Request for confirmation sequences in Yurakaré

2.1 Background on Yurakaré

Yurakaré is a language isolate spoken by around 1,600 people (INE 2015, 32) in the Andean foothill area of central Bolivia. The language is considered endangered due to a break in intergenerational transmission (e.g., Plaza Martínez et al. 2011, 243–5). There is a full reference grammar of the Yurakaré language (Van Gijn 2006) and a comprehensive ethnographic account revealing abundant links between language and culture (Hirtzel 2010). Moreover, there are works on various parts of grammar, for instance, middle voice and ideophones (Van Gijn 2010), subordination (Van Gijn 2011a), argument structure (Van Gijn 2011b), evidentiality (Gipper 2011, 2014a), and demonstratives (Gipper 2017), among others.

The expression of arguments is optional in the language so that a predicate alone can constitute a full clause. Arguments are frequently dropped, in particular subject arguments of transitive clauses (Van Gijn 2006, 267–8; see also Gipper 2016). Constituent order is relatively free, although there is a preference for the subject argument to follow the verb (Van Gijn 2006, 267). Another important syntactic feature of Yurakaré is that nouns and verbs alike can be used to form predicates (Van Gijn 2006, 267–75). In fact, evidence is accumulating^[3] that Yurakaré may best be analyzed as an omnipredicative language (Launey 1994, 2004) possessing “a logical system where nouns and verbs are only subclasses of a major class of predicable notions” (Launey 2004, 59).

2.2 Previous work on RfC sequences in Yurakaré

In Yurakaré, polar questions are not syntactically or morphologically marked; they rather exhibit all syntactic features of declaratives (Van Gijn 2006, 288). RfC sequences in Yurakaré have not been thoroughly described up to the present. The only information on RfCs comes from studies on evidentiality in the language (Gipper 2011, 2014a, 2014b). Evidentiality is a linguistic category concerned with expressing the utterance producer’s information source (e.g., Aikhenvald 2004, 3). Yurakaré has a set of three evidential enclitics, one of them being the inferential marker =tiba. This morpheme indicates that the speaker inferred the proposition expressed in the utterance rather than having direct knowledge of it (Gipper 2011, 100–2). In a conversational corpus partially overlapping with the one investigated in this article, a quarter of the uses of =tiba were found to be in RfCs (Gipper 2014b, 112). This demonstrates that flagging propositions introduced in RfCs as inferences is one of the main functions of inferential =tiba. In Section 4.4, it is shown that this does not mean that in turn, inference marking is a very prominent feature of Yurakaré RfCs overall: Only n = 9/200 (4.5%) of the cases in the collection are marked with inferential =tiba.

Other epistemic markers have been described to occur in RfCs. The commitment marker =la which usually expresses the speaker’s commitment to the truth of the proposition or the force of a directive can be used in RfCs to elicit the addressee’s commitment with respect to the truth of the proposition expressed (Van Gijn 2006, 246–7; see also Gipper 2020). Moreover, the subjective epistemic modal =laba can appear in RfCs (Gipper 2011, 137–41). Both markers are not overly frequent in the collection of 200 RfCs (Section 4.3).

While RfCs have not yet been studied in detail for Yurakaré, there is an investigation on requests for reconfirmation comparing the three languages German, Low German, and Yurakaré (Gipper et al. 2024). A request for reconfirmation is a type of polar question that does not introduce any new information into the discourse but asks the addressee to reconfirm something they previously stated (König and Pfeiffer 2024). In Yurakaré, we find a larger proportion of response tokens in responses to requests for reconfirmation (Gipper et al. 2024, 216) than in responses to RfCs as reported in this article. I will argue in Section 5.5 that this relates to the lower degree of informativity and sequential import of requests for reconfirmation when compared to RfCs.

No investigation of plain confirming or disconfirming responses to RfCs in Yurakaré has been published so far. However, Gipper and Groß (2024) examine responses that show an orientation toward confirmation but do more or less than confirming (e.g., Robinson 2020).^[4] Their study shows that in Yurakaré, modified repeats are used for two functions in this domain, namely epistemic downgrading and giving a rough rather than full confirmation (both functions in the realm of less than confirming). It is argued that these formats sustain the repeat-preferring logic of the language. The response token te ‘yeah’, in contrast, is employed as an ‘escape strategy’ when a repeat would come with some unwanted entailments or implicatures. This is often the case when the second slot of the response (Raymond 2013) deals with some terms or agendas of the question (e.g., Stivers and Hayashi 2010) or its formulation that are perceived as problematic. In such cases, a modified repeat without the preceding response token could be heard as a correction and/or as not confirming. Therefore, a preposed response token, te ‘yeah’, is employed to express the confirming part of the response. This response token is shown to be highly specialized for this particular function. These findings are consistent with the arguments put forward in this article, where it is argued that repeat format is the preferred format for formulating plain confirming responses to RfCs. Moreover, both works demonstrate a functional differentiation between repeat and response token format. An example of the response token te preceding a modified repeat is discussed in Section 5.4.

2.3 Cross-linguistic response possibility space

As the Yurakaré data on formats of responses to RfCs presented in this article provide a new data point for our understanding of the cross-linguistic response possibility space (Stivers 2022) to polar questions, it is relevant here to introduce some previous work on response formats employed in other languages.^[5] A classical and often cited classification of responses to polar questions is that by Sadock and Zwicky (1985) who propose three different types of systems: yes/no, agree/disagree, and echo. In a yes/no response system, an affirmative answer is typically expressed with a positive particle such as ‘yes’ in English, while a disaffirmative answer is indicated by a negative particle such as ‘no’. In an agree/disagree system, particles are also employed, but positive particles are used to express agreement with positive and negative questions, whereas negative particles indicate disagreement. Echo systems, in contrast, rely on repeat format to formulate affirming and disaffirming responses (Sadock and Zwicky 1985, 189–91). Enfield et al. (2019, 280) criticize this classification based on the observation that both yes/no and agree/disagree systems employ the same type of format, namely a particle, which means that not all contrasts between the three systems are located at the same level of comparison.

