This paper aims to encourage greater reflexivity about the limits of prejudice reduction as a model of social change, particularly when applied to societies characterised by historically entrenched patterns of inequality. We begin by outlining some underlying values and assumptions of this model. We then elaborate how our research on political attitudes in post-apartheid South Africa has led us to question, qualify and sometimes reject those assumptions and move towards a ‘contextualist’ perspective on the efficacy of different models of social change. We agree that the project of getting us to like one another may be crucial for producing change in some contexts. In other contexts, however, it is an epiphenomenon that distracts psychologists from the main causes of, and solutions to, problems such as race, class, or gender discrimination. In still others, with an irony that is evidenced increasingly by research, prejudice reduction may actually contribute to the very problem it is designed to resolve. That is, it may diminish the extent to which social injustice is acknowledged, rejected and challenged.