The rapid acceptance of the Web as a vehicle for survey data collection raises important questions for survey designers. Web surveys are the latest example of computerized self-administration of survey questions, and we suspect they may ultimately turn out to be the most popular. Aside from the gains from computerization and self-administration, Web data collection eliminates interviewers entirely, sharply reducing the cost of data collection. Furthermore, Web surveys can deliver rich visual content that is impossible or prohibitively expensive to incorporate in other modes. Not surprisingly, the growth in Web surveys has been dramatic. Despite serious concerns about coverage and nonresponse in Web surveys (Couper, 2001), the commercial research sector has rapidly embraced the Internet for faster and cheaper data collection, and almost daily there are reports of new surveys being done over the Web. A key characteristic of Web surveys is their reliance on visual presentation of the questions. Of course, sound can be added to Web questionnaires, but so far Web surveys have remained a visual medium. Visual presentation is not unique to Web data collection, but is shared to varying degrees with most other methods of self-administration, including mail surveys. Still, the implications of visual presentation are not especially well understood, even for the older methods; the literature on the design of mail or paper-based self-administered questionnaires is not large. Although several good texts offer practical guidelines for the design of paper self-administered questionnaires (e.g., Dillman, 1978; Mangione, 1995), there has been relatively little empirical work or theoretical analysis of the issues involved. The forms design literature is sparse in general (see, e.g., Burgess, 1984; Waller, 1984; Wright and Barnard, 1975). The one notable exception has been the work of Redline and Dillman, who have applied principles rooted in visual perception theory to the design of selfadministered forms (Dillman, Redline, and Carley-Baxter, 1999; Jenkins and Dillman, 1995; Redline and Dillman, 2002). The focus of this work has been on designing forms so that respondents are willing and able to complete them. But the design of paper forms and computer screens may affect not only whether respondents answer the questions but also which answers they give (e.g., Sanchez, 1992; Smith, 1995). The study of forms design is in its infancy, and the impact of forms design on measurement error has been almost entirely neglected. The studies we present here support a few general conclusions about the impact of visual information on responses to questions in Web surveys:
Respondents notice images in Web surveys and the content of these images can affect the answers they give;
Respondents also take in such visual cues as the spacing and relative position of the response options and these cues can alter their interpretation of survey questions;
Respondents are sensitive to information that is immediately visible and may ignore information that is equally critical but not equally available. Taken together, our results suggest that, whether we want them to or not, respondents attend to the visual design of Web questionnaires as well as to the verbal content of the questions.