Moreover, the classification proposed by Sadock and Zwicky (1985) does not fully hold up to empirical scrutiny. To give one initial example, the Yurakaré data presented in this article show that it is not necessarily the case that a language prefers the same strategy in both affirming and disaffirming answers to polar questions; in Yurakaré, confirming responses to positive polarity RfCs are most frequently formatted as repeats, whereas disconfirming responses to such RfCs do not show a preference for repeat format (Section 5.1). Furthermore, research on conversational data in a variety of languages reveals that languages mobilize the two different formats of response tokens and repeats for a functional division of labor. On the one hand, cross-linguistic evidence suggests that response tokens are more frequently employed for pragmatically or sequentially subordinate actions, and/or polar questions with a shallower epistemic gradient, such as for instance an understanding check or an echo question/request for reconfirmation (Sorjonen 2001, Keevallik 2010, Enfield et al. 2019, Rosemeyer and Schwenter 2019, Harjunpää and Ostermann 2023, Gipper et al. 2024 [this issue]).^[6] On the other hand, it has been found that in English conversations, repeats are mostly employed for responding to actions that are pragmatically marked, such as ‘confirming allusions’, i.e., confirming the content of a polar question and, at the same time, confirming that it has been implicitly conveyed by prior talk by the confirming speaker (Schegloff 1996; see also Enfield and Sidnell 2015, 138–41) or asserting the responding speaker’s epistemic rights or his/her rights regarding the course of action proposed by the question (Heritage and Raymond 2012). Repeats have been argued to be employed mostly for marked responses in other languages as well, e.g., Polish (Weidner 2023) and Russian (Bolden 2023), for instance in the form of reasserting the respondent’s epistemic primacy. Response tokens, in contrast, have been argued to constitute a format for expressing a plain confirmation that conforms to the constraints set by the question (Raymond 2003).

Some authors propose that a division of labor along these lines – employment of response tokens for unmarked, and of repeats for pragmatically marked answers – may be universal based on the semiotic properties of the two formats. The basic argument by Enfield and Sidnell (2015) and Enfield et al. (2019) is that the two formats differ with respect to the degree of agency they convey. Enfield et al. (2019, 286–7) propose that repeats manifest a higher degree of agency than response tokens by virtue of reasserting the proposition stated in the question, thereby allowing respondents to challenge the thematic and sequential agency of the questioner. This idea relies on the observation that when a question is posed, thematic and sequential agency are not equally distributed, but are tilted toward the person posing the question: The questioner introduces a new proposition and also offers a particular formulation for it, thereby exerting thematic agency. Moreover, the questioner exerts sequential agency by setting in motion a course of action that forces the addressee to produce a certain type of response, thus restricting the addressee’s possibilities for action.

Investigating the distribution of response token vs repeat format in confirming responses to polar questions in a sample of conversational data from 14 languages, Enfield et al. (2019) suggest that languages have a ‘natural’ preference for response tokens (‘interjection-type strategies’ in their terminology) when confirming polar questions, whereas confirming in repeat format is universally marked. They base this proposal on the observation that 12 out of the 14 languages in their sample exhibit a clear preference for response tokens over repeats, while there are only two languages in which response tokens do not constitute the most frequent format for confirming responses, ╪Ākhoe Hai║om (Khoisan, Namibia) and Tzeltal (Mayan, Mexico). Enfield et al. (2019) propose a cultural explanation in terms of the specific cultural values of avoiding interpersonal coercion and striving for epistemic symmetry, respectively. These cultural values, they argue, give rise to a higher frequency of repeats in the two languages, as repeat format allows speakers of ╪Ākhoe Hai║om to push back against the sequential agency of the questioner, thereby resisting the constraints set by the question, and speakers of Tzeltal to reduce epistemic asymmetry between questioner and respondent.

However, there is a growing amount of evidence that such a ‘natural’ preference for response tokens in confirming responses to polar questions is not empirically tenable. Harjunpää and Ostermann (2023) show that in Brazilian Portuguese, a repeat can in fact constitute a pragmatically unmarked plain confirming response to a polar question. The same has been shown for the three Mexican languages Tzeltal, Yucatec (both Mayan), and Zapotec (Otomanguean) (Brown et al. 2021). Repeats were furthermore found to constitute the most frequent format of confirming responses to polar questions in Latin as used in the comedies of Plautus and Terence (Potočnik 2023). The Yurakaré data presented in this article demonstrate the same. Moreover, for Yurakaré, it has already been shown that a response token, concretely te ‘yeah’, is specialized for pragmatically marked functions (Gipper and Groß 2024). These works, while emphasizing the functional differentiation of repeats vs response tokens, at the same time show that the concrete shape of the functional differentiation is not universal across languages.^[7]

One proposal that seeks to integrate semiotic properties of repeats with language-specific preferences is offered by Brown et al. (2021) who argue that repeats, while possessing the semiotic feature of informational redundancy which may suggest that interactants have equal epistemic access, may still be subject to different preferences, given that such a claim of epistemic equality can be interpreted differently according to the cultural values of a community. For instance, in English, repeats are used mostly for more competitive interactional moves, whereas in the three Mexican Indigenous languages the authors investigate, repeats constitute interactional moves that are perceived as affiliative. This proposal is basically compatible with languages differing regarding whether a repeat can constitute an unmarked plain confirmation, as it expects languages to differ regarding the interactional contexts in which repeats are employed.

The Yurakaré data offered in this article constitute a novel data point in the ongoing debate. This article contributes findings on the functional differentiation of response tokens vs repeats in the language, adding to our knowledge of the possible shapes that a division of labor between the two formats can take. Moreover, the data show that in Yurakaré, a repeat can constitute a plain and pragmatically unmarked confirming response to an RfC, which I take as evidence against the idea that repeats are universally more marked than response tokens, thus arguing against the proposal by Enfield and Sidnell (2015) and Enfield et al. (2019).

4.1 Syntactic design

The most frequent syntactic structure found in the collection is that of a verbal main clause with further material, such as additional verbal clauses, nominal predicates denoting event participants, adverbs, or postpositional clauses. The frequencies of the different syntactic designs of RfC turns in the Yurakaré collection are presented in Table 1.

An example for the most frequent format is given in (2), where Jony uses an RfC in line 001 to ask Rodolfo whether certain people have their plant seeds ‘here’ (in the village). Jony’s RfC contains a main verb, tütü ‘sit’, carrying the third person plural object prefix ma- and the third person plural subject suffix -w, resulting in the interpretation of ‘have’. In addition, there is an adverbial phrase in the form of the adverb ani ‘here’, morphologically segmentable as ana=y ‘DEM=LOC’, and a nominal predicate denoting the object participant of the main verb (masemillaw ‘their seeds’).

We can furthermore observe one of the crucial features of omnipredicative languages in line 001 of (2): the use of the same paradigm of subject markers on verbs and nouns (Launey 2004, 59). Both the verb tütü ‘sit’ and the noun semilla ‘seed’ carry the third person plural subject suffix -w.

