This paper reviews the ways in which data on sexual behavior can contribute to the understanding of HIV prevalence trends based on sentinel surveillance, building on work presented at a meeting on new strategies for HIV/AIDS surveillance in resource-constrained countries, held in Addis Ababa in January 2004. A key component of second-generation surveillance is the collection of data on the behaviors and background characteristics that may influence the course of the HIV epidemic in a population. However, the most appropriate methods for the collection and analysis of these data for various types of epidemic have yet to be established. A conceptual framework is presented outlining the relationships between background characteristics, behaviors and HIV infection. The different methods used to collect data on HIV infection, risk behaviors and background characteristics in generalized and concentrated epidemics are reviewed, including population-based surveys, and surveillance in high- and low-risk groups. The various biases inherent in different approaches are discussed. The implications of linking data at the individual and community levels are explored and recommendations made concerning appropriate analytical approaches, drawing on an example of a pilot study that linked biological and behavioral surveillance in Tanzanian antenatal clinics. The paper concludes with recommendations for the methods and frequency with which to collect the data required for second-generation HIV surveillance.