Conversation analysis has evolved over several decades as a distinct variant of ethnomethodolo-gy. Its beginnings can be traced to the mid-1960s, to the doctoral research and the unpublished but widely circulated lectures of Harvey Sacks. Sacks was a University of California sociologist who had studied with Harold Garfinkel, the founder of the ethnomethodological movement, as well as with Erving Goffman. While not an ethnomethodologist, Goffman’s proposal that face-to-face interaction could be an analytically independent domain of inquiry certainly helped inspire Sacks’s work. Two other key figures whose writings (separately and together with Sacks) contributed to the emergence of conversation analysis were GailJefferson, one of Sacks’s first students, and Emanuel A. Schegloff, another sociologist trained in the University of California system who was decisively influenced by Garfinkel and, in much the same manner as Sacks, by Goffman (Schegloff 1988).
Sacks, like Garfinkel, was preoccupied with discovering the methods or procedures by which humans coordinate and organize their activities, and thus with the procedures of practical, common-sense reasoning in and through which ”social order” is locally constituted (Garfinkel [1967] 1984). In addressing this problem, he devised a remarkably innovative approach. Working with tapes and transcripts of telephone calls to a suicide prevention center (and with recordings of other, somewhat more mundane sorts of conversations), Sacks began examining the talk as an object in its own right, as a fundamental type of social action, rather than primarily as a resource for documenting other social processes. In short, Sacks came to recognize that the talk itself was the action. It was in the details of the talk that we could discover just how what was getting done in the activity was accomplished, systematically and procedurally, then and there, by the coparticipants themselves. This appeared to be an especially fruitful way of investigating the local production of social order.
As Schegloff (1989, p. 199) later wrote in a memoir of these first years, Sacks’s strategy in his pioneering studies was to first take note of how members of society, in some actual occasion of interaction, achieved some interactional effect— for example, in the suicide center calls, how to exhibit (and have others appreciate) that one has reasonably, accountably, arrived at the finding ”I have no one to turn to”—and then to ask: Was this outcome accomplished methodically? Can we describe it as the product of a method of conduct, such that we can find other enactments of that method that will yield the same outcome, the same recognizable effect? This approach provided, Sacks suggested, an opportunity to develop formal accounts of ”members’ methods” for conducting social life.
In this way, Sacks sought to address the basic question of (as he put it in one of his early manuscripts) ”what it is that sociology can aim to do, and . . . how it can proceed” (Sacks [1964-1968] 1984, p. 21). Sociology, he argued, could be a ”natural observational science,” concerned with the methodic organization of naturally occurring events, rather than with behavior that was manipulated through experimental techniques or other interventions such as surveys, interviews, and the like. And it could be committed to direct observation of this organization in situ, rather than dependent upon analytic theorizing and a concomitant reliance on idealized models of action.
Naturalistic observation also met the ethnomethodological mandate that all evidence for the use of members’ methods, and for members’ orientation to or tacit knowledge of them, was to be derived exclusively from the observed behavior of the coparticipants in an interactional event. As Schegloff and Sacks (1973, p. 290) subsequently summarized the logic of this stance, if the event, the recorded conversational encounter, exhibited a methodically achieved orderliness, it ”did so not only to us [the observing analysts], indeed not in the first place for us, but for the co-participants who had produced” it. After all, the task was to discover members’ methods for coordinating and ordering conversational events, and these could not in any way be determined by analysts’ conceptual stipulations or deduced from inventive theories.
The contrast with other methods and approaches for studying interactional processes could not be sharper, particularly with those methods adopted by Bales (1950) and Homans (1961) and their many followers, with their commitment to theoretically derived and precise operational definitions of social phenomena as a prerequisite to any scientific investigation (see Sacks 1992, vol. 1, p. 28 and p. 105, for his thinking on Bales and Homans). But Sacks’s methodological stance contrasts even with those traditions usually regarded as neighbors of conversation analysis, such as symbolic interactionism and Goffman’s ”micro-Durkheimian” approach. These approaches assume that without an analytically stipulated conceptual scheme, there is no orderliness (or the orderliness cannot be seen) in what Garfinkel (1991) terms ”the ple-num”—the plentitude of members’ lived experience.
