We examined subjective and objective correlates among 67 women with symptomatic bacterial vaginosis before and after treatment with intravaginal clindamycin or placebo. We found no preponderance of any sexual practices among these patients. Nine patients (13.4%) had had hysterectomy. Whereas odor and discharge were the most common symptoms, 30 patients (44.8%) also complained of vulvovaginal irritation. Symptoms correlated poorly with objective therapeutic outcome. On examination the diagnosis would have been missed in seven patients (10.4%) if the clinician relied on presence of an abnormal vaginal discharge to suggest bacterial vaginosis. Vaginal pH greater than 4.5 was found immediately after curative therapy in 59.6% of patients. Mobiluncus spp. morphotypes were 99.0% specific and 52.1% sensitive and proline aminopeptidase activity in vaginal fluid was 84.4% sensitive and 70.2% specific for diagnosis. Our Gram stain criteria yielded no false-negative results, 6.1% false-positive, and frequent indeterminate results after therapy. We found little evidence for sexual transmission of bacterial vaginosis. Recurrence after effective therapy was not predicted by vaginal pH elevation, positive or indeterminate Gram stain result, or positive proline aminopeptidase test.