Estimating the probability of human immunodeficiency virus (HIV) infection per coital act has been the holy grail of HIV epidemiology for. 2 decades. These estimates are needed for modeling the epidemic and for projecting the effects of preventive interventions. However, the estimates of the probability of transmission per sex act, largely derived from empirical studies and modeling based on HIV-discordant couples, have been troublingly heterogeneous, varying between low-and high-income countries, male-to-female versus female-to-male transmission, stage of HIV infection in the positive partner, the effects of sociodemographic and behavioral characteristics, and sexually transmitted infection cofactors [1–5]. An article by Hughes and colleagues in this issue of the Journal provides valuable new estimates of HIV transmission per coital act in sub-Saharan Africa [6]. This study of 3297 HIV-discordant couples enrolled in a randomized trial of acyclovir suppressive therapy in eastern and southern Africa has the advantage of large numbers, frequent follow-up with quarterly visits, and genetic linkage of the transmitted virus to the index HIV-positive partner so as to exclude infections acquired from external partners. The acyclovir intervention did not affect HIV acquisition [7], so this trial population provides an excellent observational study with which to estimate per coital HIV transmission. In addition, the investigators used rigorous statistical methods to estimate per coital act transmission probabilities and to assess the effects of covariates. However, Hughes et al could only assess infectivity during latent stage disease, and were unable to measure transmission during early or late stages of HIV disease, which is associated with higher infectivity per coital act [1–3, 5, 8]. The investigators did not identify any acute or recent infections among the index positive partners in these stable HIV-discordant couples. Eligibility for enrollment into the randomized trial excluded HIV-infected partners with CD4 counts, 250 cells/mL, and participants whose CD4 counts declined below that level were provided antiretroviral therapy. Also, the enrolled HIV-discordant couples who had accepted voluntary couples counseling and testing had very high rates of condom use (93% of sex acts were reported as protected), so generalizability to other sub-Saharan African settings with substantially lower couples-counseling rates and condom use is problematic [9]. Despite these caveats, the study provides important new information and confirms findings from prior investigations. Overall infectivity was 1–2 cases per 1000 per coital acts, similar to that reported during latency in low-income countries but higher than estimates from many industrialized countries [1, 2, 4, 5]. The HIV load in the index infected partner was the main driver of transmission, with a 2.9-fold adjusted risk of infection per log10 increment in viral load. This is somewhat higher than prior estimates of transmission risk associated with viral load [4, 10, 11], possibly because the short follow-up intervals and frequent viral load measurements allowed more precise estimates of the association between viral burden and infectivity. The male-to-female and female-to-male transmission rates per sex act were similar after adjustment for viral load, which is compatible with other studies from lowincome populations that reported no gender-specific differentials in infectivity [1, 4], but is contrary to findings from high-income countries that suggest greater male than female infectivity [1]. Older age was associated with reduced transmission per sex act, as has been previously reported [2, 4], and male circumcision reduced female-to-male …