Sacks was also making a well-reasoned argument for the importance of studying mundane conversation, directly confronting the belief that sociology’s overriding concern should be the study of ”big issues”; that is, the belief that the search for social order should center on the analysis of large-scale, massive institutions. Social order, he insisted, can be found ”at all points,” and the close study of what from conventional sociology’s point of view seemed like small (and trivial) phenomena—the details of conversation’s organization— might actually give us an enormous understanding of the way humans do things and the kinds of methods they use to order their affairs (Sacks [1964-1968] 1984, p. 24).
This last proposition bears special emphasis, for Sacks felt that these details went unnoticed, and perhaps could not even be imagined, by conventional analytic sociology. When had sociologists concerned themselves with the profoundly methodic character of things like how to avoid giving your name without refusing to give it, he argued, or with how to get help for suicidalness without requesting it? Or with members’ methods for things like ”doing describing” and ”recognizing a description,” methods that provide for hearing the first two sentences from a story told by a young child—”The baby cried. The mommy picked it up.”—as saying: The mommy who picked up the baby is the baby’s mommy, and she picked it up because it was crying. Though apparently mundane, this observation provided the basis for a series of investigations regarding members’ categorization methods and eventually came to provide for an entirely different approach to studies of social institutions and phenomena like race and gender. Sacks noted that since every person can be categorized at any time in various ways (for example, in terms of age, gender, or stage of life), a person’s use of, or reliance on, one category rather than another to guide his or her actions with others must be grounded in one of a multitude of discoverable systems of relevance, some of which are potentially applicable in any situation, such as age, race, and gender, and others that are more limited in their use, such as occupationally defined categories or those made relevant by the organization of conversation itself, such as speaker/hearer, caller/called, and the like.
As these last examples of categories suggest, ”membership analysis”—making sense of who someone is for the purpose of appropriately designing some next action—is an unremitting problem for members of a society, one that has to be solved in real time, and for which there is no single solution. Furthermore, who a person relevantly ”is” for the purposes of some next action can change from moment to moment. As a consequence, Sacks pointed out that sociologists can no longer innocently categorize populations whatever way they (and their theories) see fit. Instead, analysts must similarly demonstrate the relevance of a category to the participants in any scene, as well as its consequentiality for them in terms of how the action proceeds, in order to ground its use in any sociological investigation (see Silverman 1998 for a useful summary of this argument).
Such grounding would stand in contrast to the epidemiological uses of categorization that underpin standard social scientific research. Sacks’s colleague, anthropologist Michael Moerman (1974, pp. 67-68), once noted that social scientists ”have an apparent inability to distinguish between warm . . . human bodies and one kind of identification device which some of those bodies sometimes use.” Further, since Sacks was primarily concerned with categorization as a thoroughly practical, procedural activity for members he was not much interested in the content of categories, with drawing the cultural grids that preoccupied cognitive anthropologists and many social psychologists. Instead, Sacks believed that by starting with the close study of actual events—such as members’ observable use of categories in situ—and showing that they happened in an endogenously, socially organized manner, a much sounder basis for studying and understanding social life could be established.
It should be evident, then, that the appellation conversation analysis does not really capture the enterprise’s commitment to addressing the most basic problem for the social sciences: the underlying character and structure of social action. As the title of one of Sacks’s first publications, ”An Initial Investigation of the Usability of Conversational Data for Doing Sociology,” makes clear, the use of recorded conversational materials was more of an opportunistic research strategy than a commitment to studying talk per se (Sacks 1972). Tape recordings of conversations constituted a record of the details of actual, singular events that could be replayed and studied extensively, and would permit other researchers direct access to exactly these same details. Still, for these identical reasons, it is conversation’s organization—its detectable, orderly properties—that has remained the concrete object of study for the enterprise.
During years since the ”initial investigations,” conversation analysis has given rise to a substantial research literature. Pursuing the lines of analysis first identified in the early studies while simultaneously opening up many new avenues of inquiry, researchers working in this tradition have produced findings that are, in the words of one contemporary practitioner, ”strikingly cumulative and interlocking” (Heritage 1987, p. 256). Important collections of papers include those of Sudnow (1972), Schenkein (1978), Atkinson and Heritage (1984), Button and Lee (1987), and ten Have and
Psathas (1995). Sacks’s lectures have now been edited and published in complete form (Sacks 1992). Special issues of Sociological Inquiry, Social Psychology Quarterly, Human Studies, Social Problems, Research on Language and Social Interaction, Text, and the Western Journal of Speech Communication have also been devoted to ethnomethodological and conversation-analytic topics.