The observation that nouns can be employed as predicates without the need for derivation in Yurakaré has the consequence that stand-alone nominal predicates are not elliptical phrases, but rather complete utterances consisting of a main clause (Launey 2004, 58). There are, however, some asymmetries between stand-alone (i.e., without any further linguistic material added) verbal and nominal predicates in the RfC collection, suggesting that at the interactional level, the two types of predicates may not be fully equivalent: In the group of stand-alone verbal clauses, only n = 2/34 (6%) can be understood as understanding checks, i.e., RfCs that identify a problem in the preceding talk by the other speaker by means of presenting a possible solution (Heritage 1984, 319).^[8] For nominal predicates, in contrast, the proportion is much higher with n = 15/27 (55.5%). It thus seems that nominal predicates tend to be employed for sequentially subordinate actions more frequently and consequently may not have the same potential for introducing new information into the discourse as verbal clauses. This pattern is illustrated in example (3), where we can observe the use of an RfC in the form of a stand-alone verbal predicate (line 004), and the use of another RfC – an understanding check – in the form of a stand-alone nominal predicate (line 006). Paulina is telling Manfredo about some events that happened to her husband when he went for a hunt and his dog entered the burrow of a peccary.

In line 001, Paulina explains that the dog was inside together with the peccary. Manfredo then uses a content question to ask how Paulina’s husband got ‘it’ out. The referents, the dog and the peccary, are not overtly mentioned in either of the two utterances. In line 003, Paulina starts explaining that her husband pierced the earth and made an opening. She uses two subordinate clauses both marked with the subordinator =ja, here indicating a succession of events with the same subject. As she has not yet produced a main clause, it is clear that her utterance is unfinished with respect to its grammatical form, and that hence some content is still to come. Nevertheless, Manfredo comes in with an RfC in the form of a stand-alone verbal clause, suggesting that he pulled ‘it’ out (note again that the referent is not mentioned overtly). This RfC arguably introduces new content into the conversation, presenting a reasonable inference from Paulina’s previous utterance. After Paulina’s confirming response in repeat format (line 005), Manfredo produces another RfC, this time in the form of a stand-alone nominal predicate (line 006). This RfC works as an understanding check, asking whether this indeed applies to the dog (and not the peccary, the other possible referent). Paulina’s response in line 007 starts out as a confirmation in the form of the response token otte ‘yes’, which is however abandoned and repaired with a correction: It was the peccary and not the dog that was pulled out by her husband. The example thus illustrates the use of a stand-alone nominal predicate in an understanding check, arguably an informationally and sequentially subordinate function.

4.2 Polarity

The majority of RfCs in the Yurakaré collection exhibit positive polarity format with n = 164/200 (82%). Of the n = 36/200 (18%) RfCs with negative polarity,^[9] the most frequent negation strategy is the negation particle nij (n = 29) followed by the existential negation predicate nijta ‘it is not the case/it is not there/it does not exist’ (n = 6).^[10] The negation predicate (which can also function as an adverb) kani ‘not yet’ plays a marginal role with n = 2 uses. While the latter two always take scope inside the proposition and contribute to its truth conditions, the negation particle nij is employed for two types of negation in the collection, taking scope inside or outside the proposition. First, it can scope inside the proposition, thus forming part of its truth conditions (n = 18/29). In (4), Asunta asks her interlocutor if she did not catch any fish. The negation particle nij negates the predicate mim ‘you took’.

In its second use, nij takes scope over the whole proposition without contributing to its truth conditions (n = 11/29). This is exemplified in (5) where Susana and Lorena are discussing the Yurakaré terms for certain animals. These animals are not present in the discourse situation, so the reference to them is not deictic. Susana inquires whether a particular bird is the one called pitchitanti in the Yurakaré language. The negator nij is placed at the beginning of the utterance, a feature obligatory for negation scoping outside the proposition.

Similar to what has been noted for instance in English (e.g., Ladd 1981, Romero and Han 2004), RfCs negated with nij (where nij occurs at the beginning of the utterance) may be ambiguous between a reading with scope inside or outside the proposition. In (5), an interpretation with scope inside the proposition where the demonstrative lati ‘that one’ is negated is possible in principle (‘That is not the one that is called pitchitanti?’). However, a feature that distinguishes negation inside and outside the proposition more reliably is the structuring of the responses which reveal the understanding of the addressee. For some instances in the corpus, this next-turn proof procedure (Hutchby and Wooffitt 1998, 15; see also Schegloff 2007, 7–8) shows that, as in English (Ladd 1981, Romero and Han 2004), in Yurakaré RfCs with negation scoping outside the proposition expect a response that confirms the non-negated proposition rather than the negative polarity proposition^[11] – the negation can thus be understood as non-literal (Hentschel 1986, 76–7) or metalinguistic (Reese 2006).

Gutzmann (2015, 4–7) proposes to analyze meaning components that do not contribute to the truth conditions of a sentence, but do still convey conventional meaning (as opposed to enrichments or implicatures that depend on the context) as pertaining to the realm of ‘use-conditional’ meaning. On this account, use-conditional meaning components, while leaving truth conditions unaffected, enforce specific conditions of felicitous usage on the sentence in which the relevant linguistic item occurs. In the following paragraphs, it is argued that the interactional contribution of non-truth-conditional negation in Yurakaré RfCs is indeed conventionalized, which makes these cases amenable to an analysis in terms of use-conditional meaning. In the remainder of this article, I therefore use the terms ‘truth-conditional negation’ and ‘use-conditional negation’ to refer to negation scoping inside and outside the proposition in Yurakaré RfCs, respectively.

The effect of RfCs with use-conditional negation in Yurakaré can be described in terms of Sudo’s (2013) notions of epistemic and evidential bias. The epistemic bias concerns the expectations of the speaker regarding whether or not the proposition is true, while the evidential bias is concerned with available contextual evidence. Sudo (2013, 281) finds that for English polar questions with negation outside the proposition, it is necessary for the speaker to entertain a positive epistemic bias, i.e., for him/her to believe the proposition to be true. Another condition for the use of such a polar question is that a positive evidential bias – i.e., contextual evidence that the proposition is true – is not available, resulting in a conflict between the two biases: The epistemic bias is positive, whereas the evidential bias is either neutral or negative. The same analysis can be applied to Yurakaré RfCs with use-conditional negation. They express the speaker’s expectation that the proposition be true, while either not being supported or even contradicted by the available contextual evidence. Possible interpretations include that the speaker claims to have knowledge of the proposition, which is however not fully available at that moment due to fading memory, or a conflict regarding what the speaker believes to be the case and some contextual evidence. This can be illustrated with example (5): In her RfC with use-conditional negation in line 001, Susana expresses her own assumption that the bird they are talking about is called pitchitanti, while at the same time acknowledging that in light of evidence from the preceding talk, this cannot be considered likely any longer.

The distribution of responses supports this analysis in terms of a conflict between epistemic and evidential bias for RfCs with use-conditional negation in Yurakaré, as the proportion of confirming responses is relatively low: Only n = 4/11 of the cases receive a confirming response. In addition, there are n = 2/11 disconfirming responses, n = 5/11 responses that neither confirm nor disconfirm, and n = 1/11 cases where no response ensues.

