REPETITION

Repetition is a basic operation that speakers of any language do in and on talk. Studied for centuries as a central ingredient of many rhetorical figures in argumentative, narrative, and poetic discourse, repetition is now understood as a pervasive feature of any form of language usage, and in fact of social and cultural life more generally (Brown Reference Brown1999). Repetition plays a fundamental role in language usage from the early stages of development, supporting the transmission and acquisition of words, sounds, and syntactic structures (Keenan Reference Keenan, Ervin-Tripp and Mitchell-Kernan1977; Bennett-Kastor Reference Bennett-Kastor1994; Tarplee Reference Tarplee, Couper-Kuhlen and Selting1996; Brown Reference Brown1998). Among adults, repetition is a resource for building cohesion into the fabric of discourse (Halliday & Hasan Reference Halliday and Hasan1976), for creating interpersonal involvement and affiliation (Tannen Reference Tannen1989), and for accomplishing many other functions ranging from agreeing to contradicting, from sharing humour to mocking (Norrick Reference Norrick1987), making it a central building block in the architecture of intersubjectivity (Sidnell Reference Sidnell, Enfield, Kockelman and Sidnell2014).

Being such an important resource in language usage and social interaction and having been researched in various fields of linguistic and social science, the phenomenon of repetition can be approached in different ways. As such, we need to specify what kind of repetitions we are concerned with in what follows. First, our focus is on verbal repetition, that is, on the resaying of words, phrases, or clauses, to be distinguished from prosodic repetition that reproduces the suprasegmental features of previous speech (Couper-Kuhlen Reference Couper-Kuhlen, Couper-Kuhlen and Selting1996:367). Second, we intend repetition as lexical replication, that is, as the resaying of the ‘same’ word, phrase, or clause. This specification is necessary as repetition may be conceptualised on a cline extending from exact replication at one extreme to paraphrase at the other. A large strand of the literature adopts a broad definition of repetition that includes the resaying of something with substantial rewording; the interest of this research is typically in the wider phenomenon of parallelism, echoing, and resonance (Tannen Reference Tannen1989; Johnstone Reference Johnstone1994; Du Bois Reference Du Bois2014). In the present project, we adopt a strict definition, by which repetition is intended as a ‘near copy’ of the original saying (Couper-Kuhlen Reference Couper-Kuhlen, Couper-Kuhlen and Selting1996:368), involving the replication of its lexical material with only limited modification, such as deictic shifts (Schegloff Reference Schegloff1996, Reference Schegloff1997; Brown Reference Brown1998). Third, repetition can be divided into two classes depending on whether it is done by the same speaker of the first saying—self-repetition—or by someone other than the speaker of the first saying; our focus here is on the latter class: other-repetition. Lastly, our object of study needs to be delimited in terms of sequential action. As already laid out by Schegloff (Reference Schegloff1996:177), other-repetitions can occupy at least three different types of sequential position, or better, enact three different types of sequential functions: initiating, responding, and third-position functions. Other-repetition may be used to initiate a sequence, such as a repair sequence aimed at clarifying a reference to a person (e.g. A: my cousin studied classics B: your cousin? A: Mary).Footnote ¹ Or it may be used responsively, for instance, to agree with an assessment (e.g. A: I remember she had a beautiful smile B: a beautiful smile [nodding]). Other-repetition may also enact a third-position action that serves to acknowledge receipt of a response, such as the answer to a question (e.g. A: were you classmates only in high school or? B: only in high school A: only in high school [nodding]). In the project reported on in this special issue, our focus is primarily on repetitions that initiate a sequence. This means, in the first place, other-initiations of repair and related actions. But we also consider repetitions embodying third-position actions as these, too, can occasion a response (see Schegloff Reference Schegloff1997:527).

In summary, our object of study in this project is verbal repetition, intended as exact or modified replication of lexical material, produced by someone other than the speaker of the first saying, so as to problematise or otherwise engage with what has been said making a response conditionally or optionally relevant (more details in later sections).

