When survey research, statistics, and electronic data processing were first introduced, they held out promise that a new level of political knowledge would be created. Applied to the study of voting behavior, survey research promised an understanding of the factors determining the outcome of an election, that political history could be based on rich and current data, and that we could begin to understand the role of elections in constitutional democracy. The truth as Peter B. Natchez shows, is that despite the opportunity provided by this revolution, voting studies have failed to make...