Request for confirmation sequences in British and American English

2.1 Requests for confirmation in BE and AE

Dedicated research on RfCs in English is limited, and much of the existing literature emphasizes their interactional role rather than aspects of their linguistic form. As stated in the introduction, RfCs are essentially a functional subtype of polar (yes/no) questions (Stivers 2010, 2776–7, Stivers and Enfield 2010, Enfield et al. 2019, 288). The grammatical formats that have been described for RfCs are declarative questions, [declarative + tag]-constructions, interrogatives with negative polarity, polar interrogatives, freestanding phrases, partial as well as full other-repetitions (Couper-Kuhlen 2020), and freestanding particles (so-called ‘newsmarks’, Thompson et al. 2015, Jefferson 1981, Heritage 1984). The latter two have been excluded from consideration here. Stivers (2010), following Stivers and Enfield (2010), contrasts confirmation requests with information requests, other-initiations of repair, and assessments. In her analysis of AE question‒response sequences, 21% of all question turns and 30% of all polar questions are RfCs. The clear majority are formatted as syntactic declaratives (77%), and a small number are [declarative + tag]-constructions (14%) or syntactic interrogatives (9%). Similarly, Weber (1993) reports that the majority of declarative questions (63%) in her AE conversational data are RfCs.

One aspect that has received considerable attention from interactional researchers is the epistemic underpinnings of RfCs and the epistemic work that participants accomplish through them. RfCs claim partial access and/or lesser rights to knowledge that belongs within the epistemic domain of the addressee, who is assumed to be authoritatively informed about the referenced state-of-affairs (Schegloff 1996, 180, Stivers et al. 2011, 3, Heritage 2012, Labov and Fanshel 1977). Therefore, by producing an RfC, a speaker proposes a flat, recipient-tilted epistemic gradient between them and their interlocutor (Heritage and Raymond 2012, König and Pfeiffer forthcoming). This also applies to RfCs that are produced as formulations of the upshot or gist of a prior speaker’s talk, the accuracy of which is for the recipient to decide (Heritage and Watson 1979, 1980). Such RfCs may be prefaced by so (Raymond 2004), or they may include other inference markers, such as then (Heritage 2012, Drew 2018). As Bolden (2010) has shown, formulations of others’ talk can also be prefaced with and to indicate that they explicate something on behalf of the co-participant that was inferable from their prior talk but has so far been left unsaid. In cases like these, RfCs serve to manage and/or secure mutual understanding through the intersubjective verification of inferential or presumptive, and thus essentially tentative, understandings that are publicly offered up for (dis)confirmation (see also Heritage 1984, Antaki 2012, cf. Zinken and Küttner 2022).

Another feature of the turn design of polar questions (including RfCs) that has been shown to matter for the epistemic and interactional work they accomplish is their syntactic polarity. First, prior research has shown that negative interrogatives (e.g., Didn’t you/Wouldn’t you say that…) strongly invite affirmative answers, so much so that they are sometimes treated as making agreement (rather than (dis)confirming answers) relevant next and are routinely used for challenging co-participants and their claims or positions (Heritage 2002, see also Keisanen 2007, as well as Küttner and Ehmer 2024).^[2] Second, Heritage and Raymond (2021), using BE and AE data, have argued that a question’s polarity indexes its speaker’s stance toward the (im)probability of the referenced state-of-affairs (see also Robinson 2020b, Pomerantz 1988), with negatively polarized questions conveying the questioner’s stance that the question’s underlying proposition is unlikely and positive polarity questions conveying the questioner’s stance that it is likely. They further argue that taking such a stance is unavoidable (since every question involves a choice between positive/negative polarity) and that it operates above and beyond other features of question design (e.g., its syntactic format).

Interestingly, Couper-Kuhlen (2012) has put forward a similar argument with respect to the intonational design, specifically the final pitch movement (rising vs falling), of polar questions in BE conversation (radio phone-ins). She argues that, within specific conversational activities (e.g., proffering topics, pursuing topics, responding to news, initiating repair), falling vs rising final pitch patterns on polar questions index different degrees of epistemic certainty about the likelihood of the question’s underlying proposition being true, as locally established by verbal and behavioral cues in the context of the question.^[3] Given the relatively shallow epistemic gradient that RfCs establish between the speaker and the recipient, one would therefore perhaps expect RfCs (and RfC-based actions, such as offering candidate understandings or upshot formulations), to exhibit a preference for intonation contours with final falling pitch movements rather than rising ones. On the other hand, Thompson et al. (2015) have observed that, in responses to informings, the difference between rising and falling final intonation appears to be less consequential on (RfC-like) turns that draw out inferences from the preceding informing. In their AE data, this appears to be true for non-repeating phrasal responses (i.e., reformulations and paraphrases; pp. 88–9) as well as clausal ones (pp. 128–34), both of which generally attract further responses that engage with the sequential relevance of (dis)confirmation of the proffered inference as next actions regardless of their final intonation.^[4] In terms of its relevance for response-mobilization (Stivers and Rossano 2010), Thompson et al. (2015) therefore conclude that the action of “drawing an inference from what the informer has just said trumps prosody” (p. 89).

As we will see below, BE and AE exhibit different preferences for the intonational design of RfCs, calling for a more differentiated, and possibly variety-specific, view on the role that rising vs falling pitch contours play (e.g., for the management of epistemic stance) in the two varieties.

2.2 Question tags in BE and AE RfCs

The canonical English tag question involves an anchor clause and an interrogative tag, consisting of an auxiliary and a personal pronoun that is co-referential with the subject of the anchor. The polarity of the tag may either be reversed or held constant relative to the polarity of the anchor (Quirk et al. 1985, 810ff.). In addition, there are so-called ‘invariant’ tag-questions. These include both lexical tags, such as right? or huh?, as well as clausal ones, such as don’t you think? (Quirk et al. 1985, 814) or the British innit?, which has been derived from the canonical question tag isn’t it but seems to be on its way to developing into an invariant tag (Biber et al. 1999, 210).

Algeo (1990, 445–6) identifies ‘confirmatory tags’ amongst other types, a classification maintained by Tottie and Hoffmann (2006, 2009) and Gómez González and Dehé (2020). In their comparison of tags in BE and AE, Tottie and Hoffmann (2006) find that ‘confirmatory tags’ (“the speaker is not sure of what s/he says, wants confirmation,” p. 300) are the most frequent tag type in BE (37%), but only the second most frequent type in AE (30%) after ‘facilitating tags’ (“the speaker is sure of the truth of what s/he says but wants to involve the listener,” p. 301). With regard to polarity, both varieties heavily favor reverse polarity tags with positive polarity in the anchor clause and negative polarity in the question tag (BE: 75%; AE: 69%). However, AE shows a larger proportion of reverse polarity tags with negative-positive polarity (BE: 17%; AE: 27%), whereas in BE constant polarity tags, with both the anchor and the tag featuring positive polarity, are more frequent (BE: 8%; AE: 4%) (p. 289). The authors find no correlation between pragmatic use and polarity type, though. Overall, Tottie and Hoffmann find nine times as many tags in BE than in AE (2,376 vs 455 per million words, p. 287).