Three major domains in conversation’s organization identified in this literature are the organization of sequences, of turn taking, and of repair. These organizations can be described as systems of naturally organized activity, systems known and used by members as courses of practical action and practical reasoning, and designed to resolve generic problems of coordination that confront any conversationalist (and perhaps members of all social species). A sketch of some research findings with respect to these organizations should serve to illustrate how they function in this fashion, as well as the interlocking nature of their domains.
Consider first the organization of sequences. Begin with the fact that even the most cursory inspection of conversational materials reveals that talk-in-interaction has a serial arrangement to it. For example, in a conversation between two parties, party A will talk first, then party B, then A, then B, and so forth. Accordingly, in two-party conversations, turns at talk constitute a series of alternately produced utterances: ABABAB. But overlaying this serial arrangement of utterances are distinctly characterizable conversational sequences, where turns at talk do not simply happen to occur one after the other but rather ”belong together” as a socio-organizational unit, and where there is thus a methodic relationship between the various turns or parts.
This methodic, structurally linked relationship between sequence parts is central to how sequences work in resolving coordination problems in conversation. This point can be demonstrated by briefly focusing on one of the earliest studies of sequence organization by Schegloff (1968), an investigation into how the initiation of conversational interactions is coordinated. Schegloff directed attention to a frequently occurring initial exchange, which was called a ”summons-answer sequence.” This sequence is composed, he discovered, of closely linked parts. The production of the first turn in the sequence, the summons, projected a relevant next action, an answer, to be accomplished by the recipient of the summons in the very next turn. Moreover, the occurrence of the expected answer cannot properly be the final turn in the exchange. The summons-answer exchange is therefore nonterminal: Upon production of the answer, the summoner is then expected to speak again, to provide the reason for the summons. This provides for a coordinated entry into conversation, and for the possibility of an extended spate of talk.
Observe that a set of mutual obligations is established by the structural relationships between these sequence parts, with each current action projecting some ”next.” In the strongest form of these obligations (sequence classes vary in this regard), the property of ”conditional relevance” holds between the parts of a sequence unit. A ”summons-answer” sequence is but one type of a large class of utterance units, known as ”adjacency pairs,” that are characterized by this property. Examples here include ”greeting-greeting,” ”question-answer,” and ”invitation-acceptance/declination.” In adjacency pairs, when one utterance or action is conditionally relevant on another, the production of the first provides for the occurrence of the second. It could be said, then, using the example above, that the issuance of a summons is an action that selects a particular next action, an answer, for its recipient. If this action does not occur, its nonoccurrence will be a noticeable event. That is to say, it is not only nonoccurring, it is notably, ”officially” absent; accordingly, this would warrant various inferences and actions. For instance, the summoner might infer that a recipient ”didn’t hear me,” which would provide for the relevance and grounds of a repetition of the summons.
The discovery that human activities like conversation were coordinated and organized in a very fundamental way by such methodic relationships between actions, with some current or ”first” action projecting and providing for some appropriate ”second,” led to investigations into the various methods by which the recipient of a first may accomplish a second, or recognizably hold its accomplishment in abeyance until issues relevant to its performance are clarified or resolved, or avoid its accomplishment altogether by undertaking some other activity. Researchers learned, for example, that for some firsts, there was not a single appropriate second but rather a range of alternative seconds. Note that in the examples of adjacency pair structures listed just above, invitations project either an acceptance or a declination as a course of action available to the recipient. In this case, and in others like ”request-granting/denial” and ”compliment-acceptance/rejection,” it was found that the alternative second parts are not generally of equal status; rather, some second parts are preferred and others dispreferred, these properties being distinct from the desires or motivations of the coparticipants. ”Preference” thus refers to a structural rather than dispositional relationship between alternative but nonequiva-lent courses of action. Evidence for this includes distributional data across a wide range of speakers and settings, and, more important, the fact that preferred and dispreferred alternatives are regularly performed in distinctively different ways. The preference status of an action is therefore exhibited in how it is done.
Related to this, conversation analytic researchers observed that the producers of a first action often dealt in systematic, methodic ways with these properties of preference organization. To take one example, the producer of a request can and often does analyze the recipient silence that follows as displaying or implicating a denial—a denial as-yet-unstated, but nevertheless projected— and seeks to preempt the occurrence of this dispreferred action by issuing a subsequent version of the request, before the recipient starts to speak. Subsequent versions attempt to make the request more acceptable and provide another opportunity for a favorable response (Davidson 1984).
