Requests for Verification across Varieties of Spanish A Comparative Approach to Gaze Behaviour

This article examines requests for verification across varieties of Spanish. It focuses on the role of gaze behaviour in the design of the request and shows how gaze is inter - related with other resource s, such as facial expression s, epistemic marker s, prosody and tag question s. By comparing speakers from Argentina, Bolivi a, Colombia and Ecuado r, the article finds similarities but also differences between the groups of speaker s. The speakers from Bolivia and Ecuador rarely establish mutual gaze or use facial expressions during the verification sequence. Instea d, more frequentl y, they deploy other resources for obtaining verifications, such as tag question s, hand gestures and rising intonatio n. The article considers differences in gaze behaviour within the larger machinery of response mobilisation and shows how gaze is intertwined with other aspects of conversational organisatio n. Results are based on qualitative and quanti - tative analysi s. The data stems from face-to-face conversations recorded in similar side-by-side settings.


Introduction
The present contribution focuses on requests for verification, a type of action that has first been described by Charles Goodwin as follows: In examining situations in which the main addressee was an unknowing recipient but a knowing recipient was also present, it was found that when the speaker moved his gaze to a knowing recipient, he produced a display of uncertainty about what he was saying, thus constructing an action -a request for verification -appropriate to a knowing recipient (C.Goodwin, 1981: 166).
This type of request is specific to multiparty interactions with the following epistemic configuration: one of the recipients is treated as an unknowing recipient, while the current speaker and the other recipient(s) share knowledge about the events of talk.For example, this is the case in games explanations (C.Goodwin, 1981) or in collaborative storytellings (C.Goodwin, 1981;Mandelbaum, 1987;Lerner, 1992).In order to make requests for verification recognizable, speakers shift their gaze direction from the unknowing to the knowing recipient.This way, they signal a change in the addressee to both recipients.In addition, speakers display uncertainty about a verifiable element, using epistemic markers, facial expressions, tag questions or rising intonation, which makes a verification by the knowing recipient relevant next.
As an illustrative example, consider excerpt (1).1 María (MAR) and Juan (JUA), a couple, are telling their friend Pedro (PED) about how they met at a party.The extract starts rather at the end of the story, when María is talking about what happened the day after the party, when they saw each other again.
(1) A la noche 'At night ' (ssat201703, Argentina) 1 A detailed description of the conversational setting is provided in section 4 (data and method).A list of the transcription symbols is provided in the appendix.
MAR: y yo me desperté al otro DÍ:a, and I woke up the next day fui al día del PAdre, (0.9) I went to father's day MAR: +*y::#1 después +hablamos a la tA:rde+ y a la NO:che, (-) and after (that) we spoke in the afternoon and in the evening mar +gazes at PED---+gazes mid.distance-+gazes at JUA--> jua *gazes at MAR--> Downloaded from Brill.com09/24/2023 07:41:25AM via free access At the beginning of the excerpt, María elaborates on what happened the day after the party.After spending Father's Day with her family (l.02), she texted Juan both in the afternoon and in the evening (l.03).Subsequently, she displays uncertainty about whether they met that day or the day after (l.04).
Juan takes the floor in overlap and confirms that they met that same night (ll.05-06), showing his orientation to this piece of information as verifiable.
After a silence, María gazes back towards Pedro (l.08, #3) and continues with the story, explaining that from that night onwards, they never stopped seeing each other (l.08).Pedro responds with head nods and a smile (l.09).
In (1), María mobilises several multimodal resources in order to make line 04 recognizable as a request for verification addressed to Juan.First, already during line 03, María changes her gaze direction from Pedro (#1), who is next to the camera in front of her, to Juan (l.03).Juan was already gazing at María, which occasions mutual gaze between them.This way, she makes available, Figure 1 María changes her gaze direction from Pedro (#1), who is sitting in front of her (out of frame), to Juan (#2).After requesting a verification from Juan, she continues the telling, gazing towards Pedro (#3).both to Juan and to Pedro, that coparticipation from Juan might become relevant, although, it is not yet certain in what form.María continues to gaze at Juan and displays uncertainty by initiating her turn with creo ('I believe').Furthermore, she changes her facial expression, frowning her eyebrows and tilting her head (#2, l.04), which can be used to downgrade an epistemic stance (Roseano et al., 2016) and initiate repair (Stolle and Pfeiffer, forth.).Indeed, she does display uncertainty before introducing a potential verifiable element, in this case whether they saw each other that evening, which allows Juan to verify this information almost immediately, even in overlap with María's turn continuation.
