Survey researchers frequently compare measures of attitudes across surveys where response scales are very similar but not identical, such as when different neutral response options are offered.  However, inconsistent neutral options may affect how respondents report their attitudes .To explore this issue, an experiment by Knowledge Networks was carried out in a web survey  involving attitudes toward how to mark pedestrian crosswalks—a topic specifically chosen due to its non-controversial nature. Respondents (n=692) were randomly assigned to receive a set of seven questions measured on a 5-point agree/disagree scale using either a “neither agree nor disagree” response option placed in the middle of the scale or a “no opinion” response option placed at the end of the scale.  In six out the seven questions, respondents were significantly more than 1.5 times as  likely to choose the “neither agree nor disagree” response when it was available to them than to choose the “no opinion” response when it was available (smallest Χ2=13.91, p<.05).  The results also show that when the neutral points are removed from the analysis, the relative distribution of all other response options is not statistically different between the question versions, which suggest that the wording and placement of neutral response options affects all other response categories equally.  Therefore, excluding inconsistent neutral response categories may allow for more accurate comparability between attitudes measured across different surveys.