Likert scales are popular for measuring attitudes, but response style, a source of measurement error associated with this type of question, can result in measurement bias. This study investigates the effect of data collection mode on both types of response styles using data from the 2012 American National Election Studies (ANES). 2012 was the 1Â year in which ANES conducted two parallel surveys, one through face-to-face interviews and another through Web, using two independent national probability samples and an identical questionnaire. We used three sets of balanced Likert scales from the survey to measure the acquiescent and extreme response styles. Using the latent class analysis modeling approach, we find that: (1) both acquiescent and extreme response styles exist in both face-to-face and Web survey modes; (2) face-to-face respondents demonstrate more acquiescent and extreme response styles than Web respondents; (3) the mode effect is similar for white, black and Hispanic respondents.