Résumé : With the implementation of health impact assessment (HIA)'s conceptual model into real-world policymaking, a number of fundamental issues arise concerning its decision-support function. Rooted in a rational vision of the decision-making process, focus regarding both conceptualisation and evaluation has been mainly on the function of instrumental policy-learning. However, in the field of social health inequalities, this function is strongly limited by the intrinsic 'wickedness' of the policy issue. Focusing almost exclusively on this instrumental function, the real influence HIA can have on policymaking in the longer term is underestimated and remains largely unexploited. Drawing insights from theoretical models developed in the field of political science and sociology, we explore the different decision-support functions HIA can fulfill and identify conceptual learning as potentially the most important. Accordingly, dominant focus on the technical engineering function, where knowledge is provided in order to 'rationalise' the policy process and to tackle 'tame' problems, should be complemented with an analysis of the conditions for conceptual learning, where knowledge introduces new information and perspectives and, as such, contributes in the longer term to a paradigm change.