As we observe in excerpt (1), gaze behaviour can play a central role in requests for verification.These requests are addressed to a recipient that, before the request, is not sequentially expected to speak next.Therefore, a change of gaze direction from one recipient to the other is constitutive of this type of action.These are general observations that have only been confirmed for English speakers (C.Goodwin, 1981;Mandelbaum, 1987;Lerner, 1992).Yet, gaze behaviour presents a certain cultural variation in conversation (Rossano et al., 2009;Blythe et al., 2018).For instance, it has been shown that the gaze behaviour of Spanish and Quechua speakers from Bolivia (Satti and Soto Rodríguez, 2021) and Ecuador (Muz, 2023) diverges from the gaze behaviour reported in previous studies (Kendon, 1967;C. Goodwin, 1981).With this in mind, the article compares the design of requests for verification across four varieties of Spanish (Argentina, Bolivia, Colombia and Ecuador) with an emphasis on gaze behaviour.Furthermore, it considers to what extent differences in gaze behaviour are intertwined with other aspects of conversational organisation.
In what follows, I consider the relationship between requests for verification and requests for confirmation (section 2) and discuss previous studies on pragmatic typology and gaze behaviour (section 3).After presenting data and method (section 4), I proceed to the analysis (section 5), which is divided into two parts.First, I present the general tendencies in the design of requests for verification across the varieties of Spanish available in the data (5.1),providing both qualitative and quantitative evidence to support the analysis.The second section of the analysis tackles two instances of requests for verification that are considered deviant cases of the general trend (5.2).In the conclusion, I summarise the results and offer some thoughts for discussion (section 6).

Requests for Verification as Subtypes of Requests for Confirmation
Requests for verification can be considered a subtype of requests for confirmation (see König and Pfeiffer, this issue).In some regards, this subtype of request is similar to other ways of requesting confirmation, but it also entails a few differences that are worth noting for the purposes of this article.
As mentioned in the introduction, the term "request for verification" was first employed by Goodwin (1981) to describe a specific type of request observable in his data.Since then, it has been taken up mainly by other studies dealing with collaborative storytellings (Mandelbaum, 1987;Lerner, 1992;Dressel and Satti, 2021).These studies show that requests for verification invite a polar response from a knowing recipient regarding a verifiable element of which the current speaker is unsure.Although the requester selects a next speaker, they also constrain the response to a type conforming response, which is made relevant by the format of the request and which can be as minimal as "yes" or even a head nod (C.Goodwin 1981: 159).
Requests for verification, as described by Goodwin, differ from other types of requests for confirmation in two main aspects.On the one hand, they are always constructed as a "side sequence" (Jefferson, 1972) to a main activity.When the speaker initiates a request for verification, the two knowing participants "confer" regarding some element of the turn in progress before the official delivery to another recipient (Lerner, 1993: 221).Gaze behaviour plays a central role in making the sequence visible as "conferring".Speakers change their gaze direction from an unknowing recipient to a knowing recipient and, in doing so, they not only make a response from the knowing recipient relevant but also avoid a response from the unknowing recipient, which in these cases is the one addressed-so-far.On the other hand, speakers accomplish a specific type of epistemic asymmetry: they downgrade their epistemic stance towards a verifiable element, but do so in a way that the verifiable element falls within an epistemic domain which is only shared with this particular recipient (and not with the recipient addressed before the request).
The distinction between confirmation and verification has also been discussed in a different sequential position, namely in responsive position to informings (Thompson et al., 2015).In this context, minimal clausal responses with final rising intonation, depending on their syntactic design, can be oriented to as requesting "confirmation" (declaratives, as in "you are?") or "verification" (interrogatives, as in "are you?").According to the authors, the main difference between these two lies on the epistemic stance embodied by the requester.While requests for verification "are generally offered as, and treated as, being from a K-position, thereby upholding the informer as K+" (Thompson et al., 2015: 95), requests for confirmation embody a higher epistemic stance on the side of the requester.Although in this study I will not discuss requests for verification in responsive position, it is interesting to note that displaying a lower epistemic stance may be a more appropriate way of addressing a knowing recipient, which supports Goodwin's claim that requests for verification "may also operate ritually, displaying deference to the other party present who could be telling the story" (Goodwin, 1981: 159).
In sum, requests for verification (as described by Goodwin) can be considered a subtype of requests for confirmation which occurs within a specific epistemic configuration: when, in a group of speakers, two or more copresent participants have a higher knowledge of the events of talk.Due to the constitutive role that gaze behaviour plays in accomplishing this type of action, I consider requests for verification particularly apposite for the main goal of this contribution, namely to undertake a comparative approach to gaze behaviour among speakers of Spanish.

