Investigative interviewers seek to obtain complete and accurate accounts of events from witnesses. Two studies examined the influence of instructions about the use of don't know (DK) responses and of clarifying the meanings of DK responses on the quality of responding to questioning. Participants watched a video, and after a delay (Study 1, 30 min; Study 2, 1 week) were randomized to a DK encouraged, DK discouraged, or control group. They then responded to answerable and unanswerable questions, after which they clarified the meanings of DK responses. Across studies, individuals encouraged to use DK responses answered fewer questions and made fewer errors at initial questioning. Discouraged and control participants showed similar performance, suggesting that interviewees assume that DK responses are not desired unless otherwise instructed. Clarifying the meanings of DK responses revealed that a majority of DK responses were correct statements about the presence or nonpresence of information in the video. The encouraged group showed greater gains in output after clarification while maintaining lower errors. Encouragement and clarification of DK responses were each associated with higher diagnosticity that substantive answers were in fact correct responses to answerable questions. Encouraging DK responses and clarifying the meaning of DK responses leads to more accurate reports in response to questioning. Encouraging DK responses reduces the tendency to overreport, which can reduce the quality of responding. DK responses frequently convey different meanings that, if clarified, can lead to useful information about the occurrence or nonoccurrence of information. (Contains 3 tables, 1 figure and 1 footnote.)