Among the many issues faced by Web survey designers is whether to offer an explicit “Don’t Know” or other non-substantive response. Prior research has demonstrated that the presence or absence of a non-substantive response option can have a significant impact on both the level of item nonresponse and the distribution of substantive responses. This paper reports on an experiment that tests three methods for handling of non-substantive responses: (1) No DK or other non-substantive option on the screen but respondents are not forced to answer; (2) a DK or some other non-substantive option presented; and (3) no DK or other non-substantive option presented but respondents who do not answer are prompted with a message encouraging them to answer. These formats are varied across a number of different question types including scalar questions presented in grids and questions known to elicit high levels of nonresponse such as income. Both attitudinal and behavior questions are included.

Analysis of these data is focused on the degree to which different presentation formats may encourage or discourage non-reporting. A second analytical issue is the impact of presentation format on the distribution of substantive responses. For example, do formats that discourage nonresponse generate different distributions that those that do not? Finally, we are very interested in the impact of different formats on undesirable respondent behaviors such as straight lining, speeding, and mid-survey termination.