Nitroimidazoles have been shown to be the drugs of choice to treat women with bacterial vaginosis, but the recurrence rate is high. Some workers have suggested that the recurrence of symptoms may in fact be reinfection by male consorts, but no controlled studies have been undertaken to confirm this. In an international, multicentre, randomised, double blind trial, the recurrence rate was studied in 241 women with a clinical diagnosis of bacterial vaginosis. All women were treated with 2 g metronidazole twice at an interval of two days. The patients were randomly allocated to two groups, one contained 123 women, whose consorts were given the same dose of metronidazole, the other consisted of 118 women whose consorts were given inert but identical placebo tablets. The women were evaluated at weeks 1, 4, and 12 after treatment. At week 1, the cure rate was 97% (115/119) in women whose consorts had been treated and 98% (111/113) in the others. At week 4 bacterial vaginosis had recurred in 17% (19/112) of women whose consorts had been treated and in 13% (14/106) of those whose consorts had received placebo. At week 12 the recurrence rates were 21% (20/95) in women with treated consorts, and 16% (15/95) in the others. The differences in recurrence rates between the two groups of women were not significant. In conclusion, treating the sexual partners of women with bacterial vaginosis does not seem to increase the cure rate.