In 3 experiments, this article compares how overhearers interpreted second speakers' contributions to a conversation depending on whether the second speaker responded to a first speaker immediately; paused and responded; said um and responded; or said um, paused, and then responded. The conversational snippets tested were unscripted and diverse; an example of one exchange is, "Are you here because of affirmative action?" (pause, um, or both) "It helped me out a little bit." Overhearers thought speakers had more production difficulty, were less honest, and were less comfortable with topics under discussion when speakers either said um or paused, and even more so with both. The best explanation for the data is that overhearers are judging, for each question asked, what it means for speakers to produce an anticipated or an unanticipated delay.Pauses, ums, and uhs are popularly understood as meaningless or as hindrances to good communication. One common conception of pauses, ums, and uhs is that they are all versions of the same thing. The traditional label for ums and uhs in the research literature, filled pauses, emphasizes this seeming interchangeability. In this view, whether a person uses an um or an uh or a pause would be a matter of personal style or habit, like speaking quickly or speaking softly. However, researchers have shown that they do have meaning; when estimating respondents' knowledge of the answers to questions, overhearers interpreted ums and uhs as amplifying what pauses did (Brennan & Williams, 1995). For example, overhearers judged an immediate "I don't know" to "What is the answer to Question 4?" as reflecting more commitment to not knowing the answer than "[pause] I don't know," which in turn reflected more commitment than "um I don't know." Similarly, overhearers thought an immediate answer, such as "Ottawa," reflected more commitment to the answer than "[pause] Ottawa," which reflected more commitment than "um Ottawa." Therefore, it mattered whether overhearers heard a pause or an um or an uh, DISCOURSE PROCESSES, 34(1), 37-55 Copyright © 2002, Lawrence Erlbaum Associates, Inc.Correspondence and requests for reprints should be sent to Jean E. Fox Tree, Psychology Department, University of California, Santa Cruz, Santa Cruz, CA 95064. E-mail: foxtree@cats.ucsc.edu although ums and uhs had similar effects (Brennan & Williams, 1995). In other work, researchers have suggested further that ums and uhs are used to signal upcoming delays, with ums signaling major delays and uhs minor (Clark & Fox Tree, 2002;Clark & Wasow, 1998;Fox Tree, 2001;Smith & Clark, 1993).There has been some experimental evidence to support the delay-signaling view. Smith and Clark (1993) found that in answering factual questions like, "What is the capital of Canada?," people reliably used um before a long delay and uh before a short delay in the amount of time it took for them to answer. Fox Tree (2001) found that people recognized words in a speech stream more quickly after uh than when the uh was digitally excised. In t...