本文主要是從總體與個體兩個層面，以統獨議題做為分析的對象，來探討臺灣民眾政治態度的變遷。首先，不同於一般的靜態分析，我們引入時間序列分析以提供清楚的動態關係解釋，並有助於理解總體政治訊息傳播對大眾政治態度形成的作用。再者，本文以理性貝氏動態模型 (Rational Bayesian Dynamics Model) 作為個體模型，主張所謂「大眾政治態度」是民眾在既有態度基礎上，對當下異質政治訊息學習與權衡的結果。基於上述對於政治行為研究個體模型與動態變遷的雙重關懷，我們以教育作為個人政治知識的間接指標，證明在不同的政治情境與資訊環境下，不管其省籍背景，個人是如何透過政治資訊的學習，理性地「調整」其統獨立場。最後，藉由統獨態度變遷分析的發現以及晚近政治行為研究文獻上的反省，我們認為所謂民調並不是在建構一個「穩定與其實的」態度。相反的，木文主張民眾對民調的反應是一個「政治過程」，取決於既存的意識形態與菁英政治論述競爭之間的互動。而在這個民調「問」與民眾「答」的過程中，議題的「不確定性」與個人的政治知識不僅影響民眾呈現「態度」的能力與意願，同時也形塑臺灣民主體制下理性政治說服的形成與運作。

　　　　　To elaborate the dynamics of the general mass attitudes in Taiwan toward the Unification-Independence issue, we explore both aggregate group behavior and individual level heterogeneity by converting consecutive cross-sectional data into a timedependence structure. Apart from the current static models, these pseudo-panel analyses identify the relations between Elites' Rhetoric and the formation of mass choices on Taiwan's future. Of modeling efforts, education and ethnic ity variables are included in both static and dynamic models to statistically estimate their respective significance, their political impacts, and changing pace. In particular, declining significance of ethnicity in dynamic behavior tends to signal the emerging importance of political knowledge in understanding political learning game. In light of rational Bayesian updating, preference revelation is defined as political processes through which mass learn to adjust with competing messages as well as limited information transmissions. In addition to unraveling underlying historical dynamics, we argue that individual uncertainty on this issue not only determines the results of reporting preference, but also reshapes the rationality of political persuasion in this newly democratic polity.