As global efforts proceed to scale up the delivery of antiretroviral therapy (ART) to HIV-infected persons in most urgent need, it is essential to understand the potential impact of treatment expansion on the transmission of new HIV infections. In this study, we use a series of simple mathematical models to explore the direction and magnitude of treatment effects on the sexual transmission of HIV. By defining the circumstances under which ART can reduce the number of new infections transmitted by treated patients, we provide critical benchmarks to aid in prioritizing efforts to maximize the population health impact of treatment and in evaluating the performance of different treatment programmes. We find that, based on the best currently available evidence of possible treatment effects on patient infectiousness, survival and behavior, the potential remains for either positive or negative changes in overall transmission. In relation to the total number of expected secondary infections caused by each infected person, however, these net treatment effects are relatively modest, particularly if treatment is initiated at advanced stages of the disease. This finding implies that treatment alone should not be expected to alter the population-level incidence of new infections dramatically, in the absence of changes in other factors including possible behavioral responses among uninfected persons and among infected persons who are not yet treatment candidates.