In their study of tag questions in BE, Kimps et al. (2014) find that 46% of tag questions are followed by confirmation. In a subset of the data comprising only information-seeking tag questions (in contrast to ‘offers’, ‘commands’, and ‘statements’, pp. 74–5), where “the speaker is a secondary knower who turns to the co-participant as a primary knower of the information s/he is seeking, and they exchange a B-event” (p. 75), 59% are followed by confirmation and 18% by disconfirmation (see also Kimps 2018). The study also confirms that, in BE, positive‒negative polarity is most frequent (67%), followed by negative‒positive (18%) and positive‒positive (15%). Kimps et al. (2014) also investigate intonation and show that 77% of tags are delivered as a separate intonation unit (p. 72). Tags are predominantly produced with falls (68%). Rises make up approximately a fifth of cases (21%). This finding is interesting in relation to Quirk et al.’s (1985) contention that only reverse polarity tags with rising pitch on the tag actually convey the speaker’s uncertainty and invite the “hearer to decide the truth of the proposition in the statement” (p. 811), while reverse polarity tags with falling pitch on the tag are presumed to invite agreement with, or acknowledgment of, a proposition the speaker already knows to be true (i.e., they are said to have the force of exclamations or assertions) (see also Huddleston and Pullum 2002, 894, Holmes 1983).^[5]

In interactional studies, tag questions are generally said to propose a flatter epistemic gradient than polar interrogative questions but a steeper one than declarative questions without tags (Heritage and Raymond 2012, Heritage 2013). As mentioned in Section 1, these studies rarely distinguish systematically between BE and AE. It is remarkable, however, that, more often than not, particularly clear examples to illustrate the use of tag-questions as features of turn design appear to come from BE data sets. As Heritage and Raymond (2005) speculate in a footnote (in relation to the design of assessments and their responses), the use of tags “as a means to downgrade epistemic claims may be much more prominent among speakers of British English than among their U.S. counterparts” (p. 25, fn. 8). While this may be true when tags are used to invite agreement with claims, assertions, and assessments (Hepburn and Potter 2011), it is unclear whether regional variability is also a factor when tags are used to construct RfCs.

2.3 Response tokens and other response types in BE and AE

As their name suggests, RfCs make confirmation/disconfirmation relevant as next actions. These actions can be implemented with a variety of response formats. Chief among them are, of course, the response tokens yes/yeah (or phonetic variants thereof, such as yep, yah, yo, yay) and no (or phonetic variants thereof, such as nope or nay).^[6] These have been referred to as ‘type-conforming responses’ (Raymond 2003, 2013, Heritage and Raymond 2012), because they do fulfill not only the action-based sequential but also the grammatical constraints imposed by (interrogatively formatted) polar questions. Similarly, when used to respond to requests for confirmation (and other polar questions), non-lexical forms (such as uh huh, hm hm, mhm, but also the negative ʔhmʔhm or ʔuhʔuh) can likewise be used to implement (dis)confirming actions (Schegloff 1982, Raymond 2003). It has been argued that – unless they are prosodically marked (Raymond 2010) – these yes/no-type responses represent the default, pragmatically unmarked way of answering polar questions (Raymond 2003, Stivers 2010, 2019, 2022, Heritage and Raymond 2012). They satisfy the sequential relevancies and pragmatic constraints established by the question, fully accept the terms and the design of the question, and align with the interpersonal asymmetries established by the very act of asking a polar question (Heritage and Raymond 2012, Stivers 2019, 2022, Enfield et al. 2019). This is also said to hold for their embodied equivalents, namely head nods and lateral head shakes (Schegloff 1987, Kendon 2002, Stivers 2019, Enfield et al. 2019, but see Deppermann et al. 2024).

Corresponding to their ‘unmarked’ status, yes/no-type responses represent the most frequent way of responding to polar questions and overwhelmingly occur as the first item in a response (Raymond 2003, Stivers 2010, 2779, Stivers and Hayashi 2010, 7, Heritage and Raymond 2012, 182, Stivers 2022, 93–4). All of the cited studies found them to be used somewhere between 75 and 80% of the time. Enfield et al. (2019) report that their proportion appears to be even greater in response to polar questions that implement RfCs than in response to polar questions that request new information (see also Stivers 2022, 150), which they attribute to the fact that RfCs tend to occur in ‘subordinate lines of action’ and that – being pragmatically unmarked and acquiescent – (minimal) yes/no-type responses maximally promote progressivity and facilitate a quick return to the main line of action (Enfield et al. 2019, 290–1).

In addition to these unmarked yes/no-type responses, English offers its speakers a range of other response tokens for confirmation. These appear to be pragmatically marked and are generally understood as “doing more than just confirming” (Stivers 2019, 197). Stivers (2019, 2022) distinguishes between what she calls (a) upgraded interjections (of course, certainly, absolutely, definitely, totally, right), which propose a problem with the asking of a question (see also Stivers 2011),^[7] (b) downgraded interjections (probably, possibly, maybe), which propose a problem with answering the question as put, and (c) acquiescent interjections (alright, sure, okay, and fine), which propose a problem with the question’s action agenda. Corresponding to their pragmatic markedness, these response tokens are relatively infrequent, being used to confirm polar questions only about 5% of the time in a mixed sample of BE and AE (Stivers 2019, 2022, see also Stivers 2010 for similar findings based on a smaller sample of AE data).

However, at least with respect to the response token right, prior research suggests differences in use between BE/Australian English and AE, specifically with regard to the particular epistemic stances conveyed with it in the two varieties (Gardner 2001, 46–52, 2007, Bolden et al. 2023). While, in AE, right appears to convey a knowing stance and can thus be used to confirm requests for confirmation, this use is remarkably absent (or at least uncommon) in BE data, where it is typically used as an information receipt that conveys a speaker’s recognition of the information as relevant to the ongoing activity (Bolden et al. 2023, see also Gardner 2007).

With respect to its differential distribution across sequential contexts in the two major varieties of English, the opposite has been observed for right’s syntactically clausal cousin that’s right, a form that is widely believed to virtually lend itself to doing confirmation (Schegloff 1996, 175). While that’s right has indeed been shown to be used quite regularly as a confirmation token in BE/Australian English (Gardner 2007, Barnes 2011a,b, 2012), this confirmatory use appears to be much less common in AE (Küttner 2016). This suggests that there is some regional variation between BE/Australian English and AE with respect to their inventories of confirmation tokens, or at least the confirmatory use of these different forms (see also Jefferson 2002, McCarthy 2002, O’Keeffe and Adolphs 2008 for similar observations with regard to other kinds of response tokens).

Another format for doing (dis)confirmation in English is repetitional responses. Compared to token-based responses, these are generally pragmatically marked in English: Using repeats to (dis)confirm polar questions enables speakers to exert some degree of (thematic) agency over the proposition expressed in the question and to thereby assert some prior or independent claim to ‘ownership’ of it (Heritage and Raymond 2012, Enfield and Sidnell 2015, see also Schegloff 1996, Stivers 2005, Heritage and Raymond 2005, as well as Enfield et al. 2019). Repetitional responses are used much less frequently to answer polar questions than yes/no-type responses. In fact, they appear to be the least frequent response type in English, being used only somewhere between 5 and 10% of the time (Stivers 2010, 2779, Stivers and Hayashi 2010, 7, Stivers 2022, 123).

Combinations of response tokens and repetitions (e.g., Yes, he is or He is, yes) are fairly uncommon as answers to polar questions, occurring about 10–15% of the time, depending on the sampling procedure being used (Stivers 2010, 2779, 2019, 195, 2022, 94, 123, 138). And although such composite responses are apparently less frequent in English than one might intuitively expect, they are relatively common when English is placed in the context of other languages (Enfield et al. 2019, 289 who report them to occur less than 4% of the time across all languages).^[8] Moreover, at least as far as confirmations are concerned, neither token-responses nor repetitions tend to attract turn-prefaces very much. Stivers (2022) reports that this only happens in about 6% of the cases in each group. However, when they are prefaced, token-based confirmations appear to exhibit a strong tendency (77% of the time) to be prefaced with oh (Stivers 2022, 94, see also Heritage 1998), whereas repetitional confirmations exhibit no such collocational preference (Stivers 2022, 123).