In general, using polar questions with negation outside the proposition to formulate biased questions seems to be cross-linguistically common, as suggested by Hentschel’s (1998, 219) typological survey. The interpretation of such questions as being able to convey a positive epistemic bias has been described for various languages, among them English (Sudo 2013), Hungarian (Gyuris 2017), Japanese (Sudo 2013, Ito and Oshima 2014), Russian (Repp and Geist 2022), and German (Deppermann et al. 2024 [this issue]). In addition, languages can tweak the format of the question to express additional evidential biases (e.g., Sudo 2013 on Japanese). The data on Yurakaré presented here add an additional data point to our cross-linguistic understanding of polar questions with negation outside the proposition. However, there is a caveat regarding the Yurakaré data, as we have to keep in mind that Yurakaré has been in contact with Spanish for around three centuries (Hirtzel 2010, 151–233) and that Yurakaré speakers are mostly bilingual with Spanish (Plaza Martínez et al. 2011, 243), a language for which negative polar questions with non-truth-conditional negation conveying different pragmatic functions are also attested (García Jiménez 2005). We therefore cannot rule out the possibility that we are dealing with a contact phenomenon here.

4.3 Modulation

Modulation with epistemic and attitudinal markers (excluding inference marking, Section 4.4, but including use-conditional negation, Section 4.2) occurs in roughly a quarter of the RfCs in the collection, with n = 51/200 (25.5%). Table 2 gives an overview of the epistemic and attitudinal markers found in Yurakaré RfCs. As some RfCs contain more than one marker, the number of markers (n = 56) is higher than the number of utterances that contain modulation. It becomes clear that three forms account for the majority of cases: the commitment marker =la, the marker =ye which in the context of RfCs indicates a problem with memory, and the intersubjective possibility epistemic modal enclitic =ya. None of the forms reaches a frequency high enough to state that it plays a crucial role in formulating RfCs – the commitment marker =la is the most frequent with n = 14/200 (7%).

In its non-interrogative use, the enclitic =la indicates the speaker’s commitment to the truth of a proposition or the force of a directive speech act. In RfCs, it comes to index a request for the addressee’s commitment regarding the truth of the expressed proposition (Van Gijn 2006, 246–7, Gipper 2020), as in (6). Rodolfo is telling Ediberto about some construction works that are being performed. He is talking about one of the men involved, stating that this man will go again when his cement is finished (lines 001–003).

In line 004, Ediberto produces an RfC marked with =la, asking whether the man is going to fetch it (more cement). As the preceding context strongly facilitates the inference that he is going to fetch cement (rather than just going to do something else), Ediberto can be relatively certain to receive a confirming response, which ensues in line 005 in the form of a repeat. The data suggest that =la is used when the degree of certainty to receive a confirming response is relatively high: n = 13/14 of the RfCs with =la receive a confirming response.

The second-most frequent type of modulation, use-conditional negation with the negation particle nij, has been discussed above in Section 4.2 and is therefore not described further here. The enclitic =ye constitutes the third-most frequent type of modulation in the collection.^[12] A detailed analysis of this enclitic has not been carried out so far, but it is clear that in RfCs it indicates that the speaker has knowledge of the information but at that moment cannot fully remember. In (7), Susana and Lorena are discussing the Yurakaré terms for some animals. They have been talking about an animal called wërri which they believe is a type of toad.

In line 001, Lorena produces an RfC, asking whether this particular animal is really called wërri in Yurakaré. The marker =ye is used here to indicate that this is something she should know but cannot remember with full certainty at that moment. In line 002, Lorena confirms with a repeat containing the commitment marker =la.^[13] The speakers’ degree of certainty regarding their memory seems relatively low for = ye, as less than half (n = 4/9) of the RfCs with =ye receive a confirming response.

4.4 Inference marking

In general, inference marking is not overly common in Yurakaré RfCs, with n = 24/200 cases (12%). Moreover, only n = 9/200 (4.5%) of the RfCs in the collection contain the evidential enclitic dedicated to marking inference, =tiba. There are two further inference marking strategies found in the RfCs in the collection: the change-of-state tokens aj, ëj, and yj (Heritage 1984) that present the proposition conveyed by an RfC as a realization based on prior talk (n = 7) and discourse markers derived from anaphoric elements (n = 8).

An example of a use of the evidential =tiba in an RfC is given in (8). Nuria and Yasmina have been talking about the possibility of finding parrots to turn into pets.

In line 001, Nuria states that she wants to get a parrot for one of her daughters. After acknowledging this as news with a change-of-state token (line 002) and a longer pause, Yasmina poses an RfC in line 003 marked with =tiba, asking whether Nuria’s daughter is going to come to visit the village. The inference marker makes explicit that this is an inference, in this case from the preceding talk: If Nuria is going to look for a parrot for her daughter, a person apparently known not to be in the village, then that daughter must come to visit the village at a certain point to be able to retrieve the parrot. In line 004, however, Nuria indicates that she does not know by means of the interjection aj ‘I don’t know’ followed by an epistemically downgraded response stating that her daughter is probably going to come if her father travels to pick her up. Some further explanation of the father’s plans ensues (omitted here).

The downgraded response in this example suggests that inference marking with =tiba in the RfC may not convey a high degree of certainty – only n = 3/9 of the RfCs marked with =tiba receive a confirming response. Another observation that points in this direction is that n = 5/9 RfCs with =tiba have rising final intonation, which seems to be associated with higher degrees of uncertainty (this topic is discussed further in Section 4.7).

In contrast to =tiba, change-of-state tokens introduce inferences with a relatively high degree of certainty, as in (9). Asunta is telling Paulina about an event where she and her group capsized with a canoe. In line 001, Paulina produces an RfC, asking Asunta to confirm that they capsized at a place where the water was shallow.

In line 002, Asunta disconfirms by means of a content question conveying a strong assertion (Gipper 2022), resulting in the effect of asserting that it was not shallow at all. This leads Paulina to produce the RfC of interest, which is preceded by the change-of-state token ëj. In this RfC, Paulina draws the inference that the water was deep at the place where Asunta and her group capsized. Arguably, this is the only possible inference from Asunta’s strong objection to the water being shallow. In line 004, Asunta also asserts that the water was indeed deep.^[14]

The overall low frequency of discourse markers derived from anaphoric forms in RfCs suggests that they are not strongly associated with the formulation of RfCs. The forms achu and lachu ‘like that’ (n = 4) are used to mark inferences in RfCs only by two young female speakers from the same village. The remaining n = 4 forms are anu ‘like this’ (n = 2), lachama ‘it is like that’ (n = 1), and lam (n = 1), a short form of lachama that is pragmaticized to some extent.

