本文旨在將艾里遜(Graham T. Allison)三個決策模式運用於美國 1995 1996 年台海危機之處理上，並且探討台北、北京與華盛頓在決 策過程中究竟犯了什麼錯誤，從而導致危機出現而且不斷擴大。在李 登輝總統 1995 年訪美事件上，三方可說都犯了一些錯誤。惟在中共對 這樣一個私人訪問過份反應而發動軍事演習威脅台灣時，柯林頓政府 輕聲細語的抗議與低調的行事作風，顯然無法遏阻北京的行動。一直 到中共人民解放軍 1996 年 3 月初對準台灣兩大港口 200 海浬的外海， 試射飛彈之後，華盛頓才意識到事態嚴重。柯林頓總統的官員立刻達 成協議，決定採取行動，以免危機昇高為台海衝突。在美國派遣兩艘 航空母艦戰鬥群之後，危機終於解除。然而，這段期間所造成的錯誤 已使三方互信降低，而且新的問題層出不窮。要言之，本文主要即在 指出，美「中」台三方因為一連串的誤判與政策失誤，終於導致台海 危機的出現，儘管危機後來終於化解，但已對三邊關係帶來深遠的影 響。本文認為，在艾里遜所著「決策本質」一書出版近 30 年的今天， 其所提出的三個決策模式仍十分有用，而且資料愈充足，其解釋力與 分析力也就愈強。

This paper intends to apply Graham Allison's three decisionmaking models to the crisis of the Taiwan Strait in 1995-96 and to analyze what went wrong in the process of decision-making in Taipei, Beijing and Washington. Indeed, all parties were at fault in regard to President Lee Teng-hui's visit to the United States in June 1995. When Beijing over-reacted to such a personal trip by launching a series of military drills, the Clinton administration's low-profile protests and warnings failed to stop Beijing's move. It was not until the Chinese People's Liberation Army test-fired missiles only twenty miles or so from the two port cities in Taiwan's north and south then Washington became alarmed in early March 1996. President Clinton's men reached a consensus immediately that they should take certain actions to prevent the crisis from being escalated to a cross-strait conflict. The crisis was over as a result of the U.S. dispatch of two carrier battle groups. However, mistakes made by Taipei, Beijing and Washington during this period have decreased mutual confidence among all three parties, thereby causing more and more new problems to arise. In sum, this paper argues that mistakes made by Taipei, Beijing and Washington triggered the crisis of the Taiwan Strait in 1995-96 and, even after the crisis was over, still have farreaching impact on the subsequent development among the three parties. This paper further confirms that, nearly three decades after his first book entitled "Essence of Decision" was published, Allison's three models are still very useful and that the more data are gathered, the more powerful the interpretative and analytical capability of these three models would be.