In addition to these different formats for (dis)confirming, there are also responses that do not (dis)confirm directly, but may be understood to do so indirectly (or to confirm ‘as best as possible’).^[9] While forming a very heterogeneous class of linguistic and interactional objects, such responses typically take issue with features of the question as put, for instance with aspects of its design (the wording of the question and thus the confirmable proposition), its underlying presuppositions, or its agenda (Stivers and Hayashi 2010, Walker et al. 2011, Pomerantz 2017, Stivers 2022). Such responses typically operate back on the question, either by adjusting its terms or by engaging with its agenda (e.g., by speaking to its inferred underlying purpose). Because they therefore essentially answer a different question than had originally been asked, they have been collectively referred to as ‘transformative’ or ‘transformational’ answers (Stivers and Hayashi 2010, Stivers 2022). Being used to answer polar questions about 12% of the time, they have been found to be slightly more common than repetitional answers in ordinary conversation (Stivers and Hayashi 2010, 6, Stivers 2022, 149). Stivers (2022) also reports that they exhibit a stronger tendency to be prefaced than other answer types – most commonly with well (see also Heritage 2015) – and that they are more likely to be produced in response to polar questions that are asked from a weaker, less presumptive knowledge base, i.e., in response to requests for information (59%) rather than requests for confirmation (25% [p. 150]).

Finally, there are non-answer responses with which respondents go on-record as either not being able to or (less commonly) refusing to answer the question (Stivers 2022, chap. 3). Respondents typically go to great lengths to answer their co-participants’ questions (as best as possible) and so non-answer responses have been found to be fairly infrequent (about 15% of polar questions in Stivers’s 2010 AE sample received non-answer responses). Inasmuch as non-answer responses still engage with the sequential relevancies established by the question that has been asked, they are different from fully non-responsive behavior, which has been found to be even less frequent (about 5% in Stivers’s 2010 sample).

Overall, then, polar questions receive answers about 80% of the time, with affirmations/confirmations outnumbering disaffirmations/disconfirmations at a ratio of about 3:1 (Stivers 2019, 194, 2022, 24–5). Stivers (2010) reports that, among those polar questions that received an answer in her AE sample, 72% were met with confirmations, whereas only 28% ended up being disconfirmed (p. 2779). It has been argued that these frequencies and distributions reflect a multi-layered underlying organization of preference for responsive behavior to polar questions in English (Stivers 2010, 2022, on the notion of preference, see Pomerantz and Heritage 2013), with

• answers being preferred over non-answers (Stivers and Robinson 2006),

• confirmations being preferred over other answer-types, namely disconfirmations and transformative answers (Stivers and Hayashi 2010),^[10]

• and simple, unmarked yes/no-type responses being preferred over other response formats for doing confirmation (Stivers 2019, 2022, see also Enfield et al. 2019).

4.1 Syntactic design

Table 1 provides an overview of the syntactic formats of the RfCs in our collections. More than half of the RfCs in both varieties are formatted as syntactic declaratives (BE: 110/182, 60.4%; AE: 115/208, 55.3%). Over a quarter of all declaratively formatted RfCs include a canonical clausal or lexical tag (BE: 33/110, 30%; AE: 32/115, 27.8%) (Section 4.6). Roughly two-fifths of RfCs (BE: 77/182, 42.3%; AE: 83/208, 39.9%), therefore, come in the form of plain declarative clauses. The other most common formats are syntactic interrogatives (BE: 35/182, 19.2%; AE: 33/208, 15.9%) and various kinds of phrases (BE: 34/182, 18.7%; AE: 48/208, 23.1%). While there is some minor variability (<5%) in terms of the relative frequency of interrogative and phrasal RfCs between our samples, taken together, they account for nearly another 40% of all RfCs in our respective collections (BE: 38.5%; AE: 39%). These distributions roughly align with those observed for polar questions in general in AE, but they are less heavily skewed toward declaratives as the main format for doing confirmation requests (Stivers 2010, 2776, Table 3). While declaratives (with and without tags) are also the main syntactic format for bringing off RfCs in our samples, interrogatively formatted RfCs do not appear to be as uncommon in English as previously suggested in the literature.

As far as phrasal RfCs are concerned, Table 2 shows that they are heavily skewed by type, with noun phrases and prepositional phrases far outweighing other syntactic phrase types.

Most of ^[11]these are produced as candidate understandings or understanding checks in other-initiated repair sequences (Schegloff et al. 1977, Schegloff 1997, Antaki 2012, Kendrick 2015). Their syntactic design as (non-repetitional) phrases minimizes the disruption to sequential progressivity that such other-initiations of repair embody and makes them recognizable as lexical replacements or paraphrases, insertions, or continuations of the prior turn (Kendrick 2015, 174–6). The following extract from the AE collection illustrates the use of a phrasal RfC to initiate repair. Here, Matthew reports what his statistics professor replied to an e-mail he sent, asking about the details of his grade.

(2) CH-En-6521 (AE #186)

In line 07, and at some distance from the trouble source, Joheved produces the prepositional phrase at the PAper? with a final pitch movement rising to high to initiate repair on Matthew’s prior turn. With its syntactic formatting as a prepositional phrase, this candidate understanding is recognizable as a possible replacement of Matthew’s on your PROject (line 03). The candidate replacement substitutes Matthew’s original reference with a form that has the status of a recognitional reference for Joheved (note the definiteness in the PAper) and works to clarify its referent. Once Matthew has minimally confirmed this candidate understanding as adequate (line 08), both participants orient toward the resumption of his telling as a relevant next action (Joheved by re-aligning as a recipient through her uh_huh in line 10 and Matthew by starting to resume his telling with a: nd_uh in line 11).

In addition, phrasal RfCs are also common as candidate answers to turns that are initiated with open questions/wh-interrogatives, both in the context of repair (e.g., WHO.=ROB? (AE #17)) and in non-repair environments (see extract (4)).

4.2 Polarity

A clear majority of RfCs in both collections exhibit positive polarity (BE: 167/182, 91.8%; AE: 161/208, 77.4%, Table 3). This is unsurprising and seems to bespeak their presumptive character: With RfCs speakers generally seek confirmation for what they presume to likely be the case (Heritage and Raymond 2021, see also Pomerantz 1988, Robinson 2020b), rather than what they presume not to be the case (unless, of course, they have reasons to believe that something is actually not the case or did not happen).

Curiously, however, there appears to be a rather strong disparity between our samples with respect to the proportions of RfCs with positive vs negative polarity. Whereas the BE sample shows a ratio of more than 9:1, the AE sample shows a ratio of only about 3:1. In fact, the BE and the AE samples appear to fall on opposite ends of the spectrum of all ten languages considered in the project (Pfeiffer et al. forthcoming). Given that the two samples correspond rather closely in most other categories, this difference is remarkable and somewhat puzzling. One possible source of this divergence lies in the diversity of interaction types and the associated activity types represented in our two samples. The BE collection was sampled from a comparatively homogeneous corpus of dinner table conversations amongst young people in their early twenties who are well-acquainted peer students or romantic partners. In contrast, the AE collection holds a more diverse range of face-to-face and telephone interactions amongst a wider range of social relationships. A number of them involve participants who have not seen or heard each other in a while, such that activities like soliciting updates about things that did or did not happen (e.g., how are YOU doing;=you didn’t (0.5) End up going to I:taly:– or you didn’t get HOMEwork done though.=did you?) as well as joint remembering or recalling/re-activating information presumed to be in the common ground (e.g., the SWEdish au pair;=didn’t she DO that?) are more prevalent in the AE data than in the BE data, which may contribute to the higher proportion of negatively polarized RfCs in AE. Another possible, though considerably more speculative explanation concerns different politeness practices in the two varieties (Brown and Levinson 1987). As has been shown, negatively polarized RfCs can also be deployed as vehicles for bringing off interactional challenges (Keisanen 2007), and this may simply be something the participants in the BE sample do less often than those in the AE sample. Note in this respect that negative interrogatives – a common format used for challenging actions (Heritage 2002, Keisanen 2007) – are somewhat less common in BE (4/36, 11.1% of all interrogatively formatted RfCs) than in AE (7/33, 21.2% of all interrogatively formatted RfCs).

