Literature Review In this thesis, I will study the characteristics of the conversational mechanism of repair in Chinese conversational discourse. To that end, it is necessary to conduct a review of the relevant literature on conversational repair. I shall start with an elaboration of the notion of “repair”, going on to researches into the organization of conversational repair and conclude with the interdisciplinary and multi-linguistic application of “repair” research.1. From Correction to RepairAs a relatively new field in conversation analysis (CA), the proper study of the conversational phenomenon of repair didnt start until the publication of Schegloff et als seminal paper in 1977. Before that, there were only some sporadic discussions of the phenomenon under such generic headings as tongue slips (Laver 1973, see Schegloff 1977) and error correction (Jefferson 1975, see Schegloff 1977). As a still often-used term, “correction”, “commonly understood to refer to the replacement of an error or mistake by what is correct” (Schegloff 1977: 363), not only limits research to a minority of the natural occurrences of repair but also misleads researchers about the nature of the trouble-sources. The shift of focus was led by Schegloff et al (1977), whose study was an empirically based effort to examine the organization of repair as a set of ordered, but not equal possibilities. The phenomenon of correction was therefore proven part of a much wider picture, i.e. repair and the scope of discussion was greatly expanded from the mere correcting of some “hearable usually linguistic errors” (1977: 363) to all possible “practices for dealing with problems or troubles in speaking, hearing, and understanding the talk in conversation” (2000: 207), a definition given by Schegloff himself some 20 years later. In deed, potential trouble-sources in conversation include not only correction of information, but also and more importantly replacement of inappropriate items or ambiguous anaphors, word search and clarification of the pragmatic function/understanding of a previous turn. These and many other occurrences may only be subsumed under the more general scope of repair. Incidentally, correction may not always be categorized under repair either, as is exemplified by the disagreement over the so-called “embedded correction” (Jefferson 1987) basically a covert form of other-correction which Schegloff (2000) ruled out as not constituting a kind of repair. Equally important as the expansion in the scope of research was the change in the view of the trouble-sources that directly occasion the repair. According to Schegloff et al (1977), trouble-sources are not self-evident but determined interactively by participants. In other words, all the segments in an utterance is, in theory, potential trouble-sources and often the existence of a trouble-source can only be evidenced by the actual mobilization of the practice of repair on the part of either the hearer or the speaker (and sometimes both). It is worth noting that just as the status of a trouble-source is an uncertainty to be interactively determined, the actual need and proper protocol of its repair is not any more certain. This dynamic and interactive view of repair has proven rewarding in terms of revealing not only its own mechanism but also other cognitive, social and psychological aspects of conversational discourse, as may be interestingly explained by such everyday wisdom: you dont know somethings at work until it goes wrong. 2. The organization of repair Many studies have been carried out with regard to the various dimensions of conversational repair itself, e.g. its classification, sites, forms and causes.Schegloff et al (1977) classified four interactional types of repair according to the subject(s) of initiation/repair, namely self/other-initiated self/other repair. This classification has been adopted by many researchers later, making it easier to tackle conversational data. Yet Geluykens (1994: 56) suggests, rightly I think, that this classification is in need of refinement as it is not always possible to draw a sharp boundary between self and other initiation. He found a sort of other-prompted self-initiation, which underlines the interactive aspect of conversational discourse.Along with the interactional four-type classification, Schegloff et al (1977) proposed the unequal distribution of the four types. To be more exact, self-repair is preferred to other-repair and self-initiation to other-initiation. It follows that the most favored type is self-initiated self-repair. Their claim was put forward with no statistical evidence so later researchers have discussed their empirical findings with reference to either or both of the two preferences. Many studies, including some based on data in languages other than English, are in support of the observation that self-repair is preferred, e.g. Geluykens (1994) and Ma (