Possible interventions to reduce the spread of human immunodeficiency virus (HIV) include actions that attempt to alter sexual behaviour, such as education aimed at reducing the rate at which individuals acquire new sexual partners, and methods that reduce the probability of transmission between partners, such as the promotion of condom use and the treatment of so-called "cofactor" sexually transmitted diseases. A mathematical model of HIV transmission that is able to mimic different approaches to the control of HIV transmission is employed to study the relative values of different approaches, either used in isolation, or in combination. The nonlinear nature of the term that describes the per capita rate of transmission dictates that for a given degree of intervention, the benefit accruing in terms of reduced HIV spread depends on the prevalence of infection before the introduction of control. Benefit is greatest when HIV prevalence is low. Combination approaches are predicted to be effective but the outcome is less than would be expected on the basis of simply summing the benefits resulting from each type of intervention used in isolation. The success of targeted interventions, aimed at those with high rates of sexual partner change, depends on the heterogeneity in levels of sexual activity within populations and what proportion of the population HIV is able to establish itself in. Targeted interventions are predicted to be very cost effective but their overall success in reducing HIV spread by a significant degree depends on the timing of their introduction (within the time frame of the development of the epidemic) and the pattern of mixing between different risk groups or sexual activity classes.