本文主旨在重新思考民意調查的本質。本文作者運用並拓展：傳播學 之框架理論觀點，指出一般媒體民調機構在產製民調結果時存有四大框 架機制，即⑴樣本代表性的框架；（2)問卷設計的框架；（3)訪員特性的框 架；以及（4）標準化訪問的框架。本文作者認為，民意調查的本質乃是一 套產製民意的框架機制，也是一套建構民意的再現系統。

This paper aims to reexamine the essence of public opinion poll. By collecting and rearranging major media polling results of the approval rating in the 2000 presidential campaign in Taiwan, the author argues that there have been significant gaps among major media poll agencies, which is beyond our common knowledge of polls from the traditional statistical sampling errors view. Facing is puzzling phenomenon, the author extends the framing theory of communication to explain the significant gap, arguing that at least four frame mechanisms in media polls have caused the differences. That is: (1) the Frame of Sample Representatives, (2) the Frame of Questionnaire Design, (3) the Frame of Interviewer Characteristics, and (4) the Frame of Standardized Interviewing. The author claims that because media poll agencies have not adopted the same framing mechanisms, we will not obtain consistent polling results. Finally, the author concludes that the essence of public opinion poll is a complicated “crystal” composed of various framing mechanisms, which shapes the public opinion; and the poll is also a cultural representative system, which constructs the public opinion.