Moreover, members were observed to orient to the properties of preference organization through their performance of actions plainly meant to be understood as specifically preliminary to some adjacency pair first action. Such ”pre”-type actions are designed to explore the likelihood that producing that first part of some pair will not be responded to in a dispreferred way. For instance, an utterance like ”Are you doing anything tonight?” provides, in a methodical way, an opportunity for its producer to determine, without yet having to actually issue the invitation, whether it would most likely be declined. Similarly, this provides an opportunity for the recipient of the ”pre” to indicate that a dispreferred action would be forthcoming without ever having to perform that action. Additionally, because ”pre” actions themselves engender sequences by making some response to them a relevant next action, they constitute the first part of a ”pre-sequence.” It follows that since these and other features of preference organization together maximize the likelihood of preferred actions and minimize the likelihood of dispreferred ones, they serve as important structural resources for maintaining social solidarity and ”preserving face.”
These interrelated observations on the organization of sequences were generalized outward in conversation analytic research from the relatively simple adjacency pair organization by the recognition that virtually every utterance occurs at some sequentially relevant, structurally defined place in talk (see especially Atkinson and Heritage 1984, pp. 5-9). Moreover, it is this placement that provides the primary context for an utterance’s intelligibility and understanding. Put another way, utterances are in the first place contextually understood by reference to their placement and participation within sequences of action, and it is therefore sequences of action, rather than single utterances or actions, that have become the primary units of analysis for the conversation-analytic enterprise. Accordingly, researchers in this tradition have not restricted themselves to studying only especially ”tight” sequence units, but have instead broadened their investigations to (mentioning just a few) the sequencing of laughter, disputes, story and joke telling, political oratory, and the initiation and closing of topics. In addition, the sequential organization of gaze and body movement in relation to turns at talk has been the focus of some truly path breaking research using video recordings (see, for example, Goodwin 1981, 1994; Heath 1986).
Now let us consider the organization of turn taking, surely a central feature of virtually all talk-in-interaction. Recall that the prior discussion on the organization of sequences frequently made reference to sequence parts as ”turns,” implicitly trading on the understanding that talk in conversation is produced in and built for turns, with recurring speaker change and a consequent serial ordering of utterances. In conversation, this turn ordering, as well as the size and content of each turn, is not predetermined or allocated in advance. Instead, it is locally determined, moment-by-moment, by the coparticipants in the talk. In fact, this completely local determination of who speaks when, how long they speak, and what they might say or do in their turn, is what provides for talk being hearable as a ”conversation,” rather than as, say, a debate or a ceremony of some kind. But this does not tell us just how—methodically— speaker change is achieved such that, ordinarily, one party talks at a time and there is little or no silence (or ”gap”) between turns. Clearly, this requires close coordination among coparticipants in any conversational encounter. The systematic practices by which this is accomplished are analyzed by Sacks, Schegloff, and Jefferson in a 1974 paper that remains one of the most important in the conversation-analysis literature.
Basic to the accomplishment of turn taking is the practice of changing speakers at possible utterance completion places, what Sacks, Schegloff, and Jefferson term transition relevance places. How are such places, where speaker change may relevantly occur but is in no way guaranteed or required, discernable by members? A key feature of the units by and through which turns are constructed offers one resource here: For an utterance to be usable as a turn constructional unit, it must have a recognizable completion, and that completion must be recognizable prior to its occurrence (Sacks, Schegloff, and Jefferson [1974] 1978, p. 12). That is to say, its completion is projectable, and a coparticipant in the conversation who wishes to speak next can therefore begin his or her turn just at the place where the current speaker projects completion.
Of course, this does not preclude this coparticipant, or any other, from starting to speak elsewhere in the course of a current speaker’s turn. (Indeed, what actually constitutes a ”turn at talk” is as locally and mutually determined as any other aspect of conversation’s organization, even as the resources for doing so are general ones.) There are various interactional moves that could involve, as one way they might be accomplished, this sort of action. At the same time, however, research on turn-taking has revealed that turns beginning elsewhere may well be met with procedures systematically designed to enforce the practice of starting at possible completion places. Further, features of the turn taking system such as that described just above account for a great deal of the overlapping speech that can occasionally be observed. For instance, a speaker might append a tag question like ”you know?” to his or her turn, while a coparticipant, having no resources available to project such an action, starts to speak just prior to or at the beginning of that appended tag, at the place that was projectably the ”first possible completion” of the turn. This would result in overlapping speech, with both parties talking simultaneously. This was just one example; studies of ”more than one party at a time” speech have uncovered massive evidence that its occurrence and its resolution (the restoration of one party at a time), as well as the solution to the problem of which overlapping action should then be consequential for next action, is methodically organized.
