This study was designed to assess reliability of interpretation of Gram-stained vaginal smears by using Nugent's scoring system for diagnosis of bacterial vaginosis (BV) across three different centers in Turkey: two in Ankara and one in Trabzon. The vaginal smears were collected from clients attending a family planning clinic in Trabzon, Turkey during October-December 1997. One slide taken from each client was prepared according to the standard methods and enumerated. One evaluator from each center examined the slides independently for the presence of BV and none of them had access to the evaluation of the others. Out of 372 slides, 301 (81%) were found to be satisfactory for scoring by all three evaluators and included in the analysis. Nugent's scores from 1-10 reported from each evaluator were compared by Spearman correlation coefficients and Kappa statistics. The difference in the proportions of BV diagnosis in three centers was evaluated by chi2 test. There was good agreement for the interpretation of Gram-stained vaginal smears by Nugent's scoring system for diagnosis of BV. These results indicate that it is a reliable method in diagnosis of BV at different settings.