Certain predictions are such that their accuracy can be affected by their dissemination, by their being believed and acted upon. Examples of such reflexive predictions are presented. Various approaches to the precise delineation of this category of predictions are explored, and a definition is proposed and defended. Next it is asked whether the possible reflexivity of predictions creates a serious methodological problem for the social sciences. A distinction between causal and logical reflexivity helps support a negative answer. Finally, we consider the thesis that “this characteristic of predictions is peculiar to human affairs” (Merton), and an attempted rebuttal (Grünbaum). It is argued that Grünbaum's rebuttal is ineffective.