This paper investigates between-voter interactions in a social network model of turnout. It shows that if 1) there is a small probability that voters imitate the behavior of one of their acquaintances, and 2) individuals are closely connected to others in a population (the “smallworld” effect), then a single voting decision may affect dozens of other voters in a “turnout cascade. ” If people tend to be ideologically similar to other people they are connected to, then these turnout cascades will produce net favorable results for their favorite candidate. By changing more than one vote with one’s own turnout decision, the turnout incentive is thus substantially larger than previously thought. We analyze conditions that are favorable to turnout cascades and show that the effect is consistent with real social network data from Huckfeldt and Sprague’s South Bend and Indianapolis-St. Louis election surveys. We also suggest that turnout cascades may help explain over-reporting of turnout and the ubiquitous belief in a duty to vote. I thank Robert Bates, Lars-Erik Cederman, Eric Dickson, Paul E. Johnson, Orit Kedar, Gary King, Ferran Martinez I Coma, and Ken Shepsle for valuable feedback on earlier drafts. How does the turnout decision of a single person affect an election? Decision-theoretic models of voting show that the probability of one vote being “pivotal ” in a large electorate is extremely small (Tullock

This article analyzes the distinction between two de®nitions of rationality, standard expected utility maximization and an alternative that says rational individuals maximize their expected utility given their acts. The latter version can explain important political phenomena that challenge the standard approach. Documenting the empirical advantages of the second de®nition for political science, the article also argues that conditional expected utility maximizers are normatively rational. Moreover, the article demonstrates that some important andrecognizedwork in political science has unwittingly reliedon this version's underlying assumptions. Finally, it formally shows how this decision theory can ground equilibria in two-person games.