Correlation of the Gram stains prior to and after rupture of the membranes (ROM) and the efficacy of Gram-stain diagnosis for bacterial vaginosis (BV) after membrane rupture has not previously been evaluated. From April 1997 to May 1998, women presenting in labor or for labor induction were invited to participate. Women with ROM prior to hospital presentation or contraindications for vaginal delivery were excluded. A Gram stain of vaginal secretions was obtained prior to and approximately 2 hours after ROM. BV was diagnosed with a Nugent score of 7-10. The Gram stain prior to and after membrane rupture was evaluated using the Spearman correlation coefficient. Population characteristics (n = 91) included maternal age of 15 to 38 years, 45 (49%) nulliparas, 59 (65%) African-Americans, 23 (25%) Hispanics, 7 (8%) Caucasian, and 2 (2%) Asians. Gestational age upon study entry was 35 to 44 weeks. Twenty-one percent (19 of 91) of the study population was diagnosed with BV. Using the Gram stain prior to membrane rupture as the standard, the efficacy of the Gram stain after ROM had a sensitivity of 26%, specificity of 97%, positive predictive value of 71%, and negative predicative value of 83%. Good correlation of pre- and post-membrane rupture Gram stains was demonstrated (r = 0.69). Gram stain may be useful for ruling out BV in the presence of membrane rupture, but is not useful for diagnosing BV due to poor sensitivity.