對於調查資料裡的極端回答，有兩種迥異且矛盾的看法，且兩者都有紮實的文獻支撐。一是極端回答代表極端而明確的態度；二則認為它僅出自回答者的回答風格，並不代表明確態度。這項矛盾意味著極端回答作為明確態度的效度程度，需同時從這兩種看法檢視，但未曾有此種研究。另一方面，訪員常需仰賴外向的人格特質完成訪問，但研究發現外向者卻較可能有極端回答的風格。這讓人想問：合作的受訪者之資料品質如何？本文將外向特質區分為受訪樂意度，及在訪問中友善度，並形成四個假設：⑴在初始態度題以極端回答者，較多是因為有極端回答風格而致，較少是因為明確態度而致。⑵樂意度越高，極端回答風格越強。⑶友善度越高，極端回答風格越強。⑷越樂於受訪的人，其態度越明確。但預期受訪者的態度明確度跟其友善度無關。假設之檢驗，則使用多層線性模式分析2014年「台灣民眾對死刑的態度與相關價值調查研究」計畫資料。假設1的檢定並以AIC比較模型適配度。結果支持研究假設。

Two contradictory perspectives concern the significance of extreme responses in survey data, and each is supported by abundant empirical studies. One school holds that extreme responses indicate extreme and hence strong attitudes. The other considers extreme responses to be merely due to the extreme response style (ERS). The contradiction implies that the extent of the validity of extreme responses as an indicator of strong attitudes may need to be investi¬gated by simultaneous examinations from the two perspectives. How¬ever, no such study exists. Furthermore, interviewers rely more or less on respondents’ characteristics of extraversion to complete an interview. Yet, people higher on extraversion are more likely to have the ERS. This leads to uncertainty about the response quality of cooperative respondents. This study distinguishes two types of extra¬verted respondents: eager participants and friendly respondents. It forms four hypotheses: (1) Initial extreme responses involve ERS more than strong attitudes. (2) Respondents with higher levels of friendliness are more likely to have ERS. (3) Respondents with higher levels of participation eagerness are more likely to have ERS. (4) Respondents with higher levels of participation eagerness are more likely to have strong attitudes. However, we do not expect that respondents with higher levels of friendliness are more likely to have strong attitudes. The hypotheses are tested by analyzing data of the “Survey Research on Attitudes toward Death Penalty and Related Values in Taiwan” in HLM models. AIC is used to compare model fits in testing hypothesis 1. Results support the hypotheses.