Few ideas have figured more centrally in the history of social theory than that of the division of labor (for a recent overview from an anthropological perspective, see Sanchez, 2018). In his early writings, Karl Marx theorized the division of labor in relation to processes of alienation. Specifically, workers, performing specialized, repetitive tasks that figured as isolated steps in a larger productive process, were alienated not only from that which they had a hand in producing but also from the creative activity of production itself and, ultimately, their own “species being” (Marx, 1977[1844]). The later Marx (1992[1867]) emphasized a distinction between the social division of labor and the division of labor in production (Mohun, 1983). The social division of labor consists of the various ways in which labor is distributed within a society between, for instance, men and women, young and old, peasants and feudal landlords, proletariat and bourgeoisie, and so on. The division of labor in production, on the other hand, refers to the ways in which processes of production under capitalism are broken down into component operations, the prototypical example being the assembly line characteristic of factory labor. While ever more minute divisions lead to greater efficiency and increased production, for Marx and Engels this simultaneously encourages the development of social classes whose interests are fundamentally at odds. Moreover, the very conditions of labor (trade and aggregation in towns and cities under feudalism; factories and union organization in the case of industrial capitalism) lead to the development of class consciousness and, eventually, a revolutionary political movement.

In contrast to the critical perspective of Marx and Engels, Durkheim (1933[1893]) emphasized the integrative fuction of the division of labor as the primary mechanism of organic solidarity. Just as the organs of the body have specialized functions, each essential to the welfare of the whole, so too the various groupings within a modern, industrialized society make a distinct and necessary contribution to the larger collective. Buoyed by a shared set of norms, values and beliefs, the organic solidarity which emerges from the division of labor prevents the destructive forces of entropy from taking root. Despite their differences, both Marx and Durkheim believed that all societies, past and present, exhibit some kind of division of labor. In what they saw as the most primitive forms of social arrangement, this was organized along lines defined by age and sex/gender. In the 1970s, a number of Marxist and feminist anthropologists noted the apparently universal association of women with the domestic domain and with the work of social reproduction (see, inter alia, Ortner, 1972; Rosaldo, 1974; Godelier, 1986). They also noted an apparently near-universal denigration of this domain in relation to “public spheres that are ostensibly sites of collective dynamism” (Sanchez, 2018).

In our contribution to this special issue of Frontiers in Sociology, we ask whether conversational interaction exhibits, like other forms of social activity, a division of labor and, if so, according to what principles it is organized and what functions it serves.¹ This initial statement of our aim requires some explanation and qualification. We may begin, then, by noting that intersubjectivity—shared understanding—requires effort. To put this another way, a conversation's participants actively work to achieve and sustain understanding, despite appearances that this emerges spontaneously in the turn-by-turn unfolding of talk.² Much of the time, this work is invisible and only its products are displayed in the form of sequentially fitted and appropriate next turns at talk. However, in sequences of repair, and especially in sequences of other-initiated repair, the work involved in the maintenance of intersubjectivity rises to the surface. On these occasions, we can see and thus describe what participants do to achieve and sustain what they take to be adequate understanding. As such what we will describe here is not so much an “interactional” division or labor, as an intersubjective one. Our claim is that the work of maintaining mutual understanding is unevenly distributed across a conversation's participants, at least in our data.

