Stigma is a recognised problem for effective prevention, treatment, and care of HIV/AIDS. However, few studies have measured changes in the magnitude and character of stigma over time. This paper provides the first quantitative evaluation in Africa of the changing nature of stigma and the potential determinants of these changes. More specifically, it evaluates the dynamic relationship between stigma and (1) increased personal contact with people living with HIV/AIDS and (2) knowing people who died of AIDS. Panel survey data collected in Cape town 2003 and 2006 for 1074 young adults aged 14-22 years were used to evaluate changes in three distinct dimensions of stigma: behavioural intentions towards people living with HIV/AIDS; instrumental stigma; and symbolic stigma. Individual fixed effects regression models are used to evaluate factors that influence stigma over time. Each dimension of stigma increased in the population as a whole, and for all racial and gender sub-groups. Symbolic stigma increased the most, followed by instrumental stigma, while negative behavioural intentions showed a modest increase. Knowing someone who died of AIDS was significantly associated with an increase in instrumental stigma and symbolic stigma, while increased personal contact with people living with HIV/AIDS was not significantly associated with any changes in stigma. Despite interventions, such as public-sector provision of antiretroviral treatment (which some hoped would have reduced stigma), stigma increased among a sample highly targeted with HIV-prevention messages. These findings emphasise that changes in stigma are difficult to predict and thus important to monitor. They also indicate the imperative for renewed efforts to reduce stigma, perhaps through interventions to weaken the association between HIV/AIDS and death, to reduce fear of HIV/AIDS, and to recast HIV as a chronic manageable disease.