The proliferation of mixed-mode surveys, where data is collected from respondents using different survey modes, raises concerns about whether respondent characteristics are being measured equivalently across modes, as the data often are combined for analysis. Mode differences often encourage survey designers to construct questions differently, depending on the mode being used to survey respondents. In this paper, we compare scalar formats using both telephone and web modes to identify ways of asking scalar questions that present the same stimulus to respondents across telephone and web modes so scalar questions can be constructed optimally for mixed-mode surveys. We include several experimental manipulations designed to compare the same scales across telephone and web modes as well as different scales within these modes. Our results indicate that mode and format independently influence responses to scalar questions; however, the interaction of mode and scalar format is not significant in any of the models. Overall, we find telephone respondents provide more positive ratings than web respondents and are more likely to select the most positive extreme endpoint category. We also find differences across scales within telephone and web modes.