This study responds, then, to a typically unarticulated assumption of scholarship in conversation analysis and related approaches: the idea that the work required to sustain intersubjectivity is evenly distributed among the participants, each having essentially equivalent responsibility to ensure that they are understood and that they understand others. This conceptualization fits with a pervasive egalitarian ideology that characterizes many of the settings in which talk takes place. However, there are social situations in which assumptions of egalitarianism do not hold. In our data, which consist of video recordings of casual conversations between Vietnamese same-generation peers, participants continuously display an orientation to relations of relative seniority through the selection of terms used to accomplish interlocutor reference (i.e., reference to the speaker and addressee of an utterance, see e.g., Luong, 1990; Sidnell and Shohet, 2013; Sidnell, 2019, 2022; Djenar and Sidnell, 2022). This pervasive orientation is also reflected in practices of repair initiation. Specifically, seniors regularly initiate repair with so-called “open class” forms such as “huh?” and “ha?” which display a minimal grasp of the talk targeted, require little effort to produce and, at the same time, push responsibility for resolving the problem onto the trouble source speaker, i.e., the junior member of the dyad (on trouble responsibility, see Robinson, 2006).³ In contrast, juniors often initiate repair of a senior participant's talk by displaying a detailed understanding of what has been said, in the form of a repeat, and inviting the senior to confirm. Not only do these practices of initiation ask little of the senior participant in terms of response and, as such, have an “assistive” feel to them, they also often mark what has been said as important, as worthy of repetition, as something that others should clearly understand and so on. We suggest then that this asymmetry in the distribution of initiation practices reflects a “division of intersubjective labor.”

In her 1978 article on differences in the contributions of men and women to everyday interaction, Fishman (1978, see also 1977) concluded:

Fishman's findings were, however, largely impressionistic and the analysis was based on an, at the time, common assumption that the functional value of a conversational “act” or “action” is the same across different sequential contexts. For instance, Fishman quantified the number of questions asked by the male and female participants in 7 h of interaction in a domestic setting. She similarly compared “minimal responses” and “statements” which “display an assumption on the part of the speaker” that they will be understood and of interest, and elicit response from their recipients. Subsequent attempts to replicate Fishman's findings failed (see McMullen et al., 1995) suggesting that, while the initial intuition of a division of interactional labor may be valid, particularly in the setting that Fishman studied, the analytic categories she employed were not sufficiently well-defined to adequately capture it.

Research on the organization of interaction done since the 1970s allows for a refinement and rethinking of Fishman's study (see, inter alia, Heritage, 1984; Moerman and Sacks, 1988; Sidnell, 2014). Specifically, we know that the maintenance of shared understanding or intersubjectivity requires effort. Much of the time, the work that participants do to achieve such understanding is invisible to analysts and only its products in the form of appropriately responsive next utterances are available to us. However, when they encounter troubles of understanding, conversationalists routinely employ practices of repair in their attempts to resolve them. This makes the work of maintaining intersubjectivity available for analytic inspection.

In what follows, we explore this work in a study of Vietnamese conversation. More specifically, we examine various practices of repair initiation and track their distribution across senior and junior interlocutors. This is made possible by the fact that Vietnamese conversationalists are pervasively oriented to locally relevant relations of seniority. Their in-situ orientation to such relations is displayed, most prominently, in the terms they use for interlocutor reference, that is, reference to speaker and addressee.

Our analysis challenges a basic assumption of work in conversation analysis—that participants in a conversation bear essentially equivalent responsibilities for the work involved in maintaining shared understanding. That assumption may be warranted in many of the settings that conversation analysts have studied—such as interaction among English speaking peers in informal conversation—but does not accurately reflect the socio-cultural realities within which Vietnamese conversation takes place. In this latter setting, relations of seniority and the different expectations in terms of interactional conduct to which they are indelibly linked, shape conversational organization in a range of significant ways.

The results of our study, and the intellectual motivation that animates it, resemble those of Ochs (1982, 1984) who, in research conducted in the early 1980s, compared what she called clarification strategies in White Middle Class American (WMC) and Samoan households. Ochs drew on work by Schegloff and other conversation analysts which seemed to show that (1) repair initiation practices exhibit a “natural ordering” based on their relative power to locate a repairable (Schegloff et al., 1977, p. 369) and (2) “speakers show a preference for using the strongest form they can in initiating repair of another's utterance” (Ochs, 1984, p. 331). Ochs found that in the Samoan context, practices of repair initiation (or what she calls “clarification”) are differentially employed depending on the relative rank of the participants:

