Psychological game theory can provide rational-choice-based framing effects; frames influence beliefs, beliefs influence motivations. We explain this theoretically and explore empirical relevance experimentally. In a 2×2 design of one-shot public good games we show that frames affect subject’s first- and second-order beliefs, and contributions. From a psychological gametheoretic framework we derive two mutually compatible hypotheses about guilt aversion and reciprocity under which contributions are related to second- and first-order beliefs, respectively. Our results are consistent with either.