The prevalence of disease was 8.19% (143 teeth) for Criterion 1 and 19.7% (344 teeth) for Criterion 2. Bleeding on probing 2 times presented the highest sensitivity (86% Criterion 1 and 64.5% Criterion 2) and the lowest predictability (11.9% Criterion 1 and 21.4% Criterion 2). Bleeding on probing frequency = 5 times showed the highest specificity (93.6% Criterion 1 and 93.4% Criterion 2) and predictability (93.4% Criterion 1 and 81.5% Criterion 2). These results demonstrated that bleeding on probing may not be substituted for clinical attachment levels when evaluating the periodontal attachment loss over time. However, the high specificity and predictability of these tests as a negative one suggests that absence of bleeding after probing may be a good indicator of maintenance of periodontal attachment levels over time.