In the Samoan context, high-ranking conversationalists typically request clarification using a minimal grasp strategy (i.e., open class repair initiators) rather than an expressed guess, as the latter requires one to more obviously take alter's perspective. In what follows we will show that, in Vietnamese conversation, we find a similar pattern in which seniors tend to initiate repair with open class initiators which (1) do not require that the speaker attempt to recover what the other has said, (2) suggest that responsibility for the encountered trouble lies with the trouble source speaker (i.e., the more junior interlocutor), and (3), require little articulatory effort for their production (this itself serving as a sign of the senior participant's low level of involvement in the junior participant's talk). At the same time, we find that junior participants rarely employ such open class repair initiators. Juniors instead show a marked tendency to use a practice of repair initiation that involves repeating a more senior participant's talk with an appended question particle. Even more striking, we find that junior interlocutors engage in an apparently distinctive sequence that involves asking a senior participant a question, receiving an answer and then requesting confirmation of that answer with a repeat appended by a question particle. This practice seems to illustrate the more general tendency of juniors to carefully reconstruct and publicly check their understanding of a senior participant's talk.

The data used in this study come from a larger investigation of other initiated repair and intersubjectivity in Vietnamese conversation. The corpus, collected in various coffee shops and restaurants in Hanoi in 2012, consists of approximately 35 hours of video recorded conversation among same-generation peers. For the present study we sampled five of these recordings. We summarize their basic features in Table 1. All instances of other initiated repair were collected from a portion of each recording (VNR 05 and VNR 20/21 = +/−50 min, VNR 10, VNR 12, and VNR 32 = +/−10 min). The result was a collection of 96 instances. The authors of the current report relistened to all these cases and discussed them in some detail. As we did this, we also sorted the examples into sub-collections according to the format used in the initiation of repair (see the next section for an overview). Once all the cases had been sorted, they were retranscribed and checked again, a process that resulted in additional observations.

Episodes of repair are composed of parts. A repair initiation marks a disjunction with the immediately preceding talk while the repair itself constitutes an attempted solution to a problem. That problem, the particular segment of talk to which the repair is addressed, is called the trouble source.⁴ Our discussion in what follows focuses on the alternative formats used in the other-initiation of repair and some of the sequential consequences that flow from the selection of one format or another. In their classic paper on the preference for self-correction, Schegloff et al. (1977, p. 367–368) distinguished five common repair initiation formats in English conversation: (1) interjections and questions words such as huh? and what?⁵; (2) question words such as who, where, when; (3) partial repeats of the trouble-source turn, plus a question word; (4) partial repeats of the trouble-source turn; and, (5) candidate understandings of a prior turn. In an important recent study, Dingemanse et al. (2014, p. 5) find that different languages make available “a wide but remarkably similar range of linguistic resources” for the other initiation of repair. According to these authors, alternative formats can be differentiated along a number of dimensions including the extent to which they characterize the trouble, the way they manage responsibility for the trouble, and what they imply about the relative distribution of knowledge among the co-participants.

Drawing on the distinctions introduced by Schegloff et al. (1977) as well as some terminological and analytic refinements introduced by Dingemanse et al. and others (e.g., Drew, 1997), we were able to sort the Vietnamese cases into five categories as shown in Table 2.⁶

Open class and repeat-formatted initiations will be examined in some detail in the discussion that follows. Here we give examples of the other formats for illustrative purposes.

In (1) the participants have been complaining about a rise in the cost of vegetables and about the dismissive attitude of those who sell them in the street markets. In line 79, Phuong, remarks that dill leaf has risen in price to 10,000 dong a bunch.⁷ Thanh, apparently unsure of what Phuong is referring to, initiates repair with cái gì ‘what' thereby targeting a noun phrase in Phuong's previous utterance that refers to something other than a person or a place (see footnote 5). Phuong provides a repair solution in the form of a repeat of the noun phrase (thì là).

Below, (2) illustrates the use of a closed class question word appended to a partial repeat of prior talk. Hoàn and Ba, along with Kiên, jointly own and run a computer software and programming company. Where the extract begins, Hoàn is asking about one of several ongoing projects referring to this as, in line 07, dự án ba ‘project three'. Ba responds with an open class repair initiator which merely indicates a problem with the immediately preceeding turn but does not specify a particular component or aspect of this as the trouble source. Hoàn continues in line 09 apparently assuming that the problem will resolve itself as the talk progresses, but Ba initiates repair again now using a combination of repeat (dự án ‘project') and question word (nào ‘which'). Hoàn then provides a repair in line 11 saying, Dự án đấy. Bank đấy ‘That project. That bank'.