4.5 Connectives

Only a relatively small percentage of the RfCs in the collection are introduced by connectives (n = 20/200, 10%), Yurakaré thus being among the languages with the lowest frequency for this variable in the sample (Pfeiffer et al. forthcoming [this issue]). The largest part of these are accounted for by anaphoric discourse markers (n = 8) and change-of-state tokens^[15] (n = 7). These forms have been described in Section 4.4 on inference marking. In addition, there is one case of the discourse marker la which awaits a detailed analysis of its meaning and use – see example (7). There are hardly any dedicated clause-combining connectives in the Yurakaré collection, the only forms found being kusu ‘maybe/or’ (n = 3) and alla ‘therefore’ (n = 1). This low proportion of dedicated grammatical connectives may in part be explained by the fact that in Yurakaré, subordinating clause-combining strategies are generally clause-final (Van Gijn 2011a). There are three enclitics that mark subordinate clauses: =ya ‘irrealis’, =ja ‘perspective continuation’, and =ti ‘perspective shift’.^[16] The adverb kusu, in contrast, is a clause-initial, but not a subordinating connector. Rather, it is mostly used clause-initially as an adverb or a discourse marker. The form kusu is a short version of kusuti which literally means ‘if it wants’ (Van Gijn 2006, 320). It has two connected readings: epistemic possibility and disjunction, where the epistemic reading can be considered the basic one (Van Gijn 2006, 322).^[17] The form alla ‘therefore’ is based on an anaphoric element, ati ‘that’, and is morphologically analyzable as ati =la ‘DEM = INS’. Like kusu, it is not a subordinating connector.^[18]

In example (10), a use of kusu in an RfC is shown. The speakers have been talking about a type of parrot whose chicks, against expectations, have been found to not yet have emerged from their eggs. In line 001, Nuria gives information about the hatching time of another bird species, the blue-crowned motmot, stating that they hatch when the water in the rivers is high. This is met with an RfC by Yasmina in line 002, asking whether it may be the case that this is also the reason the parrots are not yet ready. She introduces the RfC with the adverb kusu, indicating that this is something she considers possibly true.

Arguably, a logical disjunction interpretation (‘or’) in terms of connecting the utterance to the preceding one by Nuria is not feasible here, as Nuria is talking about blue-crowned motmots and Yasmina about parrots, comparing them to the motmots. Rather, in addition to the epistemic reading, the adverb kusu helps to connect Yasmina’s RfC to the preceding discourse in terms of a looser type of disjunction, adding a new possibility to other possible explanations for the observation that the parrots have not yet hatched.

The distribution of connectives shows that grammatical features of a language have an impact on the resources available to speakers for formulating actions. Subordinating clause-combining connectives are not readily available to the speakers, which makes the pool of connectives smaller than for languages which have clause-initial subordinating and non-subordinating connectives.

4.6 Tags

Yurakaré has no tags available in its grammar, which explains the low frequency of n = 2/200 (1%). One of the two instances coded as having a tag is a case where the tag no ‘no’ from Spanish is added by one of the young speakers to a confirmable as an ad-hoc borrowing. In the other case, a speaker adds the word nijta-ø=ya ‘NEG-3SG.SBJ = INTSUBJ’ → ‘it may not be the case’ to her confirmable, producing the effect of a reversed-polarity tag. However, this is an idiosyncratic, non-conventionalized usage.

4.7 Prosodic design

As polar questions are not syntactically or morphologically marked in Yurakaré, intonation is the only feature that may formally distinguish them from declaratives (Van Gijn 2006, 288).^[19] Different final contours can be observed for Yurakaré RfCs, so final intonation cannot be considered a fully distinctive feature of interrogativity or response mobilization in this language. RfCs in Yurakaré do not rely on final rising intonation to achieve response mobilization – n = 129/200 (64.5%) of the RfCs in the collection are delivered with falling final intonation, while only n = 51/200 (25.5%) have rising final intonation. Remarkably, these numbers are similar to those reported for confirmables in RfCs without tags in German (Deppermann et al. 2024 [this issue], 12). Moreover, in unison with the German data, the Yurakaré data support Selting’s (1995, 308) finding that falling final intonation imposes a stronger restriction on response expectations, whereas rising final intonation allows for a greater diversity in the responses (Deppermann et al. 2024 [this issue], 13). This may translate into a higher degree of uncertainty indicated by a final rise in Yurakaré RfCs. The relevant frequencies are summarized in Table 3.

Table 3 also shows that level intonation seems to pattern more like falling intonation concerning the responses received. However, as the absolute frequency of level intonation is only n = 20/200 (10%), these results are not fully conclusive.

5.1 Responsive actions and practices

Confirmations are preferred as responses to RfCs in the Yurakaré collection, accounting for n = 133/192 (69%) of the responses given (n = 8 RfCs in the collection do not receive any response), while on-record disconfirmations are clearly dispreferred with only n = 14/192 (7%) instances (Table 4).

The category neither includes a variety of actions, ranging from repair initiation over replies indicating a lack of knowledge to responses that are repaired from a confirmation to a disconfirmation or vice versa. Moreover, this category includes a number of transformational answers that assert a proposition different from the one put forward in the RfC (Stivers and Hayashi 2010). Crucially, this type of response often seems to be used to avoid giving an on-record disconfirmation while clearly still implicating that the proposition introduced by the RfC is not true, as exemplified in extract (11) where Asunta is telling Elisa about a journey she recently made.

In line 001, Elisa asks Asunta to confirm that they were planning to travel in a canoe. In line 002, after a pause of 0.7 s which foreshadows an upcoming dispreferred response, Asunta asserts that they were going to travel in a motorboat. This response is not an on-record disconfirmation, as it does not directly negate the proposition that they were going to go in a canoe. However, it still implicates that this proposition is false.

I now turn to the most frequent formats used for formulating confirming and disconfirming responses to different types of RfCs. Consistent with previous findings on response formats for confirming responses to polar questions across languages (Section 2.3), confirming and disconfirming responses to RfCs in Yurakaré rely mostly on two formats: repeats and response tokens. Only n = 14/133 (10.5%) of confirmations do not contain any of the two, while all disconfirmations comprise either a repeat or a response token. A more marginal format for expressing confirmation is that of anaphoric forms with n = 9/133 (6%) instances (only cases without additional repeat elements). The remaining five confirmations take the following formats: non-verbal response in the form of a nod (n = 3), negation predicate (n = 1, refer to extract (16)), and emphatic response with reformulation (n = 1).