Having described the function of turn constructional practices in turn taking, Sacks, Schegloff, and Jefferson still faced the issue of how coparticipants, at possible completion places, determine just who will be the ”next speaker” (note in this regard that conversation can involve more than two parties) or even if there will be a next speaker, given that a current speaker might want to continue talking. They discovered that to deal with this problem, members have available a ”turn allocational component” for the system. This component consists of a set of ordered rules that come into play at transition relevance places and which provide for the methodic allocation of the right to produce a next turn, or more accurately, a turn constructional unit. In related research, methods for securing the temporary suspension of turn-taking procedures (to tell an extended story, for example) and for coordinating exit from the system (to end the conversation) have been documented.
Finally there is the entire set of procedures by which any troubles in speaking, hearing, and understanding talk are systematically handled and ”repaired.” As Schegloff (1979, p. 269) points out, insofar as ”any of the systems and contingencies implicated in the production and reception of talk—articulatory, memory, sequential, syntactic, auditory, ambient noise, etc.—can fail,” any piece of talk is susceptible to, or can reveal, troubles in speaking, hearing, or understanding. As a consequence, members of society must have some systematically organized set of methods for managing such trouble when it arises. Further, in order for interaction to serve as a primary site for the coordination of social activity, any such troubles must be located and dealt with as quickly as possible to avoid whole stretches of talk developing on a problematic basis. Finally, this set of methods must provide the opportunity to discover and display trouble in speaking, hearing, or understanding by any of the ratified coparticipants to the interaction, while simultaneously managing such trouble from the variety of quarters from which it might arise, whether the trouble is noticed or produced by the current speaker or her recipient, and whether its source is endogenous to the interaction or impinges on it from outside.
When Schegloff, Sacks, and Jefferson (1977) began examining the related set of practices through which speakers managed such troubles they discovered two important features. First, they noticed that participants in interaction treat the initiation of repair as a separate matter from the actual accomplishment of a solution. That is, they distinguish between the various practices for locating a trouble source and making it the focus of the interaction and the set of practices for implementing a solution. Second, they observed that these two activities were not distributed evenly among the parties: The organization of repair exhibited a preference for self-repair and a preference for self-initiation of repair. And they went on to show that this latter feature is primarily a product of the way that the organization of repair relies on, and is fitted to, the system for distributing turns.
The organization of repair initiation operates in a restricted ”repair initiation opportunity space” that is organized around the trouble source or ”repairable.” Within this repair initiation opportunity space each party to an interaction moves through a series of discrete opportunities to locate and indicate potential and actual troubles. In turn, these discrete opportunities to initiate repair shape where (relative to the trouble source) a repair is effected, and by whom. The current speaker has the first opportunity to initiate repair on any trouble source within his or her own turn while still in the midst of it, or just after it is complete but before a next speaker starts. If they do initiate repair during (or immediately following the possible completion of) their own turn, such speakers also have the first opportunity to effect repair as well.
Of course, as we noted above, conversation is characterized by the alternation between current and next, thus once a current speaker completes her turn a next speaker begins, typically by addressing herself to that just-prior talk. Accordingly, if a next speaker has some trouble with the prior speaker’s turn, the next turn is the place where she can initiate repair (using a variety of forms, including ”what?” and ”huh?” and other designs that vary in the degree to which they specify the exact source of trouble). By initiating repair using one of these methods, that speaker selects the speaker of the trouble source to speak next, and to offer a solution to the trouble indicated. If the next speaker has no trouble with the prior turn, and she uses it to move the action forward (instead of stopping it to initiate repair), her turn will display a variety of understandings regarding the talk it follows. In doing so, her turn may also reveal some type of misunderstanding (from the point of view of the speaker of the prior turn). If that occurs the speaker of the prior turn can then initiate repair in ”turn after next” (or ”third position”) and offer a solution immediately. Perhaps the recurrent and recognizable format for this is ”I don’t mean x, I mean y.”
