There is mixed evidence of how framing affects contributions in public good games and different transmission channels for framing effects have been suggested. To disentangle this we study the strategy version of the public good game in a large internet experiment. Our experimental setup controls for beliefs about other players contributions and reveals that framing dramatically effects subjects’ misperceptions about which contribution strategy maximizes personal income. When controlling for the types of misperception that we measure we find that the framing effects are reduced substantially. Hence, the results suggest that framing-induced differences in the degree of misperceptions of the game are important for understanding framing effects. (with with Toke Fosgaard, Lars Gårn Hansen and Jean-Robert Tyran.)