Table 5 summarizes the frequencies of the two most frequent formats, repeats and response tokens, for confirmations and disconfirmations following RfCs with different polarity types. The table only includes responses coded as ‘confirmation’ or ‘disconfirmation’. Moreover, it only includes cases where a verbal response ensued, thus excluding the n = 3 instances where a non-verbal confirming response in the form of a nod was given.^[20] As responses may contain both a repeat and a response token, the percentages may exceed 100% when taken together for the two formats. The values in boldface indicate which of the two formats is more prominent for the particular combination of RfC and response.

In the following, the most important observations that can be drawn from the distributions shown in Table 5 are discussed. First, the formats of verbal confirming responses show a patterning according to the polarity of the RfC. RfCs with positive polarity typically receive confirming responses containing a repeat with n = 84/113 (74%) cases (13 of them combinations with response tokens). For negative polarity RfCs with truth-conditional negation, in contrast, the proportion of repeats in confirming responses is much smaller with n = 4/13 (31%) instances. In n = 7/13 (54%) cases, we find response tokens, the most frequent being the token-like negation predicate nijta (n = 6).^[21] Given that n = 4 of these confirming responses with nijta come from the same speaker, it seems that there is no clear preference across speakers for a particular response format regarding confirming responses to RfCs with truth-conditional negation.

In Section 4.2, it was noted that negative polarity RfCs with use-conditional negation are, in some of the responses, treated as asking for a confirmation of the non-negated proposition. While the number of instances is too low to draw definite conclusions on the quantitative distribution of response formats for this type of RfC, one piece of evidence is that there are n = 3/4 confirming responses containing/consisting of an anaphoric form clearly referring back to the proposition in the RfC excluding the negator, thus confirming the non-negated proposition.^[22] This is exemplified in (12). Before the start of the extract, the interlocutors have established that a meeting will take place soon. In her RfC in line 001, Lorena enquires about the presence of the two interlocutors’ brother-in-law at the time of the meeting. By means of use-conditional negation,^[23] she indicates that she believes that he is not going to be there (epistemic bias after Sudo 2013), but the fact that there is going to be a meeting makes her doubt, presumably because their brother-in-law would be expected to participate in that meeting (evidential bias after Sudo 2013).

The anaphoric response in line 002 excludes the negator nij from its scope, confirming that the brother-in-law is not going to be there. This is further supported by the expansion of the response in line 003 where Susana explains that even though their brother-in-law will not be in the village for the meeting, still everybody knows what they need to know, which means that his absence will not be a problem. The expansion is thus oriented toward accounting for the absence rather than the presence of the brother-in-law. In contrast, for RfCs with truth-conditional negation, there is only one confirming response in anaphoric format, confirming an RfC containing the negation predicate kani ‘not yet’. In the response, the anaphoric form clearly includes the negation under its scope and thus confirms the negated proposition.

A look at the overall distribution of response formats in disconfirmations in the Yurakaré collection reveals a complementary pattern to that found for confirming responses: Positive polarity RfCs are exclusively disconfirmed with response tokens (n = 7/7), mostly followed by an elaboration or correction (n = 6/7). This is exemplified in (13) where in Asunta’s disconfirming response (line 002) to Manfredo’s positive-polarity RfC (line 001), the token-like element nijta is followed by a transformational response that asserts a proposition different from the one put forward in the RfC (Stivers and Hayashi 2010).

Negative polarity RfCs with truth-conditional negation, in contrast, are exclusively disconfirmed by means of repeat format (n = 5/5) marked with the commitment marker =la but without a response token – see example (14). This pattern is reminiscent of that found for instance in European Portuguese, where disconfirming responses to negative statements are often repeats with some emphasis added (Martins 2013).

There are n = 2 cases where a response to an RfC with use-conditional negation takes the form of the token-like negation predicate nijta. In line with the observations made above, these two cases were coded as disconfirming responses, as the response is assumed to target the non-negated proposition. For an example, refer to (5). However, as there are only two cases, the findings are not fully conclusive on this point. Formally, these responses are indistinguishable from confirming responses with nijta to RfCs with truth-conditional negation – see (15) for an example. In future studies, it is advisable to carry out a closer investigation of negative polarity RfCs with both types of negation using a larger amount of data.

In sum, the factors of type of response and polarity of the RfC interact to produce the patterned distribution of repeat vs response token formats in responses to RfCs in Yurakaré. The different distributions of response formats according to polarity of the RfC make it clear that it is useful to treat positive and negative polarity RfCs (distinguishing truth-conditional and use-conditional negation) as separate types of sequence structures in comparative studies of responses. Moreover, there are clear differences in the formatting of confirmations vs disconfirmations, which underlines the importance of investigating them separately (see also Enfield et al. 2019).

5.2 Response tokens

With only six different items, the set of response tokens used in responses to RfCs in the Yurakaré collection is relatively small, in particular when compared to particle-rich languages like German which show a higher variability in response tokens (Deppermann et al. 2024 [this issue], 14). The frequencies of the response tokens used for confirmation and disconfirmation in the collection are given in Table 6.

The affirmative tokens otte ‘yes’ and its short form te ‘yeah’ are response tokens in the grammatical sense, as these forms cannot be inflected by means of subject marking and usually do not take any other morphology. The same is true for the non-lexical affirmative token mjm ‘mhm’ and its variant aja ‘uh hu’. The negative form nijta, in contrast, is not a response token in the grammatical sense, as it is a predicate that carries inflection in the form of subject indexing and other types of grammatical marking. However, to stay comparable to the other languages in the sample and to avoid a bias disfavoring response tokens, occurrences in the third person singular (marked with the zero-marker -ø) and without any tense, aspect, or modality marking were coded as response tokens, as in (15).

In this extract, a factor that underlines the token-like character of some uses of nijta is that the predicate nijta in the confirming response in line 002 does not carry congruent person marking to the RfC in line 001: In the RfC, the predicate is marked for third person plural with the suffix -w, whereas in the response, the predicate nijta is zero-marked, resulting in a third person singular interpretation. Moreover, nijta does not carry any tense, aspect, or modality markers. The only markers that are attached are the intensification prefix (in the form of CVC reduplication) and the enclitic =la indicating speaker commitment.

In contrast, it was decided to not code cases such as (16) as response tokens, where nijta in the response in line 002 carries congruent person marking to the RfC in line 001.

The predicate in the RfC in line 001 is marked for second person singular with the suffix -m, which is met by first person singular marking in the confirming response in line 002 (the suffix -y). In addition, the predicate nijta in the confirming response carries aspect marking in the form of the habitual marker -jti. In this way, this predicative use of nijta functions more like an anaphoric predicate than a response token. The case in (16) was the only occurrence of nijta where this was the case, suggesting that the predicate nijta may show some degree of pragmaticization toward a response token when used in confirming and disconfirming responses to RfCs.^[24]

5.3 Clusters of response tokens

In the Yurakaré collection, there are no clusters of response tokens. Although clusters and repetitions of response tokens have (seldomly) been observed for the language in other data, it seems that it is not an overly common strategy, given that it is not found in 200 RfC sequences.