Finally, in (3) we see the use of a candidate understanding to initiate repair. This is taken from the same recording as example (2). Here Hoàn, Ba and Kiên are discussing how much of the company money is being spent on their various projects. The extract begins with Kiên saying that Ba has recently withdrawn 100 million dong from a company account. Ba initiates repair in line 12 and Kiên repeats in part what he has just said. Ba then responds, suggesting that Kiên has misunderstood, and that he's talking about something else, leading Hoàn to initiate repair with a form which invites Ba to confirm a proposed candidate understanding of his talk. In the first of these candidates, Hoàn proposes Khoản đấy bỏ ra à ‘You excluded that amount' and in the second, at line 18-19, he suggests Nghĩa là bên kia nó nó đấu tư về à ‘Meaning the other part is what they invested?'

With respect to the initiation formats illustrated by examples (1), (2), and (3), there were no clear distributional differences according to the relative seniority of the participants.⁹

Our analysis focuses on the relation between the practices of other-initiated repair (and in particular on the use of alternative formats for initiation) and the relative seniority of the participants. Initial review of the recordings, along with native-speaker intuition, suggested that interjections (such as huh? and ha?) were used only when a senior participant initiates repair of a junior participant's talk. In addition, a slightly more sustained examination of the recordings seemed to indicate that repeats were more often used, and used in a particular way, by junior participants to initiate repair of a senior participant's talk.

In order to develop an analysis that might provide empirical grounding for these observations, we needed to operationalize a notion of “seniority.” This is an aspect of social organization toward which Vietnamese conversationalists are pervasively oriented since in almost any context a speaker must take such relations into account in designing a situationally appropriate utterance. This is seen most obviously in the terms used for interlocutor reference. As is well-established in the sociolinguistic and linguistic anthropological literature, the default means for accomplishing interlocutor reference in Vietnamese across a very wide range of contexts involves the use, not of pronouns, but rather of kin terms.¹⁰

Moreover, in Vietnamese there are no reciprocally used kin terms and, as such, interlocutor reference by such means results in a continuous display of relative seniority.¹¹ For instance, a speaker may self-refer using a term such as anh ‘elder brother' or chị ‘elder sister' while referring to the addressee as em ‘younger sibling'. These relations of seniority cannot be read directly from the ages of participants for several reasons, some of which are important to the analysis of repair initiation that follows. First, if two persons are born in the same calendar year, they may consider themselves true peers and avoid the use of sibling terms that necessarily convey relative seniority. Second, in some contexts and in some social relations, relative seniority is exaggerated whereas in others it is understated. Specifically, a difference of 5 years may be treated as significant in one dyad but not in another.¹² For these reasons, in order to operationalize seniority, we can't simply correlate some particular aspect of the speech behavior with the relative ages of the participants. Rather, we have to look at the ways in which the participants themselves orient to such relations, for instance in their practices of interlocutor reference, and use these orientations as a guide to understanding other aspects of their conduct.¹³

In a tour de force exploration of discursive practice and linguistic meaning in Vietnamese, Hy van Luong suggested that the pragmatic significance of person referring expressions (including kin terms, titles, names and pronouns) can only be understood in relation to competing models of and for reality (Luong, 1984, 1990, the notion of a model “of and for reality” is adapted from Geertz, 1973: 93). The pragmatic significance of kin-terms, in particular, is construable in relation to either of two contradictory models. Luong (1990, p. 50) explains:

Construed in terms of the male-oriented model, họ ‘last name, family name, family' refers to a “locally based patrilineage.” Construed in terms of the non-male-oriented model this same term refers to a “bilateral kindred.” Luong further suggests that these alternative kinship-relational models “conjoin at one level and contradict each other at another.” As he writes:

At this level, then, the “organic unity framework” contrasts with another possibility, which Luong refers to, drawing on the work of Turner (1974), as a communitas alternative. Thus, construed in relation to the organic unity framework, in either its male-oriented or non-male oriented guises, the use of pronouns tao and mày, for instance, suggests the absence (or suspension) of a relation based on kinship or any other positively valenced social relation and thus, by implication, contempt or denigration. Construed in relation to the communitas alternative, however, these same forms convey solidarity, extreme familiarity and even intimacy (see Zuckerman, 2023 for a similar case from Laos).