Unsurprisingly, negative polarity is overwhelmingly accomplished by means of the negation particle not, while other means of negation, such as negative determiners (no X), negative adverbs (never), or negative indefinite pronouns (nothing), are highly uncommon.

4.3 Modulation

Given that about three-quarters of the RfCs in our samples are formatted as syntactic declaratives or phrases (BE: 78.6%; AE: 78.4%, Table 1), one would expect epistemic modulations, for example with devices such as approximators, epistemic adverbs, epistemically hedging mental predicate phrases like I think/I guess, as well as some modal verbs, to play an important role in making these turns recognizable as proffering ‘confirmable’ (i.e., uncertain, tentative, presumptive) candidates, and to mobilize a (dis)confirming response from the co-participant (see also Weber 1993, Stivers 2010, Stivers and Rossano 2010).

As Table 4 shows, however, epistemic modulation is less common in RfCs than one might intuitively expect. Less than one-fifth of all RfCs in our collections feature some kind of epistemic modulation (BE: 35/182, 19.2%; AE: 39/208, 18.7%). In the BE collection, just over two-thirds of these modulated RfCs consist of syntactic declaratives (24/35, 68.6%), whereas in the AE collection, only about half of them do (20/39, 51.3%). Overall, then, epistemically modulated declaratives merely account for around 10% of all the RfCs in our respective collections (BE: 13.2%; AE: 9.6%).

Table 5 summarizes the frequencies of different modulation markers and their distributions across the two varieties. Although the two samples show some minor variability in terms of the specific forms being used for epistemic modulation, it is striking that approximators and epistemic adverbs account for the clear majority of markers in both varieties (BE: 29/41, 70.7%; AE: 24/47, 51.1%). Epistemic uses of mental predicate phrases appear to play a less central role as a mark-up strategy for RfCs (BE: 2/41, 4.9%; AE: 5/47, 10.6%), suggesting that their epistemic stance-marking function is primarily put to different interactional uses (e.g., Kärkkäinen 2006, 2007, 2012). Equally uncommon are epistemic modulations with modal auxiliaries, as are hearsay markings (e.g., via quotatives such as Deb said…, My Dad said…). Because of their strong bias toward affirmative responses and their associated capacity to convey a greater degree of certainty about the likelihood of the confirmable proposition being true (see especially Heritage 2002), negative interrogatives (e.g., dIdn’t you use to SMOKE?, the SWEdish au pair;=didn’t she DO that?) were also coded as a type of modulation. As noted above, they are slightly less common in the BE sample than in the AE sample, where they actually constitute the second most frequent modulation type.

The approximator like appears to be particularly common as a modulation marker in RfCs, accounting for over half of all markers in the BE sample (20/41, 48.8%) and about a quarter of all markers in the AE sample (12/47, 25.5%). The following extract from the BE collection shows such a modulation with like in line 09. The participants are talking about Felicity’s mother. Prior to the transcribed excerpt, Felicity mentioned that her mother is dating someone from ‘a Yorkshire farm’.

(3) Recording 9 Carolyn and Felicity (BE #91)

Interestingly, in both varieties, the approximator like appears to be particularly prominent as a modulator in RfCs that are not formatted as declarative clauses. In one fairly recurrent use, it combines with time or place references as well as other kinds of NPs, when they are offered as candidate answers to open questions (e.g., whEre is THAT;=like bUck’s COUNty or sOmething? (AE #139)). The following extract from the AE collection illustrates this use. Following a display of schadenfreude from Marie that Jodie has a class at 5 o’clock in the afternoon, to which Jodie has responded with an other-correction, stating that it is an ‘extra credit class thing’ rather than a regular class, Jodie now elaborates on what it entails.

(4) CH-En-6474 (AE #200)

In lines 04‒05, Marie responds to the informing that Jodie is going to participate in a psychology experiment with an RfC in the [open question + like phrasal candidate answer]-format: what IS it;=like sEx ED?. While this is recognizably offered as a non-serious candidate, deployed to tease Jodie (note Marie’s chuckle in line 07 and Jodie’s shift to smile voice during her response), it nonetheless receives a disconfirming response and a correction from Jodie (Pomerantz 1988) as a way of disattending the tease (cf. Drew 1987).

In connection with our previous observations concerning the syntactic design and polarity of RfCs, it is noteworthy that RfCs are very rarely modulated in our collections when they are syntactically formatted as positively polarized interrogatives. This suggests that ‘standard’ syntactic interrogativity and epistemic modulation constitute somewhat complementary turn-constructional resources for designing RfCs.

4.4 Inference marking

RfCs are commonly used to solicit confirmation for the accuracy of inferences speakers have drawn from prior talk (Heritage 1984, Bolden 2010, Antaki 2012, see also Zinken and Küttner 2022), and their inferential character can be signaled linguistically. Inference-marking is present in about 25% of the cases in each of our samples (Table 6).

As Table 7 shows, so is by far the most frequent single inference marker used in RfCs (see extract (3) for a case in point), accounting for more than half of all inference markers in both varieties (BE: 29/50, 58%; AE: 29/50, 58%). It also figures prominently when multiple inference markers are strung together to form more complex combinations. Change-of-state tokens like oh or ah (Heritage 1984) are the second most frequent inference marker in both varieties (if combinations with other markers are included), though they seem to be fairly more common in the BE sample than in the AE sample.^[12] Other inference markers are either relatively infrequent or entirely absent from our samples. In light of the attention then has received as a prototypical inference marker in the literature (Heritage 2012, Drew 2018), this may be somewhat surprising. Across our two collections, we only found four instances of its use, three of which happened to include so as a co-occurring inference marker (one composite so then…, two instances of combined use with so in RfCs that feature both as independent markers, i.e., so…then) and only one in which then was used as the sole inference marking device.

4.5 Connectives

Connectives are turn/TCU-initial items that indicate the RfC’s sequential and/or semantic-pragmatic relation to preceding utterances. This category therefore includes turn/TCU-initial inference markers, which have been double-coded and reappear in Table 9, but also (non-inference marking) conjunctions and other kinds of turn/TCU-initial sequential or pragmatic markers (e.g., wait or in other words …). Consequently, the absolute and relative frequencies for the use of connectives in RfCs are higher than, and almost double, those for inference marking (BE: 72/182, 39.6%; AE: 86/208, 41.3%, Table 8). An overwhelming majority occur with clausal RfCs (BE: 62/72, 86.1%; AE: 76/86, 88.4%).

As far as different kinds of connectives and their relative distributions are concerned, the AE and BE samples behave similarly (Table 9).

So accounts for a little more than a third of all connectives (BE: 36.1%; AE: 37.2%) and is by far the most frequent one. The vast majority of these are inference- or upshot-marking uses of so (Section 4.4, see also Heritage and Watson 1979, 1980, Raymond 2004), testifying to the central role RfCs play for the intersubjective verification of inferential understandings in the management of mutual understanding (Heritage 1984, Zinken and Küttner 2022). In addition, so is sometimes (though considerably less often) used with topic-proffering or resumptive RfCs, to mark their status as items that have been on the speaker’s agenda or previously been put ‘on hold’ (Bolden 2006, 2008, 2009).

