Recent research on political expertise indicates that what citizens know and how much they think about politics affect the political choices they make. So it would seem for issue voting in presidential elections. Unfortunately, prior work has yielded such conflicting results that we lack a clear understanding of how expertise affects the vote. Drawing on research from social and political psychology, I argue that the accessibility of policy attitudes from memory depends on political expertise. Given greater accessibility of policy attitudes, issue voting should be more pronounced at higher levels of expertise. In contrast to most previous work, this research measures expertise with interval-level knowledge scales and employs formal interaction tests. Data from the 1984 and 1988 National Election Study surveys are used to test my hypotheses. Results show that increasing expertise results in higher levels of sociotropic, ideological, and policy voting. If representative democracies are to function properly, elected officials must address the policy concerns of the public and wield government power to solve these problems. Public officials who respond effectively to these concerns should be rewarded with election to office, and those who fail to do so should be denied office. This responsiveness depends in turn on the clarity of the policy signals articulated by the electorate. In this regard, voters perform reasonably well in presidential elections. They hold elected officials and incumbent parties accountable for their governing performance (Dobson and St. editors past and present, and three anonymous reviewers for their many helpful comments and suggestions. Any and all remaining errors are of course my own.