These other-repetitions have been the topic of much research, in some of which they have been referred to as ‘echo utterances’ or ‘echoes’, including ‘echo questions’ and ‘echo exclamations’ (Bolinger Reference Bolinger1957; Quirk, Greenbaum, Leech, & Svartvik Reference Quirk, Greenbaum, Leech and Svartvik1972:408–11). One of the reasons why such other-repetitions have fascinated generations of linguists and social scientists is that they can accomplish very different interactional work, ranging from seeking clarification to expressing surprise to questioning the acceptability of what is said, spanning many of the fundamental preoccupations of human interaction: mutual understanding, alignment, agreement, and affiliation. To distinguish among all of these functions, the lexical content of the repetition is generally of little help to the interlocutor, because it is after all what they have just said themselves. The replication of words per se does little more than tie back to the original saying. So, in other-repetition sequences, speakers and recipients need to resort to other cues, such as prosody and visible behaviour, in order to determine what the repetition is doing and how to respond to it.

But other-repetitions are not only a linguistic and interactional phenomenon in their own right; they also provide for a sequentially and formally delimited environment that is an especially good locus for cross-linguistic comparison. Repeating what another has just said is an operation that is available in any language, as it largely does not depend on language-specific forms. At the same time, other-repetition is never mere reduplication of what has been said: it always requires the speaker to modulate or frame the resaying in some fashion (Sidnell Reference Sidnell2009:5–7). As such, other-repetition involves the mobilisation of the local resources of a language, including different grammatical, prosodic, and gestural systems, which have consequences for how the actions accomplished by other-repetitions are marked and distinguished.

In the next two sections, I first elaborate on the topic of cross-linguistic comparison with a presentation of the field of pragmatic typology and then focus in on the study and comparison of prosody.

PRAGMATIC TYPOLOGY

The field of pragmatic typology combines two areas of linguistic science: pragmatics—the study of language in context and of the rules that govern its use—and typology—the comparative study of linguistic systems (phonological, morphological, syntactic, semantic) to determine how and to what extent their organisation is similar or different across languages. Pragmatic typology unites the goals and tools of pragmatics and typology to investigate patterns of language use and conversation across different communities, in order to establish similarities grounded in the common infrastructure of human interaction as well as differences driven by the local resources of particular linguistic and cultural settings (see also Sidnell Reference Sidnell2009; Dingemanse & Enfield Reference Dingemanse and Enfield2015, among others).

The roots of this growing field go a long way back and can be traced to early efforts of linguists, anthropologists, and sociologists (Dingemanse & Floyd Reference Dingemanse, Enfield, Floyd, Enfield, Kockelman and Sidnell2014). One of the most important landmarks in this field is the work done by Brown & Levinson (Reference Brown and Levinson1987) as part of their theory of politeness. The development of the theory was spurred by the observation of strong similarities in the design of speech acts in three unrelated languages and cultures (Tamil, Tzeltal, and English). This suggested the existence of common universal pressures affecting people's linguistic and social behaviour, which the theory of politeness was designed to explain. The theory partially influenced the development of a prolific area of research commonly known as ‘cross-cultural pragmatics’, the specific aim of which was to establish similarities and differences in the realisation of speech acts (e.g. requests, apologies) across languages and cultures, by following a common set of procedures based on written elicitation in the form of a ‘discourse completion task’ (Blum-Kulka, House, & Kasper Reference Blum-Kulka, House and Kasper1989).

The comparative study of language use has greatly developed since the inception of politeness theory and the booming days of cross-cultural pragmatics. In a nutshell, the latest research in this area can be characterised by two main elements. The first is a commitment to the empirical analysis of recordings of natural conversation—the central arena of language use—increasingly informed by the framework of conversation analysis. This creates a substantial overlap between pragmatic typology and what can be referred to as comparative conversation analysis (see Sidnell Reference Sidnell2009). The second element is a methodology for making comparison systematic. This normally requires us to identify a formally and sequentially delimited environment, establish clear parameters for linguistic and interactional variation, and it increasingly involves coding schemes that both aid in the formulation of comparative aims and enable quantitative analysis of linguistic and interactional features.

A prerequisite for research in any area of linguistic typology is the ability to compare like with like (Greenberg Reference Greenberg1966; Croft Reference Croft2003; Dingemanse & Enfield Reference Dingemanse and Enfield2015, among others). This means being able to draw the contours of the target phenomenon so that the categories under comparison and the environment in which they operate are commensurable. Just as in the study of the sound inventories that different languages have for constructing words (phonological typology) or in the study of the vocabularies they have for referring to parts of the body (semantic typology), in pragmatic typology as well, we must be able to properly define the environment in which a pragmatic phenomenon occurs and the categories of talk and other conduct that operate in it. To take the definition of the environment first, in the framework of conversation analysis this is achieved not by constraining the actions of speakers by means of experimental design, but by identifying a conversational structure, a formally and sequentially delimited exchange of actions that is recurrent and stable across topics, speakers, settings, languages, and cultures (see Stivers et al. Reference Stivers, Enfield, Brown, Englert, Hayashi, Heinemann, Hoymann, Rossano, de Ruiter, Yoon and Levinson2009; Dingemanse & Floyd Reference Dingemanse, Enfield, Floyd, Enfield, Kockelman and Sidnell2014). As discussed in the previous section, other-repetition sequences provide just such an environment. As for the categories under comparison, these will concern both the linguistic practices operating within the environment and the distinct interactional patterns they give rise to. The articulation of such categories, too, must be sensitive to cross-linguistic applicability. To achieve this, typologists make an essential distinction between language-particular descriptive categories, which are based on the internal organisation of a given language, and language-independent comparative categories, or better ‘comparative concepts’ (Haspelmath Reference Haspelmath2010), which are designed to capture generic patterns that cut across language-particular categories. This distinction allows us to juxtapose linguistic objects that are not equivalent when considered in their respective systems but that are sufficiently similar in a relevant respect. As a linguistic object, other-repetition is relatively straightforward to capture cross-linguistically as it is largely independent of language-particular morphosyntax and lexicon.Footnote ² However, some of the formal parameters along which other-repetitions can be analysed do require us to use comparative concepts. For example, to analyse the grammatical function of a repeated word, we cannot use lexical categories such as adjective or adverb, because the criteria for assigning words to these categories are not the same across languages. However, we can use the concept of modification (as opposed to reference and predication) to capture words whose main function is to specify a property or quality of an entity or the manner of an action (see Croft Reference Croft2003:183–88). As another example, particles such as English oh, Finnish ai, and French ah are not equivalent in the range of meanings and uses they can have. When considered as prefaces to an other-repetition, however, these particles can all occupy the same turn-initial position to signal a ‘change of state’ of the speaker (Heritage Reference Heritage, Atkinson and Heritage1984). So, in this structural position, these particles are sufficiently similar for comparison. Finally, comparative concepts are necessary also in the domain of prosody. As is well known, the prosody of languages varies greatly, one area of particular variation being the intonation system (Hirst & Di Cristo Reference Hirst and Cristo1998). As an example, take a rise-fall contour, which is available in both French and Italian. Even if the contour is melodically similar in the two languages, its typical function in other-repetition sequences could not be more different: challenging in French, seeking confirmation in Italian (Persson, this issue; Rossi, this issue). As a descriptive category, therefore, the use of a rise-fall contour is not commensurable beyond its surface similarity (see also Dingemanse & Enfield Reference Dingemanse and Enfield2015:104). To compare intonation contours pragmatically, we must resort to comparative concepts that are grounded in the oppositional relations between them, formulated by reference to the functions that the contours are typically associated with. From this perspective, French and Italian draw very similar functional boundaries in the environment of other-repetition, with distinct contours associated with seeking confirmation, seeking completion, challenging, and registering (Persson, this issue; Rossi, this issue). So, above and beyond the surface form of individual intonation contours, we must look at the internal organisation of the repertoire.

This discussion shows that the articulation of categories for comparison in pragmatic typology requires us to combine formal, structural, and functional criteria. In the domain of grammar, such combination allows us to compare parts of speech by reference to cross-linguistic propositional acts such as reference, predication, and modification. In the domain of particles, we can juxtapose items that share a structural position in a turn (e.g. initial) and common functional properties (e.g. signalling the speaker's ‘change of state’). In the domain of prosody, we have seen that the comparison of intonation contours must be done by reference to the oppositional relations between them in their respective repertoires. That said, other prosodic features may be more amenable to direct comparison: as we see in the following section, these are features that are not strictly bound up in language-specific phonologies, and are therefore more free to vary according to language-independent principles (see Szczepek Reed & Persson Reference Szczepek Reed and Persson2016).

In sum, pragmatic typology is concerned with similarities and differences in situated language use and communicative behaviour across human groups. Starting from a common conversational structure, its aim is to uncover both invariants of conduct grounded in the generic organisation of interaction and distinct patterns driven by the particulars of different linguistic and cultural systems.

PROSODY

As the title of this article suggests, a key contribution of the present project is the bringing of prosody into pragmatic typology, using other-repetition as the terrain for this new development. Prosody is an established domain of linguistic analysis. However, its conceptualisation and boundaries can vary significantly between different traditions of research. So, just as with other-repetition, some specification of the domain, and of our approach to it, is in order.

Prosody can be generally understood as the nonlexical acoustic-auditory features of speech. This means sound patterns that do not concern the structure and meaning of words, but that belong to the organisation of a whole utterance, including its phrasing, information structure, pragmatic function, and other interactional properties.

Within prosody, the most well known subdomain of analysis is intonation. The simplest definition of intonation is as speech melody: the ‘ensemble of pitch variations in the course of an utterance’ (’t Hart, Collier, & Cohen Reference ’t Hart, Collier and Cohen1990:10).Footnote ³ Intonation has its own tradition of study, which has often involved a separation between aspects of the speech sound that are considered ‘linguistic’ and others that are considered ‘paralinguistic’, that is, not part of linguistic structure and meaning (Crystal Reference Crystal1969; Ladd Reference Ladd2008). This much debated distinction is, however, unnecessary when the speech sound is analysed in its full extent and relevance for social interaction (Local & Walker Reference Local and Walker2005). What is more useful to distinguish is between different ways in which the sound-meaning relation can be structured. One way is discrete, where changes in sound can be partitioned into contrastive units or categories (e.g. a rising vs. a rising-falling contour) and changes in meaning corresponding to these categories are mainly arbitrary-conventional. The other way is gradient, where changes in sound along a continuous variable (e.g. the height and depth of a rising-falling contour) correspond to gradual, analogic changes in meaning (e.g. the degree of the speaker's displayed surprise). These two kinds of sound-meaning relation, or modes of signification, often co-occur and interact in the speech signal, and both of them have been shown to give rise to orderly structures that have communicative and social import. This perspective encourages the treatment of intonation as one of several potentially relevant dimensions of the speech sound, with no a priori privileged status over other parameters (Kelly & Local Reference Kelly and Local1989; Walker Reference Walker, Sidnell and Stivers2013). So intonation—or perhaps more simply, pitch contour—is here considered on a par with parameters including pitch onset, that is, the level of pitch at the beginning of the utterance (higher or lower), and pitch span, the excursion of pitch in the course of the utterance (narrower or wider), as well as with nonpitch parameters like volume (louder or softer) and speech rate (faster or slower). This approach allows us to consider the same speech sound from different angles, either focussing on a particular prosodic feature, or identifying stable bundles of features that work together, on the condition that the categories of analysis be warranted on the basis of what is relevant to participants in producing and understanding talk. This is part of a more general approach that combines linguistics with conversation analysis (see Couper-Kuhlen & Selting Reference Couper-Kuhlen and Selting2018). While putting interaction at the centre stage of linguistic distinctions, this approach continues to draw on the knowledge accumulated in other areas of linguistics, provided that the previously identified structures and mechanisms can stand the test of what is observable in interaction. In this respect, it is worth saying a few more words about intonation to further situate our approach relative to the extensive literature on the topic.

Intonation has been often examined by looking primarily at terminal pitch (e.g. rising or falling), under the explicit or implicit assumption that this contributes to the meaning of an utterance. In fact, intonation comprises more than the end of an utterance and crucially involves the pattern created by prominent movements of pitch around one or more of the utterance's stressed syllables. These prominent movements are commonly referred to as pitch accents, but can also be more plainly seen as ‘landmarks’ (Nolan Reference Nolan, Aarts and McMahon2006:436). While agreeing on the existence of these landmarks, different traditions in intonational phonology diverge in how they analyse intonational structure and how they connect it to function and meaning. For instance, in the so-called British tradition (e.g. O'Connor & Arnold Reference O'Connor and Arnold1973), form-meaning mappings are established holistically by associating the global shape of the melody—the contour or tune (e.g. low rise, rise-fall)—to attitudes such as ‘tentative’, ‘deprecatory’, ‘impressed’, ‘pleading’. By contrast, in the autosegmental-metrical tradition (Pierrehumbert Reference Pierrehumbert1980; Beckman & Pierrehumbert Reference Beckman and Pierrehumbert1986), form and meaning are approached compositionally. The melody is decomposed into a string of localised tonal events, the atoms of which are H(igh) and L(ow) pitch targets as well as combinations of them into rising (L+H) or falling (H+L) pitch accents. These events, comprising pitch accents and edge tones, constitute distinct types of tonal ‘morphemes’, each of which makes an independent contribution to the total meaning of the melody (Pierrehumbert & Hirschberg Reference Pierrehumbert, Hirschberg, Cohen, Morgan and Pollack1990).

In our project, the formal analysis of intonation follows a somewhat hybrid approach that includes aspects from both the British and the autosegmental-metrical traditions. Our descriptions of speech melody emphasise the coherence of its global shape. At the same time, the labels we use to distinguish between alternative melodies typically focus on the section from the primary (or nuclear) pitch accent to the end of the utterance (e.g. rise from low, rise-fall-rise). We also pay special attention to the alignment of pitch components with relevant syllables, making reference to targets or levels of pitch (e.g. H[igh], L[ow]), and in some cases offer transcriptions following ToBI conventions (Silverman, Beckman, Pitrelli, Ostendorf, Wightman, Price, Pierrehumbert, & Hirschberg Reference Silverman, Beckman, Pitrelli, Ostendorf, Wightman, Price, Pierrehumbert and Hirschberg1992).

As far as the study of meaning is concerned, however, our project differs substantially from both traditions, as well as from most other frameworks. The reason is that most research on intonational meaning has been largely based either on introspection or on laboratory speech, or on speech elicited in constrained experimental settings. Such speech often sounds and works differently from conversational speech, and the contexts of use and functions of intonation in the laboratory are significantly impoverished compared to naturally occurring conversation. A few studies in the field of intonational phonology are based on more spontaneous speech produced in task-oriented dialogues (Carletta, Isard, Doherty-Sneddon, Isard, Kowtko, & Anderson Reference Carletta, Isard, Doherty-Sneddon, Isard, Kowtko and Anderson1997; e.g. Grice & Savino Reference Grice and Savino2003; Savino & Grice Reference Savino, Grice, Frota, Elordieta and Prieto2011). This approach adds a significant degree of ecological validity to the analysis, which is framed in terms of ‘conversational moves', including initiating and responding moves, and types of moves such as ‘queries', ‘checks', ‘objections', and ‘acknowledgements'. At the same time, the identification of these moves is based primarily on information status (given, new, accessible) and on the speaker's degree of confidence about it. This information-centred approach limits the reach of these studies into the range of meanings and functions that intonation serves in conversation.Footnote ⁴ In the present project, we analyse conversation as social action. Action distinctions are induced from sequential development and from the responsive behaviour of participants, and the attribution of meaning is grounded in details of talk and other conduct that exhibit the participants' own understanding of each other's actions. The phenomenon of other-repetition offers an environment featuring a diverse range of social actions, allowing us to connect intonational and prosodic form to meanings that go beyond the speaker’s degree of confidence, spanning issues of mutual understanding, alignment, agreement, and affiliation.

The breadth of social-interactional contingencies surrounding other-repetition along with the analytical precision afforded by a sequential, action-based approach to conversation gives us an advantage also over other lines of explanation of prosodic meaning based on the proposal of ‘biological codes’ (Ohala Reference Ohala1983; Gussenhoven Reference Gussenhoven2016). These codes include an iconic relation between prosodic meaning and the anatomical-physiological processes affecting vocal fold vibration: low pitch generally means ‘confidence’ because it is prototypically associated with the larynx of a large, dominant creature, whereas high pitch generally means ‘uncertainty’ by association with the larynx of a small, submissive creature. These associations, however, are rather abstract and hard to apply to language-specific systems.

Having situated the present project in relation to the study of prosodic form and function generally, we can now consider this domain more specifically with respect to cross-linguistic comparison. To pick up where we left off in the previous section, we have said that the analysis of intonation contours across languages requires the mediation of comparative concepts to formulate commensurable form-function distinctions. In what follows, we can expand the discussion to other prosodic features that are instead more amenable to direct comparison.

Among these features are the volume or loudness of an utterance and the ‘scaling’ of its pitch variations, including the height of its onset and the size of its span, as well as more complex features like register, which refers to the relative height of the whole pitch configuration. Crucially, such features can be measured on the basis of two different reference points. One is the speaker's own voice range, which gives us a measurement in terms of individual parameters; the other is previous talk by the same or another speaker, in which case the parameters are relational. Pitch and loudness relations with previous talk have been shown to be central to a wide range of interactional tasks in several languages, and to participate in the constitution of composite prosodic practices referred to as ‘upgrading’ and ‘downgrading’ (Curl Reference Curl2005; Ogden Reference Ogden2006). The former practice typically includes higher pitch onset or register, wider span, and louder volume, whereas the latter typically includes lower pitch onset or register, narrower span, and softer volume. This contrast between prosodic ‘upgrading’ and ‘downgrading’ has a counterpart in terms of individual parameters that can be referred to as ‘large’ and ‘small’ prosody (Pillet-Shore Reference Pillet-Shore2012). The use of these composite prosodic practices and more generally of the pitch and loudness features discussed above, along with features of tempo and articulation, is not strictly constrained by lexical and intonational phonology and is therefore more amenable to direct comparison across languages. For this to work, however, the use of these features has to be situated in conversational structure, including turns at talk and turn-constructional units (Sacks, Schegloff, & Jefferson Reference Sacks, Schegloff and Jefferson1974). These are the units that ‘enable us to establish structural “sameness”, and to compare “like with like” both phonetically and interactionally’ (Local Reference Local2007:1). Moreover, the conversational phenomenon of other-repetition has a further advantage for prosodic analysis, in that the varying lexical content of a repetition is, in a way, a constant across languages, and therefore provides an especially good locus for examining prosodic variation, as well as its combination with other linguistic and nonlinguistic resources.

Now that we have introduced the conversational phenomenon under analysis, situated its study in the field of pragmatic typology, and specified our approach to the analysis of prosody, we can recap the central questions of the project: what is the role of prosody in distinguishing the functions of other-repetition? How does prosody interact with other interactional resources? And how does this process differ across languages?

DATA AND METHODS

The project reported on in this special issue is based on data from naturally occurring conversation in five languages, collected and analysed by a team of researchers (see Table 1) that worked together between 2015 and 2018, with a primary base at the University of Helsinki.Footnote ⁵

Table 1.

Languages and researchers.

Detailed information on the languages, the particular variety or varieties examined, and on the corpora used is given in the language-specific articles. However, a few general notes on our corpora are in order here. All corpora include recordings of face-to-face conversation, with the English, Finnish, and French corpora also featuring telephone conversations. A large proportion of the conversations is informal, taking place in domestic settings among family, friends, and neighbours. Some French and Swedish recordings are from institutional settings, including service encounters at a ticket office, in a bakery, in a cheese shop, as well as workplace interactions and radio phone-in shows. Pooling data from these different settings is warranted by the fact that we did not observe any particular effects of setting on the resources that speakers use to design and understand other-repetitions. What we do observe is that certain kinds of institutional interactions, especially service encounters and radio phone-in shows, can increase the frequency of certain functions of other-repetition, especially registering (see extract (1) below). These imbalances in the frequency of certain functions, however, do not affect the main goal of the project, which is to account for how speakers and recipients distinguish among them.

