9309612 Clark Language use in everyday discourse is a joint activity. Speakers and their addressees coordinate closely with each other as speakers try to attend to, identify, understand, and take up what they say. This project is about one level of that process: how speakers formulate and present utterances for their addressees in real time. The problem is that time moves inexorably onward, and speakers often cannot formulate their utterances in the time available. This lads to disfluences (pauses, uh, self-corrections, repeats, restarts, etc.). It also leads to the ordering of phrases based on how heavy, or complex, they are. The goal is to develop a model of formulation and presentation that accounts for these phenomena. The project contains five series of studies. Series 1 will examine disfluences in a large corpus of spontaneous English conversation. One study will look at how speakers choose the ("thee") over the (thuh) to forecast problems in formulating noun phrases. Another will investigate the contrasting problems that marked by uh and um. A third will examine when and why speakers repeat words as in "and uh there's no point in.in doing that," and a fourth, when and why they stretch words. A fifth will compare monologues and dialogues for number and types of disfluencies. Series 2 will use experiments to examine the disfluencies speakers produce in difficult situations. One experiment will examine the pattern of disfluences that arises as speakers formulate complicated descriptions. A second and third experiment will investigate the temporal details of how speakers manage interruptions caused either by side-tracking or by bursts of noise. A fourth will compare speakers strategies in presenting utterances in face-to-face and telephone conversations. Series 3 will investigate the logic that underlies spontaneous repairs of speech. Precisely how do repairs work when speakers replace one word or phrase with another? Several alternative models will be tested against a large corpus of repairs from spontaneous talk. One study will examine the function of self-talk--when speakers interrupt their utterances apparently to address themselves. Series 4 is based on the idea that many disfluences aren't a nuisance to listeners, as ordinarily assumed, but a help. One study will examine experimentally how addressees exploit the information present in speakers' uses of pauses, uh, um, and other disfluences. A second will examine listeners' interpretations of utterances that contain various disfluencies. Series 5 is about how speakers order phrases based on their weight or complexity. Speakers ordinarily prefer, for example, "I looked up the man I had known for a long time" over I looked the man I had known for a long time up," placing the underlined heavy noun phrase late in the sentence. These studies will study spontaneous uses of many types of constructions for the syntactic, informational, and intonational factors that make phrases heavy. One study will investigate how weight is reflected in disfluencies. Series 6 will examine the process by which speakers select heavy vs. light phrases in real time. One experiment will investigate how speakers order phrases as they recall descriptions, and another, how they do so as they describe visual scenes. A final experiment will examine how quickly listeners follow instructions that contain heavy phrases in various orderings.