The second most frequent connective in both our samples is and (BE: 13/72, 18.1%; AE: 14/86, 16.3%), suggesting that (clausal) RfCs are quite regularly used to continue a previous line of talk, either by explicating something that was inferable but previously left unsaid (Bolden 2010), or by following up on a specific aspect or theme mentioned in, or related to the current business of, the co-participants’ talk. This raises questions about the claim that RfCs are primarily used in subordinate lines of action (Enfield et al. 2019, 290–1). While this appears to hold for a great majority of phrasal RfCs, which are commonly implicated in (other-initiated) repair (Kendrick 2015, see Section 4.1) and sequentially instantiate insertion sequences (Schegloff 2007, 100–6), thus making them recognizable as ‘subordinate’ lines of action qua their sequential placement, it seems less clear for clausal RfCs, especially when they are formally marked as relating to or following from, as explicating, elaborating, or continuing a previous line of talk (see also Section 5.4). In the following example from the BE collection, for instance, the causal connective (be)cause (line 14) prefaces a clausal RfC that introduces a new referent (Katie) and proffers a candidate observation about her working in the same place as Carol (‘she’ in lines 01, 03, and 09) as an account for the inquiry in line 05. This matter is then elaborated further (line 17), shifting the focus of the talk from Carol to Katie.

(5) Recording 5 Amy and Hannah (BE #48)

4.6 Tags

The vast majority of RfCs are designed without question tags in both varieties. Tags are used in only about one-fifth of our cases (BE: 34/182, 18.7%; AE: 42/208, 20.2%, Table 10). However, compared to the 6% tags Stivers (2010) reports to have found among polar questions more generally, their proportion is about three times as high. This suggests that tags do play a special role when designing RfCs.^[13] As mentioned in Section 4.1, where tags occur, they most often combine with declaratives. In the BE collection, 91.2% of tags follow declarative confirmables (31/34); in the AE collection, 76.2% follow declaratives (32/42), with most others attached to phrases (e.g., no more SNOW;=huh? (AE #125)).

Overall, our data do not support the idea that the use of tags as a turn-constructional resource for RfCs is more widespread in one or the other variety.^[14] However, the collections vary in the type of tags that are being used (Table 11). The BE corpus holds a clear majority of canonical tag questions (29/34, 85.3%), with only five invariant tags (4× right, 1× or). In contrast, only just over a third of tags in the AE corpus are canonical tag questions (15/42, 35.7%). The majority (24/42, 57.1%), though clearly less than the 77% share reported by Stivers (2010), are invariant tags, most commonly right (17/42, 40.5%). The AE collection also contains four cases of huh, which is not present in the BE collection.

One of the few instances of right in the BE collection is the following, in which Felicity and Carol are talking about a friend of Felicity’s. As it happens, this friend is half-American.

(6) Recording 9 Carolyn and Felicity (BE #85)

4.7 Prosodic design

It is in the intonational design of RfCs where the two collections show the most significant variation. While intonation contours with final falling pitch are by far the most frequent pattern for confirmables in BE (166/182, 91.2%), in AE intonation contours with final rising pitch are dominant (127/208, 61.1%). Moreover, about half of the instances with final falling intonation on the confirmable in the AE sample have tags attached to them (34/73, 46%), the clear majority of which are done with a separate rising contour (27/34, 79%). These turns therefore also end in a final rise, adding another 12% to the overall share of AE RfCs that end with final rising pitch (raising it to about 73%) and further increasing the disparity between the BE and the AE sample (Table 12).

This strong disparity between the two varieties, with BE’s strong preference for falls and AE’s somewhat less uniform preference for rises, is reminiscent of similar differences in intonational design that Couper-Kuhlen (2020) has observed in other-repetitions in AE and BE. She, too, found that “rising contours appear more frequently overall in North American English than in British English other-repetitions[,… whereas] falling contours are more frequent overall in British English than in North American English other-repetitions” (p. 547). Notably, these distinct preferences are particularly clear for the following two other-repetition actions: (1) seeking clarification or specification for a problem of hearing or understanding and (2) seeking (dis)confirmation from a position of minimal epistemic uncertainty (pp. 527–35). While other-repetitions have been excluded from the present investigation, we can note that these other-repetition actions functionally overlap with some of those of the non-repetitional RfCs considered here.^[15]

A similar difference can be observed in the intonational design of tags in the two varieties. As Tables 13 and 14 show, tags are mostly not prosodically integrated with the confirmable proposition and have their own, independent pitch contour (BE: 27/34, 79.4%; AE: 39/42, 92.9%). This separate pitch contour is mostly falling in BE (21/27, 77.8%) but mostly rising in AE (31/39, 79.5%).

These tendencies allow for the identification of variety-specific canonical intonation patterns for [confirmable + tag]-formatted RfCs: Cases from BE predominantly consist of declarative confirmables with final falling pitch, followed by a pitch reset and an added tag with another falling pitch contour, whereas comparable AE RfCs predominantly consist of declarative confirmables with final falling pitch, followed by a tag with a separate rising contour.^[16] The following cases exemplify these different canonical intonation patterns. Extract (7) from the BE collection shows a declarative confirmable in line 4, delivered with falling final pitch, followed by a non-integrated tag which also has falling pitch (line 5).

(7) Recording 3 Melissa and Zack (BE #18)

Extract (8) from the AE sample shows a series of three RfCs produced by the same speaker (Lydia) and addressed to the same recipient (Pablo) in lines 02, 06, and 14. All three of them have falling pitch on the confirmable element, one of which is phrasal (line 02), the other two of which are clausal (lines 06 and 14). Each of them is followed by a non-integrated right-tag with a separate rising contour.

(8) P-Des1a (AE #63-65)

(Lydia has just asked Pablo how his practice is going. Placenames have been changed.)

Overall then, the two varieties exhibit distinct preferences for the intonational design of RfCs. But how are we to make sense of them? Specifically, how can they be reconciled with the idea that the intonational design of polar questions is systematically linked to the speaker’s degree of epistemic certainty about the likelihood of the underlying proposition being true (with falls indexing a greater degree of certainty than rises, see Couper-Kuhlen 2012, Holmes 1983)?

In this connection, it is remarkable that our AE and BE samples converge rather clearly in their preference for falls in the case of upshot formulations. These tend to be positioned late sequentially (toward the end of the sequences whose upshot they formulate) and therefore commonly allow its speakers to claim a relatively high degree of epistemic certainty about their likely adequacy as upshots of the preceding sequence. It is here, in this ‘high certainty’ context, that the AE sample also exhibits a relatively clear preference for falling intonation contours on RfCs (unless they convey additional affective stances like surprise or incredulity). Example (9) from an AE dinner conversation is a case in point. It shows the end of an extended sequence in which Lara, Soren, and Dianne (Lara’s mother-in-law) are jointly trying to figure out which year Soren last went back to Denmark for Christmas to see his grandchildren. In lines 01‒08, Lara offers another public recollection in the service of this endeavor. Her observation that 2 years ago she and her family had gone to Florida and that the year before Lydia (Soren’s partner) was paying them a Christmas visit by herself, which Soren emphatically seconds in lines 09 and 11, strongly suggests that it must have been this year that Soren had last gone back to Denmark. This is what Dianne formulates as a concluding upshot of their discussion in lines 12–13.

