The objective of HIV surveillance is to document trends in HIV prevalence and transmission risk in order to make informed policies and to guide prevention and care/treatment programmes. Zambia established a comprehensive HIV surveillance system in 1994 that provides data on prevalence trends in both urban and rural areas based on over 20 sentinel sites using data from antenatal clinic (ANC) attendees. Furthermore, population-based HIV surveys have been conducted regularly since 1995 in selected communities to validate the representative profile of pregnant women and to measure the trends in HIV prevalence and transmission risks concomitantly. Since then, repeated national sexual behaviour surveys and one nationally representative HIV survey among adults have been conducted. This national survey from 2002 revealed that the HIV prevalence among adults in Zambia was 15.6% (23.1% in urban areas and 10.8% in rural areas). The objective of this thesis is to examine trends in HIV prevalence and sexual behaviours that are likely to have had an impact on HIV infection in Zambia since the mid 1990s. The discussion of the associations between HIV, risk behaviours and underlying sociodemographic factors builds on the proximate determinants framework and takes into account the importance of the epidemiological context. A significant declining trend in HIV prevalence among pregnant urban women and a modest decline among rural women was found in the data from the 22 antenatal sites that were part of the ANC-based HIV surveillance system throughout the period 1994 to 2002 in Zambia. The decline was clearest among urban women aged 15-24 with higher education, and it is likely that this drop reflected a reduction in incidence of infection. However, individual sites showed diverging trends. This could be a sign of different epidemic stages, but also differences in intensity and effectiveness of HIV prevention in these areas as well as cultural or structural factors affecting transmission. Both national data and data from selected communities revealed a shift towards safer sexual behaviour in Zambia since the mid 1990s with a decrease in the proportion of people reporting multiple or concurrent partners and early sexual activity, and an increase in condom use with casual partners. A parallel increase in age of the first birth for women strengthens the credibility of the reported behaviour changes. Logistic regression analyses showed that the same indicators probably contributed to the decline in HIV prevalence observed among sexually active young people between 1995 and 2003 in selected urban and rural areas. The changes in reported sexual behaviour were clearest among young urban people who had completed more than secondary education. Groups with higher education appeared also with the most marked decline in HIV prevalence. Although there are many sources of potential bias, these consistent trends are likely signs of real behaviour changes that have contributed to the decline in incidence indicated by the drop in the prevalence among young people. Higher socioeconomic status and education were associated with having more sexual partners before the HIV epidemic in Africa, probably due to a combination of factors that promote sexual networking, such as higher mobility, urban residence and financial ability. Early surveys showed higher levels of HIV prevalence among wealthy and educated groups. Both in Zambia and several other sub-Saharan countries this pattern has now reversed. Multiple sexual partners, low condom use and HIV infection have become associated with lower socioeconomic status. As knowledge about risk factors becomes available, people of higher socioeconomic status will usually possess the resources and self-efficacy to change their behaviour and avoid disease. In groups with fewer resources such changes usually occur more slowly. When a substantial proportion of a population has multiple concurrent sexual partners this can lead to rapid and widespread transmission of sexually transmitted infections, including HIV, in the population as large groups of people are connected at stages during which the infectiousness of the disease is high. Mathematical models have demonstrated that this can have a major impact on the growth of an epidemic in its early phases. There are no data on the prevalence of concurrent sexual partnerships in Zambia from the early 1980s when the HIV epidemic probably took off, but data from the period 1998 to 2003 showed a decline in concurrency among both urban and rural men. The fact that a decline was observed in groups with different educational attainment makes it plausible that a certain decrease in the prevalence of concurrency had already taken place by 1998 in higher socioeconomic groups. The observed behavioural changes in Zambia could be due to successful HIV prevention campaigns, but may also reflect changes in risk due to the visible effects of the epidemic in the communities. There has not been a national system for monitoring and evaluation of HIV prevention programs in the past in Zambia, and thus it is not possible to determine the contribution of specific programs to HIV prevalence decline. Empirical data have revealed that the effect of interventions on HIV incidence depends on the epidemiological context; i.e. the same intervention can have different impacts in different communities depending on the distribution of risk factors, sexual behaviour, stage of the epidemic, HIV prevalence/incidence in the target population, prevalence of other STIs, and sexual mixing between high-risk groups and other population groups. There is a need for a combination of research approaches - in addition to the established HIV surveillance systems - to obtain this information. In countries with high prevalence of HIV in the general population, high risk groups can be an important source of new infections although a lot of the transmission takes place among individuals perceived to have low risk. However, it may be harder to identify high-risk/core groups than in low prevalence settings as the risk of HIV transmission is high in the general population too. One hypothesis is that focusing on high risk places such as venues where people meet new sexual partners would be more effective in reducing the transmission rate of HIV at the community level rather than targeting interventions just at the perceived high risk groups. So-called PLACE-assessments of venues where people meet new sexual partners in two cities in Zambia revealed a high level of unprotected sex among people who were socializing there, but also a high likelihood of using condoms if they were available in the venue or nearby. There was a lack of the signs of current preventive campaigns targeting these high-risk venues, and the assessment identified opportunities for prevention that potentially could have a significant impact on the serious epidemics in these urban settings. Despite signs of improvements, Zambia still suffers from an extensive HIV epidemic with high incidence rates. It is of critical importance to further strengthen the prevention of new infections. This represents a particular challenge when priorities seem to be shifting from prevention to antiretroviral treatment.