Human societies are based on cooperation among large numbers of genetically unrelated individuals. This behavior is puzzling from an evolutionary perspective. Because cooperators are unrelated it cannot be the result of kin selection, and the large scale seems to preclude explanations based on direct reciprocity. Alexander (1987) has proposed that large-scale cooperation among humans can be understood as resulting from networks of "indirect" reciprocity. For example, individual A may help individual B even though A receives no direct reciprocal benefit. Instead, B might help C who helps D who finally returns the help indirectly to A. Here we describe a simple mathematical model of the evolution of indirect reciprocity. Analysis of this model suggests that indirect reciprocity is unlikely to be important unless interacting groups are fairly small.