(9) P-DinMain1a (AE #71)

Given the time references in Lara’s reconstructive recount, it is extremely likely that Dianne’s upshot formulation will turn out to be correct, and the final falling pitch with which it is delivered can be seen to betoken a corresponding epistemic stance, viz. a high degree of confidence in, or certainty about, its likely correctness. Cases like these would seem to lend further support to, and speak against a wholesale dismissal of, the idea that the intonational design of an RfC can convey its speaker’s epistemic stance toward the likelihood of the truth of the confirmable proposition. But in light of the observed regional variability in intonational design between our samples, it does not seem to hold up equally well in many other contexts of RfC use, where AE and BE preferences for intonational design clearly diverge. Consider the following sets of somewhat parallel examples from the AE and BE collections.

In this first BE case, Amy is telling Hannah about her youngest sister’s upcoming birthday party, to which the sister has invited two boys she likes.

(10a) Recording 5 Amy and Hannah (BE #35)

In a rather similar case from the AE collection, Valerie recounts how she recently met up with her ex-partner Cindy and Cindy’s new girlfriend Penny.

(10b) CH-En-6625 (AE #183)

In both of these examples, a speaker offers an inferential understanding derived from their co-participant’s prior reporting up for (dis)confirmation. Hannah in extract (10a) infers from Amy’s report of her younger sister being into two guys at the minute (lines 01‒02) that she is seeing both of them (line 07), and Lisa in extract (10b) infers from Valerie’s scene-setting description which mentions that both her parents and Cindy were present at the mall in Ashville (lines 01 and 03) that Valerie’s parents actually met Cindy as well (line 07). Both of these inferential understandings target the meaning of potentially ambiguous descriptive items in the preceding reports: Amy’s formulation of her sister being into two guys at the minute in extract (10a), and the import of Valerie’s mentioning of her parents’ simultaneous co-presence at the mall where she met Cindy in extract (10b), i.e., of them being there . So both of these inferential understandings appear to be licensed by the preceding reports, though interestingly both of them end up getting disconfirmed and corrected (lines 08‒09 in extract (10a), lines 09‒10, 12–14 in extract (10b)). From what we can see in the data, there is thus no reason to believe that one of them was uttered from a position of greater or lesser epistemic certainty about its likely correctness than the other. Still, British Hannah’s RfC is done with a step-up in pitch on the accented syllable (BOTH) and a subsequent falling intonation contour, whereas American Lisa’s RfC is done with a step-down in pitch on the accented syllable (ALso) and a subsequent rising intonation contour.

Something similar can be observed in the following two cases. In contrast to the preceding two examples, where the RfCs were syntactically formatted as declaratives, the following two instances feature RfCs formatted as negative interrogatives. In the BE case, Dan is telling Kat about one of his teachers at school who completed a doctoral degree while he was her student. He relates how he and his classmates interpreted her previous use of the honorific title ‘Ms’ (transcribed as miz in lines 03, 05, 15, and 16 of Extract (11a)). Dan’s talk contains a homophobic reference delivered in the form of reported speech of his younger self in line 08.

(11a) Recording 6 Kat and Dan (BE #56)

In a similar case from the AE collection, Beth and John are hosting Anne and Don for a joint meal, when Beth begins to solicit an update about Anne and Don’s phone situation.

(11b) Chinese Dinner (AE #26)

In both cases, the speakers who end up producing the RfC produce prior turns that exhibit their current understandings and beliefs about the matter at hand. Kat’s div- (.) diVO:RCED in line 13 of extract (11a) displays her current belief of what the form of address ‘Ms’ refers to, whereas Beth’s question did you ever getcher PHONE fixed (line 01 in extract (11b)) presupposes that there have previously been issues with Anne and Don’s phone. In both cases, these turns do not receive immediate uptake (note the 0.7 s silence in line 14 in extract (11a) and the 0.3 s silence in line 02 in extract (11b)), whereupon the same speakers put those very understandings up for (dis)confirmation (line 15 in extract (11a), lines 03 and 05 in extract (11b)). Matching their previous displays of what their current beliefs and understandings are, both speakers syntactically format their RfCs as negative interrogatives, a format that is strongly biased toward receiving confirmation (Heritage 2002, Bolinger 1957). Both RfCs can therefore be understood to be issued from similar positions of epistemic certainty about the truth or correctness of the confirmable propositions. If anything, unless Beth misremembers or confuses Don and Anne with somebody else, Kat’s RfC in extract (11a) is slightly more prone to error than Beth’s in extract (11b). And yet, it is Kat’s RfC which is done with a falling intonation contour, while Beth’s is done with two rising contours. Note also that Kat’s RfC ends up getting disconfirmed and corrected (lines 16–19, 21 in extract (11a)), whereas Beth’s ends up getting confirmed before her original question is answered with the informing that the phone has been fixed (lines 07, 09, and 14 in extract (11b)). Here too, then, RfCs that are produced in comparable sequential environments are being delivered with different final intonation contours in AE and BE, respectively.

So if we are to retain the idea that the intonational design of RfCs is somehow bound up with its speaker’s epistemic (un)certainty about the likelihood of the truth of the underlying proposition, then these rather distinct preferences in AE and BE suggest that, in many (or at least some) contexts, the two varieties have routinized distinct pitch patterns for the intonational packaging of RfCs and the marking of epistemic stance. Two alternative possibilities appear plausible here. The first would account for the observable differences in intonational packaging between the two varieties in terms of distinct, variety-specific design logics for the intonational packaging of RfCs. BE, with its strong preference for falling intonation, could be understood to more strongly foreground the speaker’s relative certainty about the RfC’s proposition and the shallowness of the epistemic gradient between the RfC-speaker and the would-be confirmer. AE, on the other hand, with its overall preference for rises, could be understood to more strongly foreground an element of uncertainty or tentativeness in the intonational design of RfCs. The other plausible possibility is that the two varieties draw on different intonational means for marking epistemic (un)certainty in RfCs. While AE may be seen to primarily draw on final pitch movements as described in the existing literature (with falling final pitch movements indexing greater epistemic certainty than rising ones), BE may primarily rely on other intonational means to accomplish this (e.g., high pitch elsewhere in the turn, as has been suggested by Couper-Kuhlen 2020 with respect to similar regional variability in the prosodic design of other-repetitions).^[17]

Future research will have to assess these proposals and a more fine-grained, action-type specific analysis is needed to unpack the role of final pitch (as well as its possible relation to other intonational means) in the two varieties. At any rate, the observed differences in intonational design between AE and BE call for more caution in data choice, especially when prosodic phenomena are investigated, and a greater sensitivity for variational differences between AE and BE when claims about the interactional function(s) of prosodic design features are being made.

4.8 Interim summary: Resources for designing RfCs in BE and AE

In summary, common features of RfCs in BE and AE are declarative syntax, positive polarity, no connective or question tag, and no modulation or inference marking. A significant difference between the two varieties exists in the pitch configuration of the confirmable and of potential question tags. For both components, final falling pitch is the dominant pattern in BE, and final rising pitch is most frequent in the AE collection. While these features show clear distributional tendencies if looked at individually, very few RfCs show all of them together (BE: 20/182, 10.1%; AE: 13/208, 6.25%). So the most common features in quantitative terms constitute an aggregate that is rarely ever found in this exact combination in actual instances in our collections (Schegloff 2009) and which, therefore, clearly cannot be considered prototypical for RfCs in either variety. In fact, our results suggest that there is no single most common turn design for RfCs. Instead, speakers evidently combine these features variably and flexibly to fit their RfCs to the local action context. Still, the epistemic gradient between the speaker and the would-be confirmer is given expression in most RfC designs. That is to say, despite the individual minority occurrences of modulation, inference marking, tags, and (syntactic or prosodic) interrogative marking, most RfCs convey the speaker’s epistemic uncertainty in one way or another. There are only 18 instances in each collection (BE: 9.9%; AE: 8.7%) that contain none of these features. Overall, our results show that speakers have an array of turn-constructional resources for building RfCs at their disposal and can flexibly combine them in multifarious ways. Moreover, while AE and BE RfCs exhibit many cross-varietal commonalities, the analyses also revealed differences between the two varieties, especially with regard to the formal inventories of tags that are being used and the intonational design of RfCs.

