Earlier studies on issue voting in Europe and North America have found that the relative performance of two competing models, the proximity model and  the directional model, depends on the methodologies used, the nature of the party systems, the issue being examined, incumbent-challenger distinctions,  and voter sophistication. The purpose of this paper is to test these two models in the Japanese electoral context.
 Empirical tests, using survey data from Japan’s 1995 Upper House election, demonstrate that the proximity model best explains the behavior of centrist  parties, and the directional model best explains the behavior of extreme parties. This paper also shows the advantages of the directional model for  predicting vote choice when voters have particular issues of strong interest. Finally, I shall suggest that the number of issue-dimensions may be of  crucial importance in evaluating the performance of the models.