Our aim was to determine the efficacy of HIV-1 voluntary counselling and testing (VCT) in reducing unprotected intercourse among individuals and sex-partner couples in Nairobi (Kenya), Dar es Salaam (Tanzania), and Port of Spain (Trinidad).Individual or couple participants were randomly assigned HIV-1 VCT or basic health information. At first follow-up (mean 7.3 months after baseline) health-information participants were offered VCT and all VCT participants were offered retesting. Sexually transmitted infections were diagnosed and treated at first follow-up. The second follow-up (mean 13.9 months after baseline) involved only behavioural assessment, and all participants were again offered VCT.3120 individuals and 586 couples were enrolled. The proportion of individuals reporting unprotected intercourse with non-primary partners declined significantly more for those receiving VCT than those receiving health information (men, 35% reduction with VCT vs 13% reduction with health information; women, 39% reduction with VCT vs 17% reduction with health information), and these results were maintained at the second follow-up. Individual HIV-1-infected men were more likely than uninfected men to reduce unprotected intercourse with primary and non-primary partners, whereas HIV-1-infected women were more likely than uninfected women to reduce unprotected intercourse with primary partners. Couples assigned VCT reduced unprotected intercourse with their enrolment partners significantly more than couples assigned health information, but no differences were found in unprotected intercourse with non-enrolment partners. Couples in which one or both members were diagnosed with HIV-1 were more likely to reduce unprotected intercourse with each other than couples in which both members were uninfected. These changes were replicated by those in the health-information group diagnosed with HIV-1 at first follow-up.These data support the efficacy of HIV-1 VCT in promoting behaviour change.