[en] This paper explores the relationship between survey rating scale and Extreme Response Style (ERS) using experimental data from Understanding Society (Innovation Panel 2008), where a self-assessment questionnaire measuring job satisfaction uses two alternative (7 and 11 points) rating options. Our results suggests that when shifting from a shorter to a longer scale, the survey design generates a tendency to choose response scales at the extreme of the distribution, thus creating a misleading quantification of the variable of interest. The experimental design of the data enables us to test our hypothesis using a non-linear estimation approach where age, gender and education level are shown to affect ERS.