Many observers of American politics have been highly critical of the proliferation of primaries that occurred in the 1970s. One of the reasons given for this unfavorable assessment is that the direct primary cannot consistently yield candidates who have broad electoral support-something that is thought necessary to win general elections. And since the major goal of parties is to win elections, this perceived shortcoming is cause for concern.In this article I show, however, that the problem lies not with the direct primary itself, but rather with the rules that govern presidential primaries. In fact, if parties would allocate delegates proportionally to candidates, adopt a preference ballot, and allow independents and "swing" voters to participate, the direct primary would offer an excellent opportunty to nominate candidates with broad support.O ne of the fundamental tenets of American politics is that political parties strive to win elections. From M. Ostrogorski's (1902) investigation of how American parties actually work to Anthony Down's (1957) analysis of how parties ought to behave, this belief has served as the anchor of much of the best discussion of party politics that we have. Yet many scholars claim that the replacement of a politician-dominated system of presidential nominations with one dominated by rank-and-file voters has weakened the capacity of the national parties to nominate candidates with broad electoral support-something that is generally thought necessary to win general elections. Perhaps for this reason, many of the most prominent scholars of political parties have been skeptical, if not openly hostile, to the new system of nominations, viewing them as antithetical to the fundamental purposes of parties (Ceasar, 1979(Ceasar, , 1982Polsby, 1983;Ranney, 1975).Few can doubt that direct primaries remove power from party professionals and give it to voters. But it does not follow that primaries prevent parties from nominating candidates that increase the party's chance for success in November. My aim in this paper is to argue that For thoughtful comments and assistance I would like to thank James D. DeNardo, Stanley Kelley, Jr., Thomas R. Rochon, and the anonymous referees of this journal. A special note of thanks goes to John Zaller, whose help was invaluable in the completion of this study. As always, I bear sole responsibility for the analyses and interpretations presented here.