Abstract: When firms or individuals stand to benefit from doing good, observers often question their underlying motivations and discount their good deeds. We propose that this attribution process is sensitive not only to the presence or absence of selfish rewards, but also to their prior likelihoods. Ten studies show that observers treat uncertain rewards as weaker signals of selfinterest than certain rewards of equal value, leading corporate prosocial initiatives associated with uncertain (vs. certain) profit forecasts and individual good deeds in response to uncertain (vs. certain) incentives to be perceived as more purely motivated and praiseworthy. Indeed, observers see actors who do good when rewards are uncertain as more likeable, benevolent, and likely to behave prosocially in the future, and they prefer products from brands that incur reward uncertainty when pursuing CSR initiatives. Even prosocial actions that turn out to benefit the actor handsomely are judged favorably if rewards were sufficiently uncertain at the outset. We argue that these effects stem from a more general process of counterfactual attribution: It is easier for observers to imagine that actors who incur reward uncertainty would have been willing to act without any incentive at all. These results deepen our understanding of how incentives impact judgments of prosocial behavior specifically, and they illustrate that acting in response to uncertain (vs. certain) incentives conveys different underlying motivations more broadly.