在研究投票行為時，政黨認同常被認為是最重要的解釋因素。但是美國選民「分 裂投票」情形的逐年增加，正反映出選民的政黨認同在減弱。兩種理論模型在本論文被用來 解釋選民的「分裂投票」行為。一是「政策中和模式」，此模式假設選民投票只是看政策， 因為選民對兩黨的政治人物不信任，希望藉由「分裂投票」產生「分治政府」，以使兩黨「 選民期望模式」，此種理論認為：選民是針對不同選舉，一一投票的。因為不同類別的選舉 ，影響選民投票決定的因素不同，選民對各種職位的期望也不相同。結果便造成選民「分裂 投票」的行為。本文的目的即在利用 1996 年美國選舉的資料，來檢測此二模式的解釋力。 結果發現，政黨認同的強弱，仍是影響選民「分裂投票」與否的關鍵因素。「政策中和模式 」及「選民期望模式」對受訪者是否「分裂投票」，都有相當的解釋力。

　　　　　When we study voting behavior, we used to say that party identification is the dominant explanatory variable. But with the increase in split-ticket voting in the American elections, party identification has been declining and voters are less affected by party affiliations. These changes also pushe researchers to look for other factors to explain split-ticket-voting behavior. There are two issue models explaining split ticket voting. The first is the "policy-moderation model," which assumes that splitticket voters wish to have a divided government, to have administrative and legislative powers belong to different parties, and to balance policy output. The second model comes from voters' expectations. This model assumes that voters have different expectations of the President and Congress, and that the two main parties excel in different issues. Consequently, voters are motivated to vote split ticket by their different expectations. The purpose of this article is to test these two models which the 1996 survey data. We found that the strength of party identification is still the dominant variable, and that both of the models have explanatory power.