5.1 Responsive actions

The overwhelming majority of RfCs receive a response, both in the BE (180/182, 98.9%) and in the AE (203/208, 97.6%) collection (Table 15). Fully non-responsive behavior is even less frequent following RfCs than after polar questions more generally (BE: 2/182, 1.1%; AE: 5/208, 2.4% vs 5% of non-responses to AE polar questions reported in Stivers 2010, 2778), suggesting that RfCs are particularly ‘successful’ in mobilizing responses (Stivers and Rossano 2010, see also Thompson et al. 2015, 89). Most responses involve the use of talk. Only about one-tenth of responses in the AE collection are fully embodied (n = 7/68, 10.3%), consisting of either head nods or lateral headshakes, while the BE collection holds only one such case (0.6%).

In line with known preference patterns, a significant majority of responses to RfCs are confirmations (BE: 134/180, 74.4%; AE: 158/203, 77.8%). Disconfirmations as well as responses that neither confirm nor disconfirm each account for approximately a tenth of cases in both collections, amounting to a total of 20–25% of non-confirming responses in both samples (Table 16).

5.2 Response tokens: Forms and their positioning

Responses to RfCs typically contain, and indeed start with, a response token. In BE, 134/179 verbal responses contain a response token (74.9%); similarly, this is the case for 146/196 verbal responses in AE (74.5%). As Table 17 shows, the most frequent token in both collections is yeah (BE: 88/134, 65.7%; AE: 72/146, 49.3%). No is the second most frequent token in AE (30/146, 20.5%), but only occurs half as often in BE (15/134, 11.2%). This difference appears to be related to the divergence we have observed with respect to the polarity of RfCs in our two samples (Section 4.2), since negatively polarized RfCs would typically be confirmed with negative response tokens (Sadock and Zwicky 1985, but recall Note 6 above and the exceptional status of negative interrogatives in this regard).

As Table 17 also shows, the overall number of lexical forms far outweighs that of non-lexical forms (with and without bi-labial closure). Inasmuch as the lexical forms may be understood to pre-index speakership incipiency, while the non-lexical forms may be understood to pre-index a more ‘passive’ mode of doing (dis)confirmation, whereby the respondent passes up on the opportunity to take a more substantial turn (Jefferson 1984, 1993, Schegloff 1982), this distribution seems to be consistent with the observation that a considerable share of responses to RfCs is continued past the initial token response (Section 5.4).

An intriguing difference between the two varieties lies in the frequency of the monosyllabic mm, which occurs 17 times in the BE collection but only twice in the AE collection; 16 of the 17 BE mms implement confirmations, and 14 of them do so as minimal responses; 15 of the 16 confirming mms are delivered with falling pitch; one with level pitch. In 2 cases, there is rising‒falling pitch, but in the remaining 14 instances, the pitch onset is low and falling only minimally. Nine are lengthened; ten are accompanied by nodding. This affirmative and confirmatory use of falling mm in BE complements Gardner’s (1997) findings for Australian English where he found falling mm to mainly function as a weak acknowledgment token, typically followed by more same-speaker talk, alongside less frequent uses of mm as a continuer (with (falling-)rising pitch) and as a positive assessment (with rising-falling pitch). Extract (12) is an example of BE mm being employed as a minimal confirming token, in combination with nodding. Meg is talking about her application for postgraduate study at an art college for which she needs to achieve a certain overall grade for her undergraduate degree. Her confirmation of Beth’s RfC in line 07 is delivered in terminal overlap and her nodding continues through a silent period and into Beth’s next turn-at-talk (lines 08‒10).

(12) Recording 10 Beth and Megan (BE #108)

As far as the lexical inventories of confirmation tokens in our two samples are concerned, it is remarkable that that’s right – a form which is often mentioned as a prototypical confirmation token, especially in BE/Australian English (Schegloff 1996, 175, Gardner 2007, Barnes 2011b, 2012) – is absent from both samples. This suggests that, if that’s right is used as a confirmation token in (one of) the two varieties, then its home environment clearly lies outside RfC sequences. In addition, the uneven distribution of right across our two samples – it being somewhat prevalent in the AE sample, but unattested in the BE sample – seems to lend mild support to previous qualitative studies arguing that its use as a confirmation token is subject to regional variability (Bolden et al. 2023, Gardner 2001, 2007). Note also that this asymmetrical distribution mirrors the one observed for the use of right as a lexical tag in RfCs (Section 4.6), suggesting a distinct pragmaticalization process for right uses in AE.

Finally, both collections also contain small numbers of double and multiple sayings of the same token (yeah yeah, no no no) (Stivers 2004) as well as combinations of different tokens (e.g., uh huh yup, mhm yeah, and yeah no).

As alluded to previously, and in line with what is stated in the existing literature (e.g., Heritage and Raymond 2012), if response tokens are used, they are overwhelmingly positioned turn-initially (BE: 128/134, 95.5%; AE: 137/146, 93.8%). Mid-turn and final response tokens are rare in both collections (Table 18).

5.3 Response prefacing

Response prefacing is not very common in responses to RfCs. Only 20/179 (11.2%) verbal responses in the BE sample and 26/196 (13%) in the AE sample featured a turn-initial preface. These numbers are about twice as high as those reported by Stivers (2022), but our samples also include non-confirmatory responses, whereas Stivers looked exclusively at confirmations. Indeed, as Table 19 shows, the inclusion of non-confirmatory responses appears to account for the doubling up of the proportion of response prefaces in our samples.

In the AE collection, the majority of prefaced confirmations are indeed token-responses that are prefaced with oh (oh yeah, oh totally; Heritage 1998, Stivers 2022). In the BE collection, such oh-prefaces are completely unattested, though. Well-prefaces, on the other hand, are attested in both collections and tend to precede confirmations that are subsequently qualified, or done indirectly (Heritage 2015). Non-confirmatory responses (including both disconfirmations and turns that neither confirm nor disconfirm) are almost exclusively prefaced with well or uh(m) in both collections.

5.4 Minimal and non-minimal responses

Responses to RfCs can be minimal and consist of response tokens only. But they can also be non-minimal, when the token response is accompanied by further confirmation-relevant talk from the respondent (Szczepek Reed and Persson 2016). Out of the 134 BE and the 146 AE responses that include response tokens in our samples, the majority includes further confirmation-relevant talk from the respondent (BE: 76/134, 56.7%; AE: 99/146, 67.8%) (Table 20).