Pragmatic Typology and Gaze Behaviour
This article approaches the comparison of language use in conversation from the perspective of pragmatic typology (Dingemanse et al., 2014).In pragmatic typology, the anchor point for linguistic comparison is a similar action in naturally-occurring talk across languages, such as other initiated repair (Dingemanse et al., 2014), other-recruitments (Floyd et al., 2020) or requests for confirmation (König and Pfeiffer, this issue).Actions are identified by a qualitative analysis, relying on conversation analytic methods (Schegloff 2007), which allows researchers to build collections of practices that are positionally sensitive (Clift and Raymond, 2018).This qualitative analysis can then be followed by a quantitative comparison across languages, which, depending on the size of the project, is supported by a large network of collaborating researchers (see Dingemanse et al., 2014or Floyd et al., 2020), or confronted with the results of previous research in other languages (see Blythe et al., 2018 or Kushida andHayashi, 2022).By comparing similar actions in typologically unrelated languages, conversation analysts have provided evidence for the universality of certain features of interaction.This has been the case for some aspects of turn taking and preference organisation (Stivers et al., 2009), other-initiated repair (Dingemanse et al., 2014) and sequence organisation (Kendrick et al., 2020), which show great similarity independently of the grammatical structure of the languages under study.Yet, this approach has also revealed some variation regarding the practices employed by participants in conversation.This is the case for the use of gaze (more detail below), other-repetition (Gipper, 2020), the management of silence (Mushkin and Gardner, 2009) or the expression of gratitude (Floyd et al., 2018), among others.Still, it is worth noting that, beyond specific Downloaded from Brill.com09/24/2023 07:41:25AM via free access contributions on variation in English (see for example Mulder and Thompson, 2008;Neumeier, 2023), conversation analysts have not primarily focused on the comparison of regional varieties of the same language, as is the case in the current article.
One innovation of Conversation Analysis as a method for comparative studies is that it allows to incorporate both verbal and non-verbal conduct in the comparison.Due to the complex interlayering of audible and visual modalities, face-to-face conversation presents a challenge for comparative approaches.Anchoring the analysis in a specific action (rather than a formal aspect of talk) provides a language independent "comparative concept" (Haspelmath, 2010) and allows for an onomasiological approach to linguistic typology (Behrens, 2012), thus permitting the inclusion of visible conduct among the elements for comparison.This is a central advantage for the current paper, which puts an emphasis on the study of gaze behaviour.
Several studies have noted the orderliness of gaze behaviour in interaction and its manifold conversational uses.Gaze allows interactants to make the participation framework of an encounter visible (C.Goodwin, 1981;Vranjes et al., 2018) or to signal response preference in conversation (Kendrick and Holler, 2017).Gaze direction is also one of the main resources involved in turn taking, both for allocating turns-at-talk (Auer, 2021) and for holding the floor (Kendon, 1967).Furthermore, specific gaze patterns are a central part of the accomplishment of courses of action, such as word searches (Dressel, 2020;Auer and Zima, 2021), reenactments (Sidnell, 2006;Pfeiffer and Weiss, 2022) or storytellings (C.Goodwin, 1984;Rühlemann et al., 2019).In other words, this means that, although participants can freely gaze at someone or remove their gaze from co-participants, this relative freedom actually depends on the gaze expectations associated with the ongoing course of action (Rossano, 2013).Another key affordance of gaze is that it allows co-participants to monitor each other's visible actions (M.Goodwin, 1980), which is particularly important in face-to-face encounters.This way, due to the ubiquity of gazing in conversation, gaze withdrawal is, on the one hand, a central resource for achieving disengagement and, on the other hand, it can be treated as accountable by participants (C.Goodwin, 1981).
Beyond the orderliness of gaze in conversation, numerous studies have noted differences in gaze behaviour and argue that patterns of gaze are not necessarily universal.Ethnographic studies have pointed out different gaze behaviour in Australian Aboriginal communities with respect to Anglo-Australians (Walsh, 1991;Evans and Wilkins, 2000).In psychology, experimental settings have evidenced differences in gaze behaviour according to gender (Bayliss  (La France and Mayo, 1976), familiarity (Broz et al., 2012) or culture (Haensel et al., 2022).These studies share similarities in two directions.On the one hand, they usually point to differences between "Western" society and either Asian speakers or indigenous speakers from America and Australia.On the other hand, they seem to agree that most differences concern the frequency and duration of mutual gaze and the meaning associated with sharing our gaze with others.
