Résumé : Understanding the origins of volunteerism and free-riding is crucial in collective action situations where a sufficient number of cooperators is necessary to achieve shared benefits, such as in vaccination campaigns and social change movements. Despite the importance of attaining behavior consistency for successful collective action, the theoretical mechanisms behind this process remain largely elusive. Here, we address this issue by studying the evolutionary dynamics of individual cooperativity levels in multi-round threshold public goods games. By using adaptive dynamics, we explore how individual behavior responses are shaped by incentives, with and without rewards or punishment. We demonstrate that rewarding consistent cooperators can lead to the emergence of two distinct populations: volunteers who consistently cooperate and free-riders who consistently defect. In contrast, punishing consistent defectors does not lead to similar evolutionary branching. Our results help offer insights into designing effective interventions that promote collective action and address collective risk dilemmas ranging from climate mitigation to pandemic control.