Table 21 shows the distribution of minimal and non-minimal responses by response type. As one would expect, minimal (token only) responses are almost invariably confirmatory. Non-minimal responses, on the other hand, show a more diverse distribution. Interestingly, it is not the case that they are strongly (or even clearly) associated with disconfirming responses (BE: 15/76, 19.7%; AE: 12/99, 12.1%), which one would expect to be accompanied by elaborations, accounts, explanations, or the like (e.g., Heritage 1988, Ford 2001). Instead, the vast majority of non-minimal responses are also confirmatory (BE: 60/76, 78.9%; AE: 86/99, 86.9%). Put differently, token-implemented confirmations are commonly expanded or followed up with additional talk beyond the initial response token itself.^[18]

But why are so many token-implemented confirmations expanded beyond the initial token? There is no simple answer to this, as the types of turn-continuations vary greatly. On the one hand, there are cases in which the token is followed by elements that primarily work to calibrate the ‘strength’ of the confirmation, e.g., through the addition of (partial) modified repeats as in YES.=he HA:S. (AE #112; see also Stivers 2005, Enfield and Sidnell 2015) or (epistemic) qualifications (e.g., yeah i THINK so, AE #143). In other cases, confirmers expand their confirmations by taking up a stance toward the state of affairs they just confirmed, e.g., by adding assessments (as in yeah (.) which I find really weird but apparently that’s how it works). In yet other cases, confirmers expand their token-based confirmations with further specifications and elaborations of, or accounts for, ^[19]what they just confirmed (Gubina et al. 2021, 2024, see also Zinken and Küttner 2022, Seuren and Huiskes 2017). Here, too, a variety of factors can give rise to such, typically more substantial, expansions. It may be a matter of recipient design, as when a respondent volunteers an additional account for, or an elaboration of, a token confirmation that the RfC-speaker might otherwise find unintelligible, confusing, or misleading. On other occasions, the RfC can be (and evidently is) understood to look for more than mere confirmation (e.g., in the service of elaborating a topic or topicalizing and unpacking a prior informing). In the following extract, for instance, Briar reports that she has not received a call she was supposed to receive that night, wherefore she now entertains the idea of just going to bed (lines 04‒06, 08). Maya infers from this that Briar’s original plan consisted of going out and doing something fun (lines 10–11).

(13) CFEngn5000 (AE #108)

The designedly unspecific reference to doing something fun employed in the RfC (rather than, say, like you were gonna go out and party?) can be heard to solicit a more specific description of what Briar’s plans were. And Briar hears it that way: after producing the relevant confirmation with a type-conforming response token (line 12), she continues with a rather elaborate description of her original evening plans, in which she explains the arrangements she had made, the reasons for making them, and the role of the as-yet pending phone call she had mentioned earlier as part of these arrangements (lines 13–17, 19–21). The entire description serves to unpack what the token-implemented confirmation of ‘you were gonna go out and do something fun’ entails.

On some occasions, the information-soliciting character of the RfC may be understood to interlock with considerations of recipient design in responding. Consider the following example in which Jean is telling her Dad how she went for a run and saw the local tennis team practice.

(14) CFEngn5635 (AE #157)

Jean’s announcement that she even played a little bit herself (line 04) is met with a newsmark that is prosodically marked for surprise (line 05) and solicits further talk about this matter (Thompson et al. 2015, Wilkinson and Kitzinger 2006). Indeed, when Jean merely produces a positive assessment after her confirmation token (lines 06‒07), Dad solicits further information about the team (line 09). Jean’s response ends with a report of how it came about that she ended up playing with them, viz. her having been invited to hit a few shots by the team’s trainer (line 11). Dad’s RfC in line 16 draws on presumptive knowledge that would have constituted a potential obstacle to her participation, namely that Jean did not have her tennis racket with her, which Jean confirms with a negative token-response in line 17. On its own, this token-based confirmation would be somewhat confusing, as it would fail to address the issue of how she was able to play nevertheless. And indeed, soliciting such a ‘how come’-type account can be understood to be the underlying purpose of Dad’s RfC (compare Pomerantz 1980, Pomerantz 2017). Jean’s turn continuation past the confirmatory token engages with the relevance of this issue by adding the account i bOrrowed uhm MIKE’S (line 18) and a subsequent specification of the person reference (the COAch’s, line 19).

Cases like these call for a more differentiated view as regards the sequential implicativeness of RfCs. Beyond their use as candidate understandings in other-initiated repair sequences (or similar, sequentially embedded insert expansions) and possibly some more strongly sequence-closing implicative upshot formulations (see extract (9)), it appears difficult to maintain the view that RfCs tend to occur in subordinate lines of action that attract minimal confirmatory responses (cf. Enfield et al. 2019, 290–1). As the above cases show, for various reasons, respondents quite regularly orient to RfCs as requiring or deserving more than minimal confirmation with a response token.

5.5 Full and expanded repeats

As previously suggested in the existing literature, repetition is not a major resource speakers of English draw on for building responses to RfCs. The clear majority of responses in our samples do not re-use any lexical material from the RfC at all. Both full as well as expanded repeats are uncommon as a response format in English RfC sequences (Table 22). The observed frequencies are even lower than the ones reported for answers to polar questions more generally (∼5–10%), despite the fact that our sample includes confirming as well as dis- and non-confirming responses.

Given that repetitional answers have been shown to convey the respondent’s claim to epistemic ownership over the asserted proposition (Schegloff 1996, Stivers 2005, Heritage and Raymond 2012, Enfield and Sidnell 2015, Enfield et al. 2019), this relative paucity of use as part of sequences that are fundamentally characterized by an underlying recipient-tilted epistemic asymmetry may be somewhat surprising. However, it may be precisely this characteristic feature of RfC sequences which further inhibits the use of full or expanded repetitions as a response format. Claiming or asserting epistemic ownership over the matter at hand may be tantamount to an ‘epistemic overkill’ in a context where the RfC speaker already conceded epistemic authority to the respondent in and by seeking confirmation for its accuracy or correctness from them. In fact, this same pragmatic mechanism may provide the foundation for the use of repetitional confirmations as a practice for ‘confirming allusions’ (Schegloff 1996), allowing a respondent to index that the RfC has (adequately) explicated something the respondent had already conveyed inexplicitly before.

That repetitions are used to confirm RfCs in rather special interactional circumstances is vividly illustrated by the only case of a full repeat from the BE collection. Here, the repeat confirms an oh-prefaced self-repair and follows an aborted polar interrogative. In line 01, Amy initiates a new telling concerning her two sisters Kay and Emma. Emma, her younger sister, is pregnant.

(15) Recording 5 Amy and Hannah (BE #181)

In line 06, Hannah aborts an information-seeking turn was that (.) one that’s preg- which is heard as asking about whether Kay is the sister who is pregnant and therefore receives a slightly delayed no response from Amy in line 09. Hannah’s RfC in line 08 proffers a revised candidate understanding, oh ↓EMma’s pregnant, and is met with Amy’s full repeat (line 09). Amy’s turn does not only identically repeat Hannah’s RfC lexically but also matches it prosodically (primary accent on ↓EMma, falling final pitch from there and no further accents). While the lexical and prosodic mirroring designs Amy’s turn as responsive to the turn it matches (Szczepek Reed 2012, 2020), her turn also provides a response to Hannah’s aborted enquiry and unambiguously clarifies the identity of the pregnant sister. In this context of displayed referential trouble, then, a full repeat is chosen over a token or other form of simple affirmative to achieve greater clarity.

5.6 Interim summary: Building responses to RfCs in BE and AE

In summary, the response slot following RfCs is almost always filled with a response, which is most often a confirmation. Responses to RfCs frequently contain response tokens, which are typically positioned initially. In a solid number of cases, response tokens constitute the whole of the response, and such minimal responses are virtually always confirmations. However, quite regularly responses that contain response tokens also contain other talk, which shows that RfCs are frequently deployed to do more than deal with brief obstacles of uncertainty about the matter at hand. Full repeats of the RfC turn are extremely rare, which may be related to the fact that epistemic authority over the confirmable proposition has typically already been conceded to the would-be confirmer with the preceding RfC. In such an environment, the use of repetitional confirmations appears to be highly constrained.