本文比較受訪者在完全平衡型量表與最低平衡型量表中發生極端回答、中立回答與默從的傾向，並分析此三種回答模式受到受訪者、訪員、社會距離以及訪問情境等相關因素的影響，以確定回答模式對兩類型量表信效度的影響程度。比較分析以面對面訪問的問卷中跨不同主題且五刻度同意形式的量表為例，資料來自台灣社會變遷基本調查四期四次國家認同問卷組。配對樣本t檢定結果顯示，三種回答模式的傾向在最低平衡型量表中明顯的較高。兩階層的波爾松迴歸與線性迴歸模式的分析結果顯示，社會距離對三種回答模式有明顯的影響，包含對極端回答有影響的性別、婚姻、教育與年齡距離；對中立回答有影響的性別、婚姻與年齡距離；以及對默從有影響的婚姻與年齡距離。極端與中立回答在最低平衡量表中的社會距離效應比較明顯，默從則無類似的發現。雖然最低平衡型量表同樣可以偵測到回答模式，比較建議的仍是較不會產生回答誤差的完全平衡型量表。

In order to explore the extent to which response styles reduce the reliability and validity of attitudinal scales, this article compares three types of response styles including extreme response, middle response, and acquiescence, and their associations with the characteristics of the respondent and the interviewer, social distance and interview situation between fully balanced and minimally balanced scales. Five-point agreement scales across different contents in a questionnaire collected from a face-to-face survey were employed as the example. Data analyzed is from the Taiwan Social Change Survey concerning national identity. A paired-samples t test indicated that response styles are more frequent in minimally balanced scales than in fully balanced scales. Two-level Hierarchical Poisson and Linear Regression Models showed that response styles are significantly affected by social distances. Extreme response is influenced by gender, marital status, education and age distances, while middle response is affected by gender, marital status and age distances. Marital status and age distances influence the tendency to acquiesce. The social-distance effects on extreme and middle response are more significant in minimally balanced scales than in fully balanced scales. Similar results were not found on acquiescence. Although minimally balanced scales can detect response styles, fully balanced scales are recommended due to less response bias in comparison with minimally balanced scales.