Abstract
The debate over the relative merits of directional and proximity models as theories of candidate evaluation has turned largely on the criterion of predictiveness. This article adds to the predictive record for France and the United States in 1988 but also focuses on the criterion of verisimilitude. By examining the interaction between model choice and measurement technique I show that the proximity model based on idiosyncratic voter perceptions of candidate issue positions produces a more plausible representation of the distribution of issue-based candidate preferences of the voters than the directional model based on national mean perceptions of candidate issue positions.