5.4 Position of the first response token

When used in non-minimal (response token plus further material) confirming responses, roughly two-thirds of response tokens occur in turn-initial position, with n = 11/17 (65%). Out of the n = 6 cases where a response token stands in turn-final position, four come from the same speaker, Rodolfo, for whom the construction seems to be somewhat conventionalized, as it is always a full repeat of an RfC consisting of a single word as the first part of the response, followed by the response token otte ‘yes’. The other two cases come from two different speakers, one being a partial repeat followed by the token mjm ‘mhm’, the other a modified repeat followed by the token te ‘yeah’.

With n = 7/14, half of all disconfirming responses consist of a response token plus continuation. In all cases, the response token occurs turn-initially. There are n = 6 cases of the token-like negation predicate nijta and one case where the negating response token no is spontaneously borrowed from Spanish by a young speaker.

5.5 Minimal and non-minimal responses

Minimal responses consisting solely of a response token are not very frequent in the collection in verbal confirming and disconfirming responses, with n = 25/130 (19%) cases in confirmations and n = 2/14 (14%) cases in disconfirmations (both instances of the token-like negation predicate nijta).^[25] Table 7 shows the frequencies of the response tokens used in minimal confirming responses.

Given that repeats constitute the preferred format for confirming responses to positive polarity RfCs (Section 5.1), the question arises as to whether there is a functional differentiation between repeats and minimal response token responses when used in such confirmations. In the following, I discuss evidence showing that response tokens in Yurakaré are preferred as responses to polar questions with a low informational and/or sequential load. The functional differentiation between repeats and response tokens, then, is one of informativity and action status, in line with cross-linguistic evidence showing that choosing a response token in confirming responses to polar questions becomes more likely when the question implements a pragmatically subordinate action (Section 2.3). In Yurakaré, response tokens show a higher proportion of use in confirming responses to requests for reconfirmation that, in contrast to RfCs, do not introduce a new proposition into the discourse but rather ask the other interactant to reconfirm their prior turn (Gipper et al. 2024, 216; Section 2.2). A request for reconfirmation can be considered sequentially and also informationally subordinate, as it will usually initiate a side sequence rather than a main sequence, and does not introduce a new proposition to the conversation.

If we now look at the RfCs in the collection that receive a minimal confirming response in the form of a response token only, we find that this explanation can also be applied to at least some of the cases in the corpus. For the minimal uses of the response token otte, n = 9/11 cases are responses to RfCs that implement understanding checks, arguably a sequentially subordinate action (see also Enfield et al. 2019, 290). Consider extract (17) where Manfredo is telling Asunta a Yurakaré mythological narrative about the son of the Thunder (tërërë abonto). In line 001, he explains that this character was looking at some people. Earlier in the story, he had narrated that those young people were bathing and that then a thunderstorm started. In his utterance in line 001, he uses the demonstrative naa marked with the ‘source’ marker =jsha, while at the same time pointing and looking upwards to indicate the location from where the son of the Thunder was watching.

When Manfredo starts to continue his telling (line 002), Asunta comes in with an RfC (line 003), asking whether it was from the sky/from above that the boy was watching. Arguably, this can be understood as an understanding check, given that Manfredo had already indicated that it was from above and that it is well known that the Thunder is located there. Manfredo’s confirming response in line 004 takes the form of the response token otte. After a pause, he continues his narrative in line 005 without expanding the confirming response, starting with a repetition of his utterance from line 002 where he is interrupted by Asunta’s RfC.

Another situation where a response token is regularly used is when the speaker has already made the assertion s/he is asked to confirm. In RfC sequences, this is the case when there is some overlap between the RfC and an identical assertion made by the speaker who is asked to confirm. In extract (18), Lorena is telling Susana about some events she experienced in the past. In line 001, she asserts that she and her group were going to fetch bananas when suddenly their grandmother heard something that sounded like a jaguar.

In line 002, Susana produces an RfC, asking whether the events took place during Lorena’s childhood, interrupting Lorena’s utterance in line 001, which is why Lorena is not able to complete the last word starting with ta-la- ‘1PL.OBJ-MAL-’. Susana’s RfC is confirmed by Lorena with a full repeat in line 003. After a request for reconfirmation by Susana in line 004 that remains without a response, both women start producing an utterance with the same content at the same time (lines 005 and 006), with Susana producing an RfC asking whether Lorena was the age of a certain child, and Lorena stating that she was the age of exactly that child. To respond to the RfC with a default repeat response, Lorena would have to repeat that utterance again which would result in a doubling of her own utterance. It seems that this is avoided, as she gives a confirming response in the form of the response token otte in line 007.

In both examples above, otte seems to take on a sequence organizing function, basically indicating that the matter brought up in the RfC is now considered sufficiently established and the conversation can move on. Thereby, otte may serve as a device for closing side sequences, albeit this is clearly different from the closing of problematic side sequences described by Oloff (2017) for the German response particle genau ‘exactly’. The side sequences in the two examples above are not oriented to as problematic by the speakers; rather, otte is merely employed to indicate that the topic has been sufficiently discussed and can thus be closed.

As argued by Gipper and Groß (2024), the second function of response tokens in Yurakaré is to work as an ‘escape strategy’ when a repeat would come with unwanted semantic or sequential entailments. Specifically, this function is associated with the response token te ‘yeah’ when used in non-minimal responses. When employed on its own in a minimal response, te tends to occur in contexts when some informational trouble has already been on the conversational table. Another context where stand-alone te is used seems to be when the speaker wishes to address some trouble with the terms of the question but leaves the actual trouble unexpressed. Consider extract (19) where Nuria is telling Yasmina about her plans for one of her fields. In line 001, she states that it would be good if she could finish her rice field before the seeds germinate. After a pause, Yasmina produces an RfC, asking whether Nuria has weeded there (line 002). Nuria gives a confirming response in line 003 in the form of the response token te.

While the sequential context is not conclusive regarding whether any trouble is addressed with this response format, the prosodic design of the response token te in line 003 suggests that the response may be doing more than merely confirming. As shown in Figure 1 extracted with Praat (Boersma and Weenink 2022), the token is lengthened and produced with a fall-rise contour with some devoicing toward the end. Given that in non-minimal responses, te expresses confirmation where a second response part deals with some portion of the question that is perceived as problematic, it is plausible that this function may carry over to stand-alone uses. In the case of (19), a possible analysis is that the response token te treats the RfC as an unnecessary question, challenging its askability (Stivers 2011): As weeding is a necessary activity, Nuria may consider the RfC asking whether she has weeded her field as unnecessary given that the answer is too obvious. However, as the collection contains only n = 7 cases, more research is needed to determine the range of functions of stand-alone te.