In this way, Luong recasts Durkheim's notion of organic solidarity (based on notions of differentiation and specialization within a larger whole, here a family) as a semiotically mediating ideological orientation rather that as the inevitable product of the division of labor characteristic of a particular social formation. We propose that the materials considered above fit well with this conceptualization. Specifically, in the patterns of other initiated repair here documented, we see a pervasive concern among conversationalists with relations of relative seniority and with the duties and entitlements that normatively attach to positions within an asymmetrically organized social arrangement.

Does this suggest a system of exploitation similar to that which Marx found in the division of labor associated with capitalism and which Fishman proposed could also be identified in cross-sex interactions among white middle-class Americans in the 1970s? Two features of the present case speak against this. First, relations of seniority lack the stability of class or gender relations within a particular encounter. For instance, a participant may be positioned as junior relative to one co-participant and as senior relative to another. If the analysis proposed here is correct, such a participant will be obligated to support the maintenance of intersubjectivity at one moment and entitled to expect such support from another at the next. Second, the relations of seniority which organize the intersubjective division of labor lack the stability of class or gender relations across the life course. Any given individual will find themselves gradually occupying the senior role across more and more interactional encounters as they age. For these reasons, the division of labor we have identified here seems not to be a system of exploitation per se, but rather an asymmetrical organization of duties and entitlements.

To conclude, our study suggests that, in Vietnamese conversation, participants are oriented to a normative division of labor which demands junior interlocutors expend more effort than senior ones in the maintenance of intersubjectivity. Specifically, whereas senior interlocutors regularly initiate repair with a form that pushes responsibility for the problem onto another participant, junior interlocutors more often initiate repair in ways that display close attention to, and detailed understanding of, a senior interlocutor's talk. In terms of larger theoretical questions, our study points to some of the complex ways in which the “social” bears on the “interactional.” We note that much research in CA that attempts to address the question of when and how perduring social facts bear on the organization of interaction focuses on participants' invocation of these facts (whether explicitly or implicitly). This approach appears to assume that the social order is brought to bear on interaction when the vernacular categories of everyday or institutional life (such as, e.g., “men” and “women,” “queer” and “straight,” “old” and “young” etc. for English) are imported into it. We have come at the problem from a different direction, and this has revealed a quite different way in which the social bears on the interactional. Specifically, beginning with participants' displayed orientations to seniority (displayed, that is, in their selection of terms for interlocutor reference), we discovered a robust correlation with the practices involved in the other-initiation of repair. We have proposed that this reflects an unequal distribution of the work involved in the maintenance of intersubjectivity. Notice then that the perduring facts of age are integrated, lockstep, with the organization of interaction. Age is not being “invoked” by the participants as relevant to the organization of interaction. Rather, the practices of interaction are, in part, organized by reference to it. But note also that the social facts (of age) which are built into these sequences are not entirely isomorphic with the vernacular categories of explicit reflection and ratiocination. Rather, “age” is integrated as a wholly indexical variable (“indexical” in the sense of Garfinkel, Sacks, and Schegloff), always calculated in relation to the age of those others with whom a given participant finds themselves interacting. The social is not, as it were, plucked from the sky and made to serve interactional ends. Rather, the social is woven into the warp and weft of interaction as it unfolds moment-by-moment, turn-by-turn.

The datasets presented in this article are not readily available because of restricted access. Requests to access the datasets should be directed to jack.sidnell@utoronto.ca.

The studies involving humans were approved by Research Ethics Review Board, University of Toronto. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.

All authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication.