Structurally Similar Formats are Not Functionally Equivalent across Languages: Requests for Reconfirmation in Comparative Perspective

In: Contrastive Pragmatics
Sonja Gipper Department of Linguistics, University of Cologne Cologne Germany

Search for other papers by Sonja Gipper in
Current site
Google Scholar
Katharina König German Department, University of Münster Münster Germany

Search for other papers by Katharina König in
Current site
Google Scholar
, and
Kathrin Weber Department of German Linguistics, University of Jena Jena Germany

Search for other papers by Kathrin Weber in
Current site
Google Scholar
Open Access


This study is concerned with a cross-linguistic comparison of requests for reconfirmation (RfRCs) in Yurakaré, German, and Low German mundane spoken conversations. We focus on two different RfRC formats, namely token RfRCs (such as German echt? ‘really’) and RfRCs that repeat (part of) another speaker’s prior turn. We show that these RfRC formats are used to accomplish different social actions in the three languages under study, ranging from registering news to challenging prior information. Token RfRCs serve as versatile resources for responding to news in all three languages. Repeat RfRCs, in contrast, are used in divergent ways: In German, they are mainly treated as challenges, while showing only a weak contextualisation of challenges in Low German and virtually no potential for being understood as challenges in Yurakaré. Our study thus demonstrates that structurally similar RfRC formats can be put to use for different actions and sequence trajectories across languages.

1 Introduction

Recent years have seen a growing interest in the cross-linguistic sequential analysis of polar questions, i.e., utterances that invite a transfer of knowledge from an addressee to a speaker and at the same time prefigure the answer space in that they make responses relevant that affirm/deny or confirm/disconfirm. Languages afford different linguistic resources for formulating polar questions, such as verb position, tags, morphological markers, and prosodic contours (König and Siemund, 2007), which shape the gradient of stances polar questions index and the range of actions they implement (as e.g., the distinctiveness of requests for affirmation and requests for confirmation, Enfield et al., 2010). Concerning responses to polar questions, languages exhibit different profiles in their choice of either repeat formats or conventionalised particle formats to affirm/confirm a prior question (Holmberg, 2016; Enfield et al., 2019).

This paper is concerned with a particular type of polar question, requests for reconfirmation (RfRCs), which ask for the reconfirmation of something another speaker has stated in a previous turn. They can be realised in token format, as in excerpt (1), or in repeat format, as in excerpt (2).

RfRCs are sometimes pooled with requests for confirmation (e.g. Raymond and Stivers, 2016: 326) as they are “routinely treated as seeking confirmation” (Stivers and Enfield, 2010: 2621). Unlike with requests for confirmation, however, speakers producing an RfRC do not introduce a new proposition to the discourse but operate back on prior talk. Moreover, RfRCs can be routinised in form (Thompson et al., 2015), they merely invite a response rather than making it relevant (Gubina and Betz, 2021), and they favour confirmation over disconfirmation more strongly than requests for confirmation (Marmorstein and Szczepek Reed, this issue). This paper, thus, sets out to study RfRCs separately from other types of polar questions.

In previous research on selected languages, RfRCs have been referred to as “assertions of ritualized disbelief” (Heritage, 1984: 339) or as “show repairs” (Imo, 2009), whose main job is to express astonishment or incredulity. As a starting point for cross-linguistic interactional research, such functional ascriptions can be problematic as they presuppose that RfRCs accomplish comparable actions across languages. Starting from particular forms can be equally challenging as they may not be found in all languages (or not with the same relative frequency). Moreover, each form can come with language-specific ‘collateral effects’ (Sidnell and Enfield, 2012) and may contribute to distinct interactional styles in particular ways (Brown, Sicoli, and Le Guen, 2021). In order to conduct a comparative analysis in line with recent approaches to pragmatic typology (Dingemanse et al., 2014; Rossi, 2020), we start from a sequential definition of RfRCs (see Dingemanse and Floyd, 2014 for this “natural control method”) that ties them to their conversational ecology rather than to a particular function or form.

To date, much research on RfRCs is based on British and American English and a few other spoken standard languages (e.g. Heritage, 1984; Thompson et al., 2015; Gubina and Betz, 2021); there are only very few studies that present comparative analyses (Kaimaki, 2012; Aldrup, this issue; Marmorstein and Szczepek Reed, this issue). This paper aims to expand this research by presenting a cross-linguistic interactional study on RfRCs in German and Low German, two typologically closely related Germanic languages, and Yurakaré, a language isolate spoken in Bolivia. The sample thus includes a standard language, a non-standard dialect variety, and an Indigenous language. Our study focuses on the sequential environments in which RfRCs occur and the practices speakers deploy to invite minimal or more elaborate responses while also analysing their quantitative distribution. Despite some formal commonalities, we show that the two main RfRC formats – token and repeat RfRCs – accomplish very different actions in the languages under study, i.e. structural similarity does not necessarily imply functional equivalence. Methodologically, we hope to show how a quantitative approach engenders questions for a qualitative analysis, which can help to unravel the functional specificities of RfRCs across languages.

In Section 2, we provide an outline of previous conversation-analytic research on token and repeat RfRCs. After that, we introduce our data and methods used for building our collections in Section 3. Here, we also discuss in more detail how RfRCs can be delimited from other interactional devices such as requests for confirmation. The section furthermore introduces the coding scheme applied in this study. Section 4 presents a quantitative analysis of language-specific formats of RfRCs, their prosodic design, preferred response types and their sequential embeddings, identifying commonalities as well as divergent patterns across the languages. Then, in Section 5, we zoom in on the response trajectories after token and repeat RfRCs following informings, illustrating how they are deployed as different social actions in the three languages by means of a quantitative and a qualitative analysis. We conclude the paper by discussing the results and their implications for comparative and interactional typological research (Section 6).

2 Formats and Functions of RfRCs

Requests for reconfirmation are responsive turns that treat a prior as new and somewhat notable or remarkable information (see Marmorstein and Szczepek Reed, this issue). RfRCs are usually distinguished from change-of-state tokens or news receipts, which also “accept prior talk as informative” (Heritage, 1984: 335) but are routinely treated as closing implicative as they do not ask for additional reconfirmation (Imo, 2009; Heinemann, 2017). However, some studies call into question such a clear-cut distinction: In Finnish, for instance, some change-of-state tokens can be followed by reconfirmation when placed in specific sequential environments (Koivisto, 2015).

RfRCs can implement various actions: They are organised on a continuum of merely registering or acknowledging news (Maynard, 1997), expressing affiliation (Imo, 2011; Gubina and Betz, 2021), framing the news as not yet fully accepted or processed (Thompson et al., 2015), counter to expectation (Robinson, 2009), surprising (Selting, 1996), or challenging the just presented claim (Heritage, 1984; Benjamin and Walker, 2013). Therefore, RfRCs are involved in different sequential trajectories. Much like continuers, they can receive no notable uptake (Thompson et al., 2015) or they can work to expand the sequence by eliciting (at least) a simple reconfirmation (Heritage, 1984), by inviting elaboration on the matter at hand (Gubina and Betz, 2021), or by eliciting accounts that substantiate or explain a previous claim (Raymond and Stivers, 2016). Based on this wide range of opportunities, Gubina and Betz (2021) argue that it is better to think of RfRCs’ conditional relevance on a gradient; they open up a response space in which different forms of uptake are possible and it is up to the addressee to decide which trajectory to follow.

Moreover, responses are fitted to the local contexts in which RfRCs emerge. RfRCs referring to unsolicited informings tend to receive minimal uptake (Thompson et al., 2015) in contrast to RfRCs in contexts in which some kind of trouble (e.g. issues of epistemic primacy, overt disconfirmation of an assumption) has been registered earlier (Gubina and Betz, 2021). Following informings on a person’s behaviour, personal opinions or choices, RfRCs are regularly understood as off-record account solicitations, which do “not initially appear to be soliciting an account at all” (Raymond and Stivers, 2016: 325). In doctor-patient interactions, they can invite further informings and thus work as a minimal and indirect form of resistance to a diagnosis (Stivers, 2011b).

Concerning their form, previous studies document that RfRCs can be realised as (partial) repeats or in more or less routinised formats, which differ in the strength with which they identify their point of reference. In English, for instance, formats range from particles or lexically fixed expressions to minimal or expanded repeats (Thompson et al., 2015). Across languages, routinised formats are often derived from expressions dealing with the truthfulness, adequacy or veracity of a previous turn, such as English really (Jefferson, 1981; Heritage, 1984; Maynard, 1997), German ehrlich (‘honestly’, Imo, 2009) or Arabic waḷḷāhi (‘by God’, Marmorstein and Szczepek Reed, this issue). Usually studies tend to concentrate on either such routinised formats or on repeat RfRCs (see Aldrup, this issue). With this study, we aim at a comprehensive approach that takes into account the different formats languages recruit for requesting reconfirmation.

Prosody and embodied conduct are important for indexing epistemic stances and scaling emotive involvement in RfRCs. For instance, sustained gaze, raised eyebrows or sudden head tilts can be deployed with RfRCs as expressions of disbelief and surprise (Kendrick, 2015; Thompson et al., 2015; Couper-Kuhlen, 2020; Gubina and Betz, 2021). However, these embodied displays heighten the stances of the verbal action rather than acting as distinctive cues. Prosodic resources, in contrast, seem to play a more important role in the formation of RfRCs. For other-repeats, recent studies show that prosody can help to distinguish RfRCs from other repeat actions (see papers in Rossi, 2020). In particular, repeats expressing surprise or disbelief tend to stand out prosodically (Selting, 1996; Benjamin and Walker, 2013). In other formats such as clausal RfRCs in English (do I?, you did?), too, ‘expressive’ prosody (high pitch register, loudness, rise-fall pitch contours, lengthening) can index “heightened emotive involvement on the part of the responder” (Thompson et al., 2015: 92). However, there is no straightforward mapping of one prosodic cue with a particular function. Rather, an assemblage of varying prosodic resources (pitch contours, pitch range, loudness, lengthening) as well as verbal and contextual factors are used for distinguishing RfRCs from other actions such as repair or news registering (Couper-Kuhlen, 2020).1

Concerning final intonation of routinised RfRC formats, studies report differences within and across languages. In English, there is a general tendency for particles with final rising intonation such as oh and really to make a response relevant whereas particles with final falling intonation merely register news (Local, 1996 for a fine-grained phonetic analysis; Thompson et al., 2015: 64–84).2 For German ja (‘yes’) and was (‘what’), Selting (1996) reports instances in which it is solely the prosodic design (pitch peaks, marked length or loudness) that distinguishes RfRCs from other actions such as signalling a problem of hearing or understanding (also see Imo, 2011). For German echt, however, Gubina and Betz (2021) find no systematic link between the token’s final intonation and its uptake. Similarly, for English really and Arabic waḷḷāhi Marmorstein and Szczepek Reed (this issue) report that no patterns could be identified regarding the forms’ prosodic design. To sum up, languages differ not only in the assemblage of prosodic resources used for making reconfirmation relevant but also in how stable this prosodic distinction is in different RfRC formats.

So far, most studies focus on RfRCs in spoken Indo-European languages such as English or German, and only few take a comparative approach (Kaimaki, 2012; Rossi, 2020). Moreover, studies often look at either repeat (Aldrup, this issue) or (selected) routinised RfRC formats separately (Marmorstein and Szczepek Reed, this issue). It still has to be determined, though, how languages make use of the distinction between routinised token forms and repeats. For English, Thompson et al. (2015: 135–136) note a tendency that particle RfRCs often do not receive any uptake and primarily register something as news while repeat formats tend to manifest some problem with the prior and engender more elaborate answers (also see Heritage, 1984: 343). Moreover, the authors note differences in frequency: Minimal formats are the most frequent in their data while expanded repeat formats are used to a lesser degree (Thompson et al., 2015: 53–54). However, their study – like many others – offers “impressions of frequency” (Thompson et al., 2015: 53) rather than presenting statistical results. This, we argue, calls for a combined qualitative and quantitative cross-linguistic approach. Our quantitative analysis provides valid frequency ratios of RfRCs and the different formats across languages. These distributions give rise to a subsequent qualitative analysis of the actions the different RfRC formats implement and the response trajectories they invite in each language.

3 Data and Methodology

In this paper, we combine quantitative and qualitative approaches in the analysis of RfRC sequences in mundane spoken interaction. In the following, we elaborate the methodological aspects of the study. Section 3.1 introduces the languages under investigation. Subsequently, Section 3.2 presents the procedures for building our collections and discusses our data-driven criteria for the inclusion and exclusion of boundary cases. Section 3.3 delimits RfRCs from similar social actions, while Section 3.4 presents the coding scheme. Finally, the method of comparison is presented in Section 3.5.

3.1 Description of Datasets

Yurakaré [glottocode: yura1255, denoted as YUR in this paper] is an agglutinative language isolate spoken by around 1,600 people (INE, 2015: 32) in the Andean foothill area of central Bolivia (see e.g. van Gijn, 2006). The Yurakaré dataset used in this study consists of a subset of the Yurakaré section (van Gijn et al., 2011) of the DobeS archive at The Language Archive (TLA), MPI for Psycholinguistics, Nijmegen, The Netherlands. The data were collected during a language documentation project (2006–2011) by members of the project team. All data used for this paper contain video-recorded conversations between people who are familiar with each other, in many cases members of the same family. The data were transcribed and translated into Spanish by speakers of Yurakaré in the ELAN annotation tool developed at The Language Archive, MPI Nijmegen, The Netherlands (e.g. Brugman and Russel, 2004).

German [glottocode: stan1295, denoted as GER in this paper] is an inflectional standardised West Germanic language (Haspelmath, 2008). The dataset comprises approximately 24 hours of mundane conversation between friends and family. The speakers come from the northwest of Germany (mostly Westphalia and Rhineland). Data were collected as audio and video as part of a larger corpus project “Korpus multimodale Interaktion (KoMI)” by Pepe Droste in 2016 and 2017. The conversations are transcribed according to the GAT 2 conventions (Selting et al., 2009) in EXMARaLDA (Schmidt and Wörner, 2005).

Low German [glottocode: lowg1239, denoted as LoG in this paper] is an inflectional non-standardised (dialect) Germanic language. Due to the historical kontinentalwestgermanisches Dialektkontinuum (‘continental West Germanic dialect continuum’), Low German is structurally different from (High) German (Goossens, 1980). In line with the general decline of dialects in Germany, Low German is a primarily spoken variety that is losing importance as an everyday language. The Low German data comprise approximately 60 hours of mundane talk-in-interaction and were collected in the north-western part of Germany in the Westphalian language area (for a description of the project see Weber, 2020). All speakers are native Westphalians, who acquired Low German as a first language or in a bilingual fashion with High German. In their everyday talk, they speak both Low German and (regional) High German. The data were transcribed using the GAT 2 conventions (Selting et al., 2009) in EXMARaLDA (Schmidt and Wörner, 2005). They are only available in audio and, thus, cannot be analysed multimodally.

Our datasets for the three languages exhibit important differences regarding the sociodemographic background of the speakers. The German data exclusively contains talk from younger speakers, while the speakers in the Low German collection are mostly over 50 years old. The Yurakaré collection contains a smaller number of speakers covering different age groups (see Section 3.2 for further details). Low German and Yurakaré are both languages that are spoken and acquired in a contact situation. Low German speakers are usually bilingual with (regional) High German and Yurakaré speakers with Spanish. This is not regularly the case for the speakers in the German dataset. Finally, the Yurakaré speakers who participated in the data collection live in small and remote communities, which is not the case for the German and Low German speakers. This raises questions regarding the comparability of the data, and of possibilities of controlling for these factors. Comparative studies based on near-natural interaction often rely on pre-existing data that have not originally been collected for a comparative endeavour (Kornfeld et al., 2023: 106). This is also the situation in our case: Our study relies on datasets that already existed before we decided to compare them.

However, we still would like to make a case for the validity of observations based on the comparison of our three datasets. While not representing the same age ranges of speakers, the datasets still represent practices that are observable across certain communities of speakers. Moreover, sometimes it is not possible to control for sociodemographic factors, as for instance in the case of Low German, where a break in intergenerational transmission has taken place and young speakers are therefore rare. Moreover, our study constitutes a first cross-linguistic exploration of RfRC sequences in the three given languages. It will be of interest to investigate the role of sociodemographic factors in RfRC formulation in the future.

In terms of the social situations investigated, our corpora are comparable in that for all languages they predominatly consist of informal conversations among family and friends where no other activities were carried out in parallel. Another strand of investigation in the future would be to study RfRCs as they are employed in different types of situations or activities across languages. Recent endeavours of building parallel (or parallax, see Barth and Evans, 2017) corpora such as the Parallel European Corpus of Informal Interaction (PECII, Kornfeld et al., 2023) or the Social Cognition Parallax Interview Corpus (SCOPIC, Barth and Evans, 2017) will play an important role in making such work possible.

3.2 Building the Collections

This study builds on the research of the Scientific Network “Interactional Linguistics”, which conducts a cross-linguistic analysis of approximately 200 instances of request for confirmation sequences in each of 10 languages (cf. König and Pfeiffer, in prep.). The network’s definition of this type of polar question excludes RfRCs as they do not introduce the relevant proposition but rather operate back on a prior speaker’s turn (cf. König, Pfeiffer and Weber, in prep., also see Section 3.3). Our interest to study RfRCs was motivated by this decision, i.e. we wanted to build an additional collection to determine the differences between these two types of polar questions in terms of their linguistic design and the responses they invite. To this end, we decided to use the same stretch of data needed to identify 200 requests for confirmation for collecting RfRCs in Yurakaré, Low German and German. Even though this yields collections of different sizes, it allows us to determine the relative frequency of RfRCs in comparison to requests for confirmation. Each of the authors was responsible for building the collection for the language of their scientific expertise: The German collection was created by König, the Low German collection by Weber, and the Yurakaré collection by Gipper.

The network collection was compiled by identifying the first 15 requests for confirmation (or less if the recording did not contain as many) through an exhaustive search of a continuous stretch of interaction starting at the beginning of a given recording until 200 cases were gathered (see König, Pfeiffer and Weber, in prep.).3 As a next step, we inspected the same stretches of conversation for RfRC sequences. We first cast a very wide net, including all cases that could potentially be identified as RfRCs. Data sessions were conducted to discuss the inclusion of different types of cases. Based on a more detailed sequential analysis, we then iteratively excluded some cases (e.g. those that could be straightforwardly classified as repair) and included others (e.g. change-of-state tokens and continuers that invite reconfirmation as well as instances bordering on other-initiation of repair, see Section 3.3). In this way, iterative qualitative analyses without a pre-set formal focus shaped the data collection process.

For Yurakaré, in the stretch of data in which the 200 RfC sequences were found 167 instances of RfRCs by 11 speakers (aged from approx. 15 to approx. 65) could be identified. The cases come from a total of around 163 minutes of conversational data (10 interactions; 1 multiparty, 9 dyadic). The normalised token frequency for Yurakaré is 61.3 RfRCs per hour. In the German corpus, the 200 RfCs were identified in 569 minutes of talk (14 interactions; 8 multiparty, 6 dyadic). In this stretch of data, 67 RfRCs issued by 25 speakers could be collected. This results in a normalised frequency of 7.1 RfRCs per hour for the German data. The speakers’ age ranges from 19 to 33 years. The 200 RfCs in the Low German data were found in 354 minutes of talk (15 interactions; 11 multiparty; 4 dyadic). Within this stretch of conversation, the corpus yielded a total of 77 RfRCs by 21 speakers. The normalised token frequency per hour is 13.1 in the Low German data. The speakers are between 40 to 80 years of age.

3.3 Delimiting RfRCs

This section presents relevant characteristics of RfRCs that emerged as a result of our collection building process and reports on actions bordering on this understanding. An RfRC has a particular sequential structure as it operates back on another speaker’s prior utterance. The previous speaker is asked to reconfirm what they have just said, i.e. the RfRC identifies a foregoing utterance (or parts of it) as a reconfirmable. In contrast to requests for confirmation (see König and Pfeiffer, in prep.), the speaker issuing the RfRC does not introduce a new proposition to the conversation. Instead, RfRCs present “known-answer questions” (Steensig and Heinemann, 2013), which invite reconfirmation or commitment (Marmorstein and Szczepek Reed, this issue). Responses to RfRCs usually – but not necessarily – offer a reconfirmation and may be expanded by further talk related to the matter at hand. Consider the sequential structure in excerpt (3).

This sequential structure is similar to that of other-initiated repair (OIR). However, previous research has also noted relevant differences between the two: For instance, Dingemanse and Enfield (2015: 100) exclude routinised RfRCs such as Dutch echt (‘really’) from their collections of OIR as they are not treated as the instantiation of a problem of hearing or understanding (Did you say X? Did you mean X?) but rather as taking issue with the truthfulness, plausibility or credibility of the prior utterance. This, however, also applies to some other-repeats (Selting, 1996; Schegloff, 2007: 151–155; Rossi, 2020; also see Aldrup, this issue), which Dingemanse and Enfield (2015: 100–101) include in their study of OIR due to the format’s similarity to candidate understandings. Kendrick (2015: 181) coins the term ‘pseudo OIR’ for utterances which resemble repair but are used to accomplish a different focal action. Another relevant aspect in which RfRCs can be distinguished from OIR sequences concerns their fittedness to the reconfirmable. Repair is a generic practice for orienting to trouble whenever it occurs. RfRCs, in contrast, regularly follow utterances which are framed as informings (Thompson et al., 2015: 60–64), i.e. in contexts in which RfRC-speakers are recognised as being less knowledgeable, for instance as a recipient of tellings or of answers to questions they have previously posed (Maynard, 1997; Robinson, 2013; Thompson et al., 2015).

Crucial for our study is the kind of responsive action that RfRCs call for and regularly receive, a reconfirmation. Such a response accomplishes more than attesting that something holds true, it does more than simply confirm. By virtue of operating back on a prior turn (without altering its content), it offers a reconfirmation and thus a recommitment that something is the case (also see Aldrup, this issue; Marmorstein and Szczepek Reed, this issue). Notably, the design of RfRC responses differs from that of responses to more repair-like practices (Thompson et al., 2015: 60–64). Our collections corroborate this observation: We did not find response tokens which highlight the adequacy of the reconfirmable (such as German genau ‘exactly’, das stimmt ‘that’s right’) but – among others – tokens dealing with askability (natürlich ‘of course’, see Stivers, 2011a). In elaborated responses, speakers usually do not correct, rephrase, clarify or complete their previous turns, as would be expected for OIRs, but rather offer more details or substantiations of their claims. Moreover, we do not find adjustments of articulation in the form of more carefully pronounced repeats that would deal with problems of hearing (Robinson, 2013). Further evidence for the non-repair character of RfRCs can be found in follow-up turns by the RfRC speakers, who often do not offer claims of now-understanding (Koivisto, 2019) but rather assessments (Heritage, 1984), markers of surprise and unexpectedness (Wilkinson and Kitzinger, 2006), metacommunicative comments on their epistemic position (I did not know that), requests for accounts (Kendrick, 2015: 187), follow-up questions about the matter at hand, news receipts, or additional RfRCs. While these observations corroborate that there are functional differences between OIRs and RfRCs, we also came across borderline cases in which no clear-cut distinction could be made (see also Aldrup, this issue). We included such cases in our collections so that they can be related to core cases of RfRCs.

RfRCs open a space for responding to them. In our collections we included formats which regularly receive uptake in the form of a reconfirmation. However, other types of uptake or a lack of uptake are also possible. To be able to determine the proportions of different kinds of uptake, we included all formats that regularly receive a reconfirmation in the respective languages. For German this means that we included all instances of echt, as this form regularly invites the reconfirmation of a prior, even though not all instances receive a verbal response (Gubina and Betz, 2021). Moreover, we included two special formats: (1) change-of-state tokens in all three languages, and (2) the Yurakaré response token , which is otherwise used as a continuer. For these two formats, we only included cases where a reconfirming response was given.

3.4 Coding

In order to enable a quantitative analysis which can help to identify language- specific tendencies, we decided to code the data for features concerning the RfRC turn-design, its sequential embedding, and the ensuing response. Part of the coding scheme is based on the categories developed and tested for interrater reliability in the Scientific Network “Interactional Linguistics” (cf. König, Pfeiffer and Weber, in prep.). All coding was gradually adjusted by iterative qualitative analyses and is thus motivated by the data (see also Stivers, 2015). In the data collection process, each of the authors was responsible for coding the language of their scientific expertise. However, through joint examinations of cases in all languages in data sessions, we ensured that we had a shared understanding of the coding categories that was applicable to all three languages. As the Low German dataset does not provide video recordings, we decided to focus on verbal responses only in the coding of responses.4 To make the datasets comparable, we classified German and Yurakaré cases where only a non-verbal confirming response was given as ‘no response’. Given that these cases are very rare (three for Yurakaré and two for German), this decision does not have a great impact on the analysis. In the following, we list the variables and the values which we coded for.

RfRC Turn

RfRC format
Values: token RfRC; repeat RfRC; other

We first coded for different types of formats used for requesting reconfirmation. In accordance with the literature (see Section 2), we coded for repeat formats (repeat RfRC), including ‘near copies’ (Rossi, 2020: 496) or partial repeats, and token-like formats (token RfRC). The latter category encompasses short and lexically fixed expressions such as Yurakaré achama (‘is that so?’) and German wirklich (‘for real?’). The term token was chosen explicitly to refrain from an unwarranted grammatical categorisation which might not be applicable to all languages and to capture various linguistic formats that languages recruit to request reconfirmation (which in our data range from verbs, adjectives or response tokens to items usually described as change-of-state tokens). The few cases that could not be categorised as either token or repeat RfRCs were coded as other.

RfRC format specific
Values: e.g. echt ‘really’; ja ‘yes’

For non-repeat formats, the language-specific lexical form of the RfRC was noted (e.g. echt ‘really’ for German, achama ‘is that so?’ for Yurakaré, or ja/jo ‘yes’ for Low German).

Prosodic upgrade5 (of RfRC)
Values: 0=without prosodic upgrade; 1= with prosodic upgrade

Following the literature on the prosodic design of RfRCs (see Section 2), we analysed the cases for features of loudness, lengthening, pitch jump, and marked intonation contour which may cluster (e.g. Pillet-Shore, 2012: 383) to produce what we call a ‘prosodic upgrade’ (other terms used in the literature are ‘punched-up prosody’ (Wilkinson and Kitzinger, 2006) and ‘expressive prosody’ (Thompson et al., 2015)). We coded for the presence of one or more of these features in a binary fashion (1=with prosodic upgrade (at least one of the features is present); 0=without prosodic upgrade (none of the features is present)).

Final intonation (of RfRC)
Values: fall; level; rise

As a second aspect of the prosodic design, we coded for the final pitch contours of the RfRCs in our data following the British-School taxonomy (fall; level; rise contours; see Couper-Kuhlen, 2020). Our focus is on the nuclear syllable and its pitch configuration up to the final syllable of the unit.

Sequential Embedding (of RfRC)

Values: informings; responses to information questions; confirmatory responses to some confirmable; contradictory responses to some confirmable; confirmatory responses to RfRCs; other

Following Thompson et al. (2015), we coded the action of the turn preceding the RfRC. In our collections, we found the following actions to regularly precede RfRCs: (1) informings (volunteered or elaborated, i.e. response turns which elaborate on a topic), (2) responses to information questions (i.e. questions which target a specific piece of information), (3) confirmatory responses to some confirmable (i.e. simple, non-elaborated confirmations, which have been invited by polar questions), (4) contradictory responses to some confirmable (i.e. responses which disconfirm or reject a polar question) and (5) confirmatory responses to RfRCs (i.e. reconfirmations which have been invited by an RfRC). We coded cases which do not pertain to any of these domains as (6) other.

Response Turn (to RfRC)

Response type
Values: confirmation; disconfirmation; neither; no response

If there was a response to an RfRC, we coded the response type concerning the values (1) confirmation, (2) disconfirmation, and (3) neither. No uptake was coded as (4) no response.

Response format
Values: response token; repeat; response token + repeat; response token + elaboration

We coded the format of the responses. These include the following values: (1) response token, (2) repeat, (3) combination of response token and repeat, and (4) response token plus elaboration.

3.5 Method of Comparison

After the data were coded, we first calculated the relative frequencies for all variables to be able to compare them across the languages. Our study is exploratory in nature, hence our quantitative results are not meant to test pre-formulated hypotheses. We report descriptive statistics in the form of relative frequencies of variables and statistical association measurements based on Cramér’s V (see supplementary material: As the absolute frequencies of RfRC instances vary across the three languages, and for some categories the absolute numbers are very small, the reported distributions represent tendencies that will need to be statistically confirmed with much larger data sets. We use the quantitative results to generate questions for a qualitative analysis regarding the similarities and differences in functions of the two RfRC formats – tokens and repeats – in the three languages. All statistical analyses were conducted in R version 4.2.3 (R Core Team, 2023), in parts with the package DescTools (Signorell et mult. al., 2023). The plots were generated with the ggplot2 package (Wickham, 2016).

4 Formatting, Sequential Embedding, and Preference Organisation of RfRCs

In this section, we present the quantitative findings on the two main RfRC formats, their sequential embedding, and the response types they solicit. The corpus frequencies reveal substantial differences between the languages which led us to conduct a qualitative analysis that looks into the social actions that RfRCs accomplish in the given languages (Sections 5.3 and 5.4). As Section 3.2 showed, a first general difference is the overall rate of occurrence of RfRCs in the corpora: The Yurakaré collection comprises 167 RfRC sequences compared to 77 cases in Low German and 67 in German. These frequencies become even more relevant when we look at them relative to the time span in which these cases occur: The 167 instances of Yurakaré stem from less than three hours of conversation, whereas the Low German cases occur in almost 6 hours of talk and the German instances in roughly 9.5 hours.

Regarding the typical formats of RfRCs (see variable RfRC format in Section 3.4), we find that in all three languages they split into two main categories: More or less routinised short (mostly one-word) formats, and different types of repeats. In this paper, we refer to the non-repeat formats as ‘token RfRCs’, as they are at least to some extent conventionalised for expressing RfRCs. Since our study focuses on token and repeat RfRCs as the most frequent formats in all three languages, formats that deviate from these were removed in a next step. In Yurakaré, six cases were found that take other formats, such as reformulations and self-repeats. Moreover, in Low German, there were three cases where the RfRC contained an explicit evaluation (e.g. so viel? ‘that much?’). These cases are disregarded in the following analysis. This results in 161 RfRCs in Yurakaré, 74 in Low German, and 67 in German that form the basis of our quantitative analysis.

Figure 1 shows the differences in the relative frequency ratios of the RfRC formats (token and repeat) in all three languages. It shows that the proportion of repeat formats is much higher in Yurakaré,6 while German and Low German prefer token formats. The statistics in the supplementary material (p. 1) show that there is a large association (E=0.43; df=2) between language and RfRC format, which is also highly significant7 (p<0.01).

Figure 1
Figure 1

Relative frequencies of RfRC formats per language

Citation: Contrastive Pragmatics 2024; 10.1163/26660393-bja10097

While all three languages employ RfRCs in token format – albeit with different frequency ratios – these are formed with different linguistic resources across the languages. Table 1 shows the formats and frequencies for the token RfRC formats in our collections. In German, we mostly find adverbs such as echt ‘really’ and ehrlich ‘honestly’, where echt ‘really’ serves as the most frequent RfRC format in this language. In Low German, formats that have been described as change-of-state tokens for German (such as achso and aha, Golato and Betz, 2008; Imo, 2009) are recurrently used as RfRCs and receive a reconfirming response, while this is not frequently observed in German (a single case) and Yurakaré (four cases, but non-lexicalised forms). In Yurakaré, the most frequent format of token RfRCs are anaphoric expressions, such as the pro-verb achama ‘is it so?’, with or without further morphological marking. Finally, while all languages employ response tokens as token RfRCs – in the Germanic languages with the meaning of ‘yes’, in Yurakaré a dedicated continuer – this is done in Low German with a much higher relative frequency.

As already mentioned in Section 3.1, Yurakaré is typologically different from the two Germanic languages. Where German and Low German are inflectional languages, Yurakaré exhibits an agglutinative structure. This typological difference may explain the finding that in Yurakaré, ‘token’ RfRCs show a lower degree of conventionalisation and invariance compared to German or Low German. For instance, the German token format echt? ‘really?’ is fully conventionalised, whereas the most frequent token format in Yurakaré, the verb achama ‘it is like that’, is formally variable in that it can take derivational, inflectional, and propositional morphology. The agglutinative structure of Yurakaré may possibly inhibit a conventionalisation of the form achama as an invariant token RfRC format. This typological difference shows that a purely form-based comparative approach is not applicable for these languages.

Table 1
Table 1

Formats of token RfRCs in German, Low German, and Yurakaré

Citation: Contrastive Pragmatics 2024; 10.1163/26660393-bja10097

Figure 2
Figure 2

Relative frequency of RfRCs’ final intonation per language

Citation: Contrastive Pragmatics 2024; 10.1163/26660393-bja10097

Another comparative difference in RfRC formatting is that the two Germanic languages mostly deliver RfRCs with rising intonation, whereas Yurakaré speakers more frequently employ RfRCs with falling intonation, as shown in Figure 2 (see variable final intonation in Section 3.4). The supplementary material reveals a large association between these variables (E=0.43; df=4), which is highly significant (p<0.01) (see supplementary material, p. 2). This is basically comparable to previous findings regarding final intonation in information-seeking and assertive content questions in Yurakaré, both of which show a very low frequency of clearly rising intonation with around 5% only (Gipper, 2022). Moreover, in requests for confirmation, the proportion of rising intonation is also low in Yurakaré when compared to other languages (Pfeiffer et al., in prep.). This suggests that in Yurakaré, final rising intonation is not associated with interrogative actions to the same degree as in German and Low German.

Comparing the sequential integration of RfRCs in Yurakaré, German, and Low German (see variable sequential embedding in Section 3.4), informings proved to be the most prominent preceding action in all languages with about 60% of all cases. In these contexts, the RfRCs refer back to informing actions which are not made relevant by a prior turn (volunteered informings, cf. Thompson et al. 2015) or they refer to turns which first respond to a prior but also offer more than the required answer (elaborated informings).

Concerning preference organisation, RfRCs show a similarly strong tendency toward receiving a confirming response in all three languages with a proportion of roughly 75% (see Figure 3; variable response type in Section 3.4), demonstrating a strong cross-linguistic preference for confirmations and a virtual ban on disconfirmations (also see Marmorstein and Szczepek Reed, this issue).8 The languages differ, however, in the remaining quarter of cases: In German these are mostly made up of instances without any uptake (no response-category). In the other two languages, in addition to cases without any uptake, we found instances where a response is given but is not (fully) oriented toward (dis)confirmation (neither-category). These responses are of very different kinds in the two languages: Whereas in Yurakaré they mostly deal with elaborating the information further, in Low German they often adjust the information given in the turn for which reconfirmation is requested and thus border on OIRs. Especially the language-specific differences in the response types disconfirmation and neither affect a medium association (E=0.16; df=6) between the variables language and response type, which is significant (p<0.01) (see supplementary material, p. 2).

Figure 3
Figure 3

Relative frequency of response type per language

Citation: Contrastive Pragmatics 2024; 10.1163/26660393-bja10097

Comparing the preference organisation of RfRCs visualised in Figure 3 with those of requests for confirmation in the respective languages (cf. Pfeiffer et al., in prep.), an important difference becomes apparent: For all three languages, the percentage of RfRCs that do not receive any response is substantially higher than for requests for confirmation where proportions are lower. This reveals that RfRCs do not mobilise responses to the same degree as requests for confirmation. This finding is consistent with Betz and Gubina’s (2021) observation that RfRCs invite confirmation rather than making it conditionally relevant. It is therefore warranted to treat RfRCs as not merely a subtype of requests for confirmation, but rather as an action in its own right with specific interactional contingencies. Our findings from this section can be summarised as follows:

  • In Yurakaré, RfRCs are used with a much higher normalised frequency than in German and Low German.

  • The three languages use similar RfRC formats, differing, however, in the relative frequency with which these formats are deployed. Yurakaré shows a preference for repeat RfRCs, while in German and Low German token RfRCs prevail.

  • In all languages, RfRCs occur mainly after informings (about 60% of the cases).

  • The three languages show striking similarities regarding the RfRCs’ preference for confirming responses. There are almost no disconfirmations; however, Yurakaré and Low German show a certain proportion of responses that neither confirm nor disconfirm the RfRC.

These observations call for a closer quantitative and qualitative sequential analysis which tries to determine whether different RfRC formats are used to accomplish similar or different social actions in the respective languages. In the following section, we will focus on different response formats in RfRC sequences (ranging from minimal to more elaborate responses) and the responsive actions they implement. This ‘next-turn proof procedure’ (Hutchby and Wooffitt, 1998: 15) can be used as a key to determine the actions speakers ascribe to the RfRC turn.

5 Functions of RfRCs Following Informing Actions

To keep the sequential contexts constant across languages, our analyses in this section concentrate on RfRCs following informings, the most frequent preceding actions to RfRCs in all languages (see Section 4). This ensures that differences in the responses do not result from differences in the RfRCs’ sequential embedding. Note that the responses include both cases of confirmation and of responses that are not oriented toward confirmation (the latter in Low German and Yurakaré, see Figure 3 above). In Section 5.1, we first examine the extent to which different RfRC formats have an impact on the response type in the respective languages. We then investigate whether the prosodic design of the RfRC influences the ensuing response behaviour (Section 5.2). Based on the quantitative findings, we then provide a more detailed sequential analysis of the response types following token RfRCs (Section 5.3) and repeat RfRCs (Section 5.4) in each language.

5.1 RfRC Format and Response Format

Figure 4 demonstrates that the languages differ in the degree and type of functional differentiation between the two RfRC formats based on the responses they solicit. In German, the difference between the two formats is particularly evident: Repeat RfRCs trigger confirmations consisting of a response token and further elaborations in about 80% of the cases.

The statistics in the supplementary material (p. 3) confirms a large association (E=0.47; df=1) between the response format Token + Elaboration and RfRC format in the German data, which is highly significant (p<0.01). What is more, repeat RfRCs in German are never responded to with a minimal response in the form of a response token only. Token RfRCs in German, in contrast, are much more versatile: They can trigger elaborations, minimal token confirmations, and also show a relatively high percentage of cases that receive no uptake at all. The statistical association between the response format no response and RfRC format is, however, only small (E=0.25; df=1) and shows no significance (p=0.23) for the German data (see supplementary material, p. 3). The sequential analysis in Section 5.4 will confirm that repeat RfRCs often work as challenges, which elicit further talk in German, while token RfRCs can be employed flexibly for functions from simply registering news to challenging.

Figure 4