It has been argued that the visual composition of web survey pages influences response rates, but there is little evidence that such effects are significant. In addition, many admonitions have been proposed for font selection, background color, and other features, but there is little experimental evidence that such features affect responses. Our purpose in this paper is to present results from an experiment in which three quite different web page designs were tested with regard to their impacts on response rates, early terminations, item omissions, subjective evaluations by respondents, amount of time spent answering, and measurement effects. Two web designs were created, which were theorized to create negative respondent reactions and compared with a control that had been successfully used in many previous surveys, and appeared to produce few if any volunteered negative reactions. The alternative designs used as a frame of reference Donald Norman’s theory on emotional usability. One treatment was designed to create viscerally negative reactions while the second was designed to create behaviorally displeasing reactions. The attempt to create designs that might have a negative effect on response was undertaken in hopes of identifying elements that create barriers to achieving high response rates. The alternative designs were administered in a 29 question survey in February, 2007, to a random sample of Washington State University undergraduates, with 600 students assigned to each panel. We found that although item omissions, response time, and subjective reactions varied significantly across treatments, the final response rates (51-54%) were quite similar. Overall, the results suggest that visual composition of web survey pages may have less influence on response rates and measurement that has often been theorized. The results also suggest several important avenues for follow-up research.