OBJECTIVE: To describe the dynamics of HIV transmission in stable sexual partnerships in rural Tanzania. DESIGN: Retrospective cohort study nested within community-randomized trial to investigate the impact of a sexually transmitted disease treatment program. METHODS: A cohort of 1802 couples was followed up for 2 years, with the HIV status of each couple assessed at baseline and follow-up. RESULTS: At baseline, 96.7% of couples were concordant-negative, 0.9% were concordant-positive, 1.2% were discordant with the male partner being HIV-positive, and 1.2% were discordant with the female partner being HIV-positive. Individuals living with an HIV-positive partner were more likely to be HIV-positive at baseline (women: odds ratio [OR] = 75.7, 95% confidence interval [CI]: 33.4-172; men: OR = 62.4, CI: 28.5-137). Seroincidence rates in discordant couples were 10 per 100 person-years (py) and 5 per 100 py for women and men, respectively (rate ratio [RR] = 2.0, CI: 0.28-22.1). In concordant-negative couples, seroincidence rates were 0.17 per 100 py in women and 0.45 per 100 py in men (RR = 0.38, CI: 0.12-1.04). Individuals living in discordant couples were at a greatly increased risk of infection compared with individuals in concordant-negative couples (RR = 57.9, CI: 12.0-244 for women; RR = 11.0, CI: 1.2-47.5 for men). CONCLUSION: Men were more likely than women to introduce HIV infection in concordant-negative partnerships. In discordant couples, incidence in HIV-negative women was twice as high as in men. HIV-negative individuals in discordant partnerships are at high risk of infection, and preventive interventions targeted at such individuals are urgently needed.