We present the results of a lab-in-the-field experiment run with ex-combatants, victims, and non-victims of conflict in Colombia. In our setting, subjects make contributions to a public good paired with members of the three groups. By comparing contributions when paired with members of their own group with contributions when paired with members of the other two groups, we elicit both outgroup discrimination (towards ex-combatants) and discrimination of the ‘other ingroup’ (non-victims for victims, and victims for non-victims). As participants make contribution decisions first unconditionally and then conditionally using the strategy method, we distinguish beliefs-based from preferences-based discrimination. We also elicit empirical and normative expectations of all participants in the different interactions, for which we study the role played by social norms. Our results suggest that victims and non-victims discriminate against ex-combatants significantly more than they discriminate each other, and that the discrimination of ex-combatants is preference-based. Consistently with needs-based and inclusive victimhood models, victims discriminate against ex-combatants significantly less than non-victims. In line with stigma reversal, victims are discriminated by non-victims, even when they do not discriminate against them, and discrimination is driven by high status individuals. While victims and non-victims hold different expectations about each other, and act accordingly, they both exhibit the same discriminatory empirical and normative expectations towards ex-combatants, but non-victims discriminate against them more intensely.