Thus, the movement of talk through these three positions—current, next turn, and turn after next—systematically provides the various parties to the interaction the opportunity to detect any trouble in speaking, hearing, and understanding, whatever its source, and initiate repair on it. As a consequence almost all instances of repair are initiated in one of these adjacent locations. The localization of repair initiation opportunities, and the distribution of them over three turns, has several consequences for the organization of social life. First, the localization of repair within a finite, and relatively restricted, space ensures that trouble is dealt with swiftly. Second, and related to this, given the systematic relevance of repair, if speakers move through these three positions without any party initiating repair, a shared understanding of the talk is thereby confirmed en passant.
Finally, as with sequence organization, the issue of preference is best grasped as a structural property of the organization of talk-in-interaction (rather than being a product of concerns regarding the private desires of the parties). The two preferences observed by Schegloff, Sacks, and Jefferson are a product of the distribution of opportunities to initiate and effect repair that systematically favors the speaker of the trouble source over others. As Sacks, Schegloff, and Jefferson ([1974] 1978, p. 40) put it, the organization of turn taking and the organization of repair ”are thus ‘made for each other’ in a double sense.” It is worth noting in this regard that insofar as interaction provides the primary site for the achievement of intersubjectivity, for what makes sociality possible, the organization of repair constitutes its last line of defense (Schegloff 1994).
Taken together, the operation of the turn-taking system and the practices involved in the organization of sequences and repair account for many of the detectable, orderly features of conversation. This orderliness was shown to be locally organized and managed, the product of members’ methods. It will be useful to make note once again of the research strategy that enabled such findings. Because the data consisted of recordings of naturally occurring activity, a scientific account of the phenomenon under investigation could be empirically grounded in the details of actual occurrences. The investigation began with a set of observable outcomes of these occurrences—in the case of turn taking, for example, speaker change overwhelmingly recurred; overwhelmingly, one party talked at a time; turn order, size, and content were not fixed, but varied; and so on. It was then asked: Could these outcomes be described as products of certain social organized practices, of methods of conduct? At the same time, if members of society did in fact use such formal methods, how were they systematically employed to produce just those outcomes, in just those occurrences, in all their specificity? In addressing the problem in this way, then, conversation analysis was able to discover how cardinal forms of social order were locally constituted.
The research on turn taking in conversation has provided one starting point for more recent studies of interaction in ”institutional” settings, such as news interviews, doctor-patient and other clinical consultations, courtrooms, plea bargaining sessions, job interviews, and citizen calls to emergency services. In many of these studies, researchers pursued Sacks, Schegloff, and Jefferson’s ([1974] 1978, pp. 45-47) suggestion that the practices underlying the management of ordinary conversation are the ”base” or primary ones (for an example, see Heritage and Maynard in press).
Other forms of interaction—in this case, so-called ”institutional” forms—are in part constituted and recognizable through systematic variations from conversational turn taking, or through the narrowing and respecification of particular conversational practices involved in the organization of sequences, repair, and other activities.
Take the case of courtroom interaction. The turn-taking system operative in these encounters places restrictions on turn construction and allocation: Coparticipants ordinarily restrict themselves to producing turns that are at least minimally recognizable as ”questions” and ”answers,” and these turn types are pre-allocated to different parties rather than locally determined. The relatively restricted patterns of conduct observable in these settings is, in large part, the product of this form of turn taking. Accordingly, variation in turn taking in such settings has been shown to have a ”pervasive influence both on the range and design of the interactional activities which the different parties routinely undertake and on the detailed management of such encounters” (Heritage 1987, p. 261; Atkinson and Drew 1979).
Note that throughout the above discussion, the term ”institutional” has been presented with quotation marks around it. This was done to emphasize ethnomethodology’s preoccupation with the local production of social order. From this view, that some activity or encounter is recognizably either an ”ordinary conversation” or more ”institutional” in nature—for example, is recognizably a ”cross-examination,” a ”call to the police,” a ”clinical consultation,” or whatever—is something that the coparticipants can and do realize, procedurally, at each and every moment of the encounter. The task for the analyst is to demonstrate how they actually do this; how, for example, they construct their conduct, turn by turn, so as to progressively constitute and thus jointly and collaboratively realize the occasion of their encounter, together with their own social roles in it, as having some distinctively institutional sense (Heritage and Greatbatch 1991). Conversation analytic research on ”institutional” interaction has therefore undertaken, through its investigations into the methodic practices by which this gets done, a systematic study of a wide range of human activities.
