本文利用「臺灣社會變遷基本調查四期四次國家認同組」面訪調查資料，探索臺灣民眾在相關的兩個態度題組（包含外來人口影響與兩岸統合兩個五點量表）的回答模式。首先利用回答模式指標比較每一量表各種可能的回答模式。其次進一步作兩題組共九個題目的探索性因素分析，連續性指標之驗證性潛藏類別因素模型，以及包含與不含回答模式的類別性指標之潛藏類別因素模型分析；同時作各類分析的比較，進而確定可能影響測量誤差的回答模式。最後以多變量分析比較以上三種潛藏類別因素模型中態度類型的推估，尤其著重在包含回答模式與不含回答模式之間的差異。結果顯示，四種回答模式指標指出非極端回答模式的傾向最高。三種因素分析結果顯示，利用類別性指標之四因素模型，產生兩個因素分別代表中立與極端兩回答模式，其模型適合度比兩因素模型比較理想。比較連續性指標兩因素以及類別性指標的兩因素與四因素共三種的潛藏類別因素模型之後發現，考慮到兩種回答模式的因果模式分析效果比較好。其中明顯容易產生中立回答模式者為男性、非大陸各省市的少數族群、中上層以上或下層階級、與小學及以下或專科以上教育程度者；而極端回答模式則較容易出現在女性、非大陸各省市的少數族群、國中及以下教育程度、中層及以上階級與29歲以下或60歲以上的受訪者身上。綜合各種潛藏類別因素模型的結論可發現若未將回答模式從原始的態度區辨出來，性別與自評社會階層的對國家認同態度的影響效果會是「虛假」的，換言之，回答模式在檢視受訪民眾的此兩種量表態度時是不應被忽略的。

This article explores different response styles by using two scales representing attitudes toward immigrants and reunion across the Taiwan Strait. Data were collected from a face-to-face survey-the 2003 Taiwan Social Change Survey. A response style index was first used to compare response style across scales. Exploratory Factor Analysis, Latent Class Factor Model-continuous indicators and Latent Class Factor Model-nominal indicators were then used to compare the response patterns (styles) of the nine attitudinal items. Multivariate analyses with and without including response styles were finally compared across the three factor analysis-type models. The results based on the response style index exhibited a tendency toward a mild response style. The three types of models indicated that the Latent Class Factor Model-nominal indicators for four latent factors provided a better fit than the others, while two response styles were derived from this model-the middle response style and the extreme response style. Concerning social-demographics, males, the ethnic minority in Taiwan, those who identified themselves as high, middle, or low class, and those with less than a primary school or more than a college education tended to choose middle points among the answering scales. On the other hand, females, the ethnic minority in Taiwan, those with less than a junior high school education, those who considered themselves to be middle or high class, and those who were less than 29 or more than 60 years old tended to choose extreme response categories. In conclusion, the identification of response styles revealed that the association between attitudes and two characteristics, namely, gender and social class, could be spurious. In other words, response styles should not be neglected in the examination of the two attitudes demonstrated in this article.