訪問調查時很重要的前提是確保訪員與受訪者進行訪談時的個人隱私。但現實中很難避免第三者在場的訪問情境。相關研究結果雖然分歧（Blair, 1979; Hartmann, 1995），一些針對高敏感與社會規範性題目的研究仍發現顯著的第三者在場效應（Taietz, 1962; Aquilino, 1993; Smith, 1997），只是從不同性質的第三者在場深入探討其對訪問回答的影響。本研究探討相較於事實性題目的回答時，三面向的第三者在場（分別是不同類型的第三者、在場人數與在場的期間）是否較容易造成受訪者被詢問敏感性及社會規範性題目時提供偏差的回答（例如：低報，社會規範的回答方向）。研究資料來自一項針對大都會地區的調查，對象是戶籍設在台灣，年滿20 歲的民眾。分析的結果顯示，有50–60% 的訪問是在有第三者在場情況下進行的，在場人數多半是1 人。約40% 的第三者在場是持續到整個訪問結束為止。在個人收入與性經驗題目的回答中，相對於無偏差的實質回答，偏差回答較容易受到一到三個面向的第三者在場的影響。在回答傾向方面，無論是偏差回答刪除之前或之後的實質回答分布，顯著的第三者在場效應多發生在同居、婚外情、墮胎、未婚媽媽、個人收入、以及政黨支持等題目的回答，尤其個人收入的偏差回答均受到三面向第三者在場顯著的影響。本研究再次確認第三者在場對社會規範性與敏感性題目的回答品質與回答傾向之影響符合理論解釋，也呼應前人研究的提示，情境效應會因問卷題目的性質不同而有差異。

The assurance of confidentiality in a dyad interview (private interview) is central to survey interviewing, and it is a premise of such interviewing. In reality, however, the presence of third-parties, a situational variable in the interview, is hard to avoid. Despite the mixed results found in previous studies (Blair, 1979; Hartmann, 1995), some findings were supportive of a significant third-party effect, especially those with a high degree of sensitivity and social norms (Taietz, 1962; Aquilino, 1993; Smith, 1997). Furthermore, scant attention has been paid to the in-depth nature of the presence of others. Three dimensions of the presence of others, respectively the number of others, the types of others and the duration of their presence, were examined in this study under the assumption that they would lead the respondents to underreport or to provide socially desirable answers to sensitive or social normative questions, as seen in comparison with factual questions. Data from a regional survey of metropolitan residents aged 20 or older in Taiwan were used. The results indicated that 50–60% of interviews carried out when third-parties were present, mostly one person. About 40% of third-parties were present all the time. Concerning personal income and sexual experience, in contrast to bias-reduced substantial responses, biased responses are more likely to occur in the interviews characterized by one to three dimensions of third-party presence. In terms of response tendency, significant third-party effects were found among substantial responses and bias-reduced substantial responses to the questions including cohabitation, extra-marital relations, abortion, unmarried women, personal income, and political-party identification. In particular, the third-party effect on biased response to personal income was consistently found significant across three dimensions of the presence of others. The present findings reconfirm the theoretical expectations for the third-party effect on response quality and response tendency to social normative and sensitive questions. They also echo what the previous studies have indicated: the situational effect varies with questions of different nature.