Figure 1

Pitch contour for response token te in extract (19).

Finally, the two cases where mjm is used to form a minimal response react to RfCs (both understanding checks, which accounts for the non-repeat format) that directly follow a confirming response to another RfC that has been responded to with a lexical response token (one otte, one te). These may be cases where doubling response tokens is avoided.

Non-minimal responses consisting of a response token and other material in any order partially show a patterning regarding the formats with which the response tokens occur (Table 8). These different formats relate to whether a response constitutes a plain (full repeat) or a non-straightforward (repeat with modifications) confirming response.

Basically, we can observe that the token otte ‘yes’, when used in a non-minimal response, always co-occurs with a full repeat,^[26] whereas the token te ‘yeah’ co-occurs with formulations that either provide modifications or expansions of repeats or other formulations. This provides further evidence for the functional differentiation between the two response tokens: The response token otte, which is reserved for plain unproblematic confirmations, is combined with full repeats which indicate a plain confirmation in Yurakaré without any problematization of the question (Section 5.6). The formats co-occurring with te, in contrast, all deal with reworking the RfC, showing that te is employed when the speaker deals with some problematic aspect of the RfC.

An example of a non-minimal response with the response token te ‘yeah’ is given in (20), where te (line 002) precedes a repeat with modifications (line 003) where the placeholder element itta (indicating problems with remembering the word) from the RfC in line 001 is replaced by a verb.

Had Paulina only produced the modified repeat without the response token te, she might have been heard as not confirming but only correcting Asunta, whereas the combination of the response token with the modified repeat allows her to confirm while also implementing modifications in the repeat following it. This exemplifies how in non-minimal responses, te indicates an orientation toward confirmation in contexts where the default format for confirmation, a repeat, would come with some unwanted entailments or implications (Gipper and Groß 2024).

In sum, in Yurakaré we find a functional differentiation between response tokens and repeats, as well as between different response tokens. The token otte ‘yes’ is specialized for confirming RfCs with a low degree of informativity and/or sequential import such as understanding checks, whereas the response token te is employed for responses that, in addition to confirming, work on some terms or agendas of the question or its formulation. Repeats, as shown in Section 5.6, constitute the default format for plain unmarked confirming responses to RfCs with a higher informational and/or sequential import.

5.6 Full and expanded repeats

In Yurakaré, a repeat is the default format for a) confirming responses to positive polarity RfCs, and b) disconfirming responses to negative polarity RfCs with truth-conditional negation (Section 5.1). These repeats can be full, expanded, partial, or partial with modifications. This section focuses on full and expanded repeats without additional response tokens in confirming responses to positive polarity RfCs.^[27]

Together with Low German (Weber 2024 [this issue]), Yurakaré is among the languages in the sample with the highest proportion of full repeats overall, with n = 40/189 (21%) of all verbal responses containing a full repeat. An additional n = 14/189 (7.5%) cases contain a full expanded repeat. Of the n = 71/113 verbal confirming responses to positive polarity RfCs formatted as repeats without an additional response token, n = 30/71 (42%) are full and n = 9/71 (12.5%) full expanded repeats.

Full repeats often confirm positive polarity RfCs that consist of one single content word (n = 21/30). In these cases, a repeat cannot take a less-than-full format, as there is only one content word constituting the target for repetition. An example is given in (21). Before the start of the extract, Manfredo told Paulina about an event where another person accidentally shot at a dog while hunting peccaries, having mistaken the dog for a peccary. In line 001, Paulina asks whether this happened during the day, which is met with a confirming response in repeat format by Manfredo in line 002.

It is also possible for utterances consisting of more than a single content word to be confirmed with a full repeat (n = 9/30), as in (22) where an RfC containing an adverbial element (adojla ‘on top’) and a main clause (balip ‘you walked’) are repeated in the confirming response in line 002 (including the necessary deictic shift from a second-person plural to a first-person plural subject). This shows that it is not only one-word utterances that license the use of a confirmation formatted as a full repeat.

What also becomes apparent in examples (21) and (22) is that the interactants do not orient to the confirming responses in repeat format as expressing meaning components in addition to confirmation such as underlining the respondent’s epistemic authority. In (21), the response is treated as informative and surprising in the form of the interjection ij produced by Paulina in line 003. Manfredo’s ensuing self-repeat in line 004 offers a recommitment to the information which orients to Paulina’s treating it as surprising. In line 006, Paulina follows up on her treating the information as surprising by asking how it is possible that the person in question did not recognize the dog, which is met by an explanation offered by Manfredo (not shown). The interactants orient to the remarkability of the fact that the events happened during the day, but not toward any issues regarding epistemic authority. In (22), the response is not further oriented to by the participants; rather, the responding speaker simply continues her telling in line 005. This demonstrates that in Yurakaré, repeats have the capacity to express plain and simple confirmations that do not orient toward issues of epistemic authority.

Moreover, we can observe that in Yurakaré repeats are not regularly implicative of sequence expansion (Heritage and Raymond 2012, 186). The question of sequence expansion rather seems to hinge on the degree to which the information given in the repeat is noteworthy or unexpected. In (21), the confirming response in repeat format engenders an expansion of the sequence, as the speakers orient to the matter introduced by the RfC as having a high informational and sequential import, treating the information as noteworthy. In (22), the confirming response formatted as a full repeat does not provoke an expansion of the sequence. In line 005, Lorena simply continues telling her story without further elaborating on the content brought up in the RfC in line 003.

In sum, it has been shown that full repeats in Yurakaré can be employed as plain confirming responses that do not deal with issues of epistemic authority and that do not necessarily engender an expansion of the sequence. Let us now look at an example of a full expanded repeat. Expanding a repeat allows speakers to offer additional information to that already given in the repeating part. Expanded full repeats seem to fulfill a range of functions in the data. One of them is to offer more precise information than that requested in the RfC. In (23), such a case is shown. In line 001, Ediberto produces an RfC asking whether a certain worker makes thirty (Bolivianos) with his work.

In his confirming response, Rodolfo uses a full expanded repeat to confirm this and at the same time add that the worker makes this amount of money per day. The response thus gives information that is more precise than that requested in the RfC. There is no evidence in the extract that would point to an interpretation of the response as dealing with issues of epistemic primacy.

In addition to the full and expanded repeats, there are n = 23/71 (32%) repeat confirmations without an additional response token formatted as partial repeats and n = 9/71 (13%) formatted as partial repeats with modifications, showing that speakers of Yurakaré have diverse repeat formats available to formulate confirming responses.