Abstract
Identification of causes of changes in prevalence and incidence of HIV at a national level is important for planning future prevention and intervention needs. However, the slow progression to disease and the sensitive and stigmatising nature of the associated behaviours can make this difficult. Changing rates of incidence are to be expected as an epidemic progresses, but separating background changes from those brought about by changes in behaviour and interventions requires careful analysis. This paper discusses the criteria required to determine whether observed changes in HIV prevalence are the result of changes in behaviour.