All of the studies in this special issue draw on methods from linguistics and conversation analysis, including two particular intersections between these fields: the prosody and phonetics of talk-in-interaction, and pragmatic typology.Footnote ⁶ The main tenets and specifics of the methods used in these fields have already been extensively discussed in the previous two sections. With respect to pragmatic typology, however, it may be worth going over some of the practical steps of its mixed methodology to better situate the work reported in this special issue within the larger project of which it is part. As already discussed, the project as a whole joins a recent lineage of research involving the structured comparison of a formally and sequentially delimited phenomenon across languages. This begins with data sessions devoted to the qualitative analysis of audiovisual recordings in each language, with a thoroughgoing examination of each exchange in its own right, situated in the richness of its social context and enmeshed in a flux of other events. This allows analysts to get a firm grasp of the details of single cases before making generalisations across cases, and to identify deviations from a pattern that may reveal the violation of an underlying norm, thereby providing evidence for it. This qualitative analysis is also aimed at developing a principled definition of the phenomenon, which requires the discussion of borderline cases. Firming up the boundaries of the phenomenon is a prerequisite for the systematic collection of cases on a large scale; together with the qualitative analysis of linguistic and interactional patterns, this provides the foundation for the subsequent quantitative analysis. This phase begins with the design of a coding scheme to capture the core set of formal and functional categories, parameters, and variables that have emerged from the qualitative analysis and that are amenable to quantification. The coding scheme is piloted, revised, and then further refined after group coding exercises, which serve to increase the calibration among coders. When the coding scheme has been sufficiently tested and assimilated, the independent coding of each language's dataset can begin. The last step in the process is a test of intercoder reliability.

The coding scheme designed for the quantitative stage of the present project and the main quantitative cross-linguistic analysis are in preparation for publication at a later stage. The findings reported in the language-specific articles in this special issue are primarily the product of qualitative analysis.

DELIMITING THE PHENOMENON

We have already delimited our phenomenon as: verbal repetition, intended as exact or modified replication of lexical material, produced by someone other than the speaker of the first saying, so as to problematise or otherwise engage with what has been said making a response conditionally or optionally relevant. Here we draw the boundaries of the phenomenon in greater detail.

Let us begin by describing its basic sequential structure. Extract (1) reproduces the final moments of a service encounter at a ticket office (from Huhtamäki, Lindström, & Londen, this issue). Before the extract begins, the customer has sought and obtained information about an upcoming event from the ticket seller, and has begun moving away from the counter. The customer then stops and turns back to the ticket seller to double check on the time of the event (line 1).

1. (1) Swedish (HEL_LUC_RL_004_2 00:05)

   

This example makes very clear the three-part structure of an other-repetition sequence: in the original turn, Speaker A says something; in the repetition turn, Speaker B repeats all or part of what Speaker A has said; in the response turn, Speaker A responds to the repetition.Footnote ⁷ This pattern is in many respects analogous to that of an other-initiated repair sequence (Schegloff, Jefferson, & Sacks Reference Schegloff, Jefferson; and Sacks1977; Dingemanse & Enfield Reference Dingemanse and Enfield2015). However, the phenomenon of other-repetition is formally narrower and functionally wider than other-initiated repair, so it is important to further specify its boundaries in terms of linguistic construction, turn structure, sequential position, and response relevance.

RECURRENT TYPES OF ACTION

One of the basic findings of the present project is that other-repetition is used to accomplish certain recurrent types of action that are common across languages. These types of action can be seen as generic functions that emerge out of the complex social-interactional contingencies and nuances that other-repetitions are enmeshed in—regular patterns that cut across the thick particulars of each case. Most types of action that we identify here have already been described, more or less extensively, in earlier research, including in studies of other-repetition. Some types have also already been found to be cross-linguistically recurrent in an earlier comparative project on other-initiated repair (Dingemanse & Enfield Reference Dingemanse and Enfield2015). In the present project, we build on this previous research to map out a larger space of action types ranging from actions that fall within repair initiation to actions that extend beyond repair initiation, to actions that are alternative to it. Our systematisation of this action space and of the distinctions within it goes hand in hand with our goal of explaining how people design and recognise the actions accomplished by other-repetition, and how this process is affected by cross-linguistic variation. Some of the action types discussed below are strongly associated with particular types of responses discussed in the previous section, including by relations of conditional relevance and type conformity (Schegloff & Sacks Reference Schegloff and Sacks1973; Raymond Reference Raymond2003); for other action types, the response space is wider and less constrained.