This mode of research, with its commitment to understanding precisely how any activity becomes what it recognizably and accountably is— that is to say, how it acquires its social facticity-has tended to focus in the 1990s on work activities and settings, under the rubric of ”workplace studies.” The scope of investigation has expanded to encompass all forms of ”embodied action” (that is, not only the talk), with extensive use of video recordings and, influenced by Suchman’s (1987) pioneering study of human-machine interaction, with careful attention to how the machines, technologies, and other artifacts that saturate the modern work site are taken up and enter into the endogenous organization of work tasks (see, for example, Whalen (1995), and the papers collected in Luff, Hindmarsh, and Heath in press).
Research into conversation’s organization also continues to evolve. While there has been relatively little work that attempts to fundamentally deepen the original account of the turn-taking system developed by Sacks, Schegloff, and Jefferson (for a notable exception see Lerner 1996), there has been important research at the intersection of grammar and interaction—recognizing that talk-in-interaction is in fact the natural home of human language. This work demonstrates that the approach to language taken Chomsky’s Transformational Grammar could be supplanted by one based on naturalistic study of the ”grammar for conversation.” Given that Chomsky’s approach has decisively shaped, both directly and indirectly, the understanding of language in cognitive science, psychology, computational linguistics, and the other disciplines that rely on a model of grammatical organization for their own research, these findings are plainly significant.
Conversation analytic work on graamar and interaction was launched by Schegloff s (1979) paper on the ”relevance of repair for a syntax for conversation.” This line of work has underscored the need for studies of language to draw on naturally occurring spates of talk. As Schegloff observed, while nearly every episode of ordinary talk contains instances of repair within the ”sentences” (or sentential turn constructional units) out of which it is built, the entire view of language developed by linguists is based on imagined (or what might as well be imagined) instances of language that are free of such repair. Schegloff went on to show that most instances of talk-in-interaction, at least in English, are organized by reference to the systematic relevance of repair, whether an instance of it actually occurs in the sentence or not.
Of course repair is not the only organization relevant for grammar, and so more recently scholars have begun to examine what more might be learned about language by studying it as produced in naturally occurring interaction. With respect to this problem, conversation analysts have argued that insofar as language most likely evolved in face-to-face encounters by members of our species, its structure and organization must have evolved, at least in part, to manage the basic exigencies confronted by speakers and hearers. Thus, in addition to the systematic relevance of repair, the structure and organization of grammar most likely evolved as resources that shape, and are shaped by, how opportunities to speak are distributed, what constraints are introduced by a current turn on subsequent ones, and how speakers’ formulations of the events, persons, and objects are organized. Perhaps most developed are a series of findings that link the organization of grammar and the system for distributing turns at talk briefly described above.
As we stated earlier regarding turn-constructional units, one of their key features is that each sentence, or utterance, projects from its beginning roughly what it will take for it to be possibly complete. And over its course, each utterance projects in finer and finer detail the exact moment that a speaker may end her utterance. Thus, instead of expressing logical predicates or cognitive states, grammar may be best understood, in the first instance, as a sequentially sensitive resource that progressively projects the course and duration of turns at talk (Ford and Thompson 1996).
One of the most striking consequences of such a view of grammar is that the locus of its organization is transformed. While most approaches to grammar rely on the sentence as the basic unit of organization (with occasional nods to the organization of ”discourse”), the grammatical units produced in interaction are fundamentally organized relative to their sequential environment, most proximally the just prior, current, and next turns. Thus, rather than the sentence, or even discourse, being the fundamental unit or environment of analysis, interaction and sequences of turns appear to be that within which grammar is most proximally organized. This appears to be true even at levels beneath the turn whether a sentence, clause, or phrase. As Schegloff (1996) shows, turn beginnings and turn endings, as well as what happens in between, are sites of strategic manipulation. Through this manipulation, both grammatical and prosodic, speakers fit their utterances to prior talk, launch new actions, and shape when they will be heard as possibly complete. Any scientific analysis of language, then, must take into account this central function.
Thus, as the collection of papers assembled in Ochs, Schegloff, and Thompson (1996) suggest, rather than viewing grammar as an independent, clearly delineated, and internally coherent structure, it is best approached as one more of the interrelated set of resources through which interaction, and social life more broadly, is organized.
