Publication selection processes can distort the probabilities associated with hypothesis tests. The paper defines a framework in which such distortion can be explored and considers,for illustration, the distortion of the power function of a particular test.