This article examines a fundamental aspect of democracy: the relationship between the policy positions of candidates and the choices of voters. Researchers have suggested three criteria—proximity, direction, and discounting—by which voters might judge candidates' policy positions. More than 50 peer-reviewed articles, employing data from more than 20 countries, have attempted to adjudicate among these theories. We explain why existing data and methods are insufficient to estimate the prevalence of these criteria in the electorate. We then formally derive an exhaustive set of critical tests: situations in which the criteria predict different vote choices. Finally, through survey experiments concerning health care policy, we administer the tests to a nationally representative sample. We find that proximity voting is about twice as common as discounting and four times as common as directional voting. Furthermore, discounting is most prevalent among ideological centrists and nonpartisans, who make sophisticated judgments that help align policy with their preferences. These findings demonstrate the promise of combining formal theory and experiments to answer previously intractable questions about democracy.