Lack of respondent engagement is a concern in web surveys. In an attempt to combat this problem, survey practitioners often try to design web questionnaires in ways that they believe respondents will find visually appealing. However, prior research has found that, when answering survey questions, respondents may draw on features that researchers intended to be purely stylistic. For example, when the appearance of response option labels varies across the options, respondents may perceive the options to be further apart conceptually than when they are presented in a similar fashion to one another. This may in turn affect the survey responses that respondents provide. This presentation will discuss the results of an experiment included in a web survey with more than 2,000 respondents. The respondents were randomly assigned to one of four experimental conditions that varied the format of the response options: one group got standard labels for each option, the second got labels that increased in boldness from left to right, the third got labels that decreased in boldness from left to right, and the fourth group got labels with emoticons. In addition, respondents were randomly assigned to one of two response option orders, and to either a fully-labeled or endpoint-only-labeled scale. The results of this experiment indicate that these three factors may influence survey responses, suggesting that practitioners should use caution when adding stylistic features to web questionnaires.