From a conversation analytic perspective, there are two main studies that approach variation in gaze behaviour.Rossano et al. (2009) show that in a Tzeltal community recipients gaze at speakers less frequently in question answer sequences, while Blythe et al. (2018) evidence differences in gaze behaviour in four aboriginal communities in Australia.In both cases, the results show that speaker gaze to recipient is common, while the establishment of mutual gaze is less frequent or less prolonged in these communities.This is due to recipients gazing less frequently at speakers than what is observably the norm in other communities (Kendon, 1967;C. Goodwin, 1980).One central contribution of these studies is that differences in gaze behaviour are intertwined with other aspects of conversational organisation, such as the practices for responding to questions or the practices for next speaker selection.
As stated above, gaze behaviour in conversation is tightly linked to the course of action within which it is deployed (Rossano, 2013).This has two central implications.On the one hand, a comparative study of gaze behaviour has to rely on action necessarily.Otherwise, differences in the use of gaze among participants can be easily mistaken with differences in the role of gaze between courses of action or in the participation framework of the encounter (see Auer and Zima, 2021).On the other hand, it is difficult to isolate the use of gaze from the larger machinery of response mobilisation (Stivers andRossano, 2010, Blythe et al., 2018).Indeed, gaze direction is not the only resource that can be deployed for obtaining a response from a recipient.Rather, it intertwines in complex ways with other response mobilising features of turn design, such as interrogative lexico-morphosyntax, interrogative prosody and recipient-tilted epistemic asymmetry (Stivers and Rossano, 2010: 4), and with the degree of response relevance that a certain type of action may carry (for example, assessments and requests for information are quite different in this regard).This means that a speaker can select a next speaker in more or less explicit ways (using name vocatives or gaze direction) or they can adjust the epistemic relation towards this recipient by displaying uncertainty, i.e., more recipient-tilted epistemic asymmetry mobilises a response by the more knowing recipient.In other words, eliciting a response by a recipient is arguably a universal interactional task.However, participants can deploy different resources for doing so, which makes it an interesting locus for pragmatic comparison.In sum, there is a good base to assume differences in gaze behaviour across groups of speakers.Moreover, the few studies that have addressed this issue using conversation analytic methods show that differences in gaze behaviour are linked to other aspects of conversational organisation.In what follows, I aim to contribute to this line of inquiry by comparing the design of one specific action (requests for verification) across speakers of different varieties of Spanish.This way, the article approaches variation in gaze behaviour from a holistic perspective (see Gipper, 2020) and reflects on how differences in gaze behaviour may be intertwined with other response-mobilising features of turn design.

Data and Method
The data stems from the Freiburg Sofa-Talks Corpus, which comprises video recordings of couples, family members or close friends.2This conversational setting was designed to provide authentic opportunity spaces for a range of narrative practices to be deployed to make shared experiences accessible for others.The participants are recorded in a place familiar to them, where they feel comfortable sharing memories and stories with the researcher, with whom they have a personal relationship (friend, family member, co-worker, etc.).The main camera is positioned next to the researcher, facing the two participants (see figure 2).Whenever possible, a second camera is set up in a way that captures the visible behaviour of the researcher (see figure 3).3This way, the corpus provides a large set of comparable videodata with complete visual access to the gestures and gaze of the participants.The similar setup of the recordings across different languages (and language varieties) provides the basis for cross-linguistic studies, and the size of the corpus allows for both qualitative and quantitative investigations of multimodal practices.Data was collected by different researchers at the University of Freiburg (Germany).All participants have signed informed consent forms before producing the material.The data can be considered "semi-experimental" (see Mondada, 2009).The researchers are present and the participants are encouraged to sit side-by-side in front of the camera to allow for reliable identification and coding of 10.1163/26660393-bja10092 | Contrastive PragmaticS (2023) 1-33 multimodal behaviour.However, the setting was designed to be as "naturalistic" as possible.The participants choose freely what they talk about and they self-manage the process of turn-taking, choosing topics and assigning participation roles.Moreover, they freely decide how and to what extent they address the researcher, which leads to dynamic changes in the participation framework.As a result, the recordings encompass both storytelling activities and other forms of talk.Furthermore, researchers have a personal relationship with participants, which often leads to them being treated as participants themselves (see Hofstetter, 2021 for a detailed reflection on this issue).The advantage of the corpus selected for this study lies in that it allows both for high comparability and visibility while using a simple recording technique (one or two small cameras).This procedure does not require to invite participants to a laboratory, which facilitates the inclusion of individuals with diverse social and geographical backgrounds.
The study also includes a group of Spanish speaking immigrants in Germany (12 participants), which will be treated separately, since the participants speak different varieties of Spanish and, therefore, cannot be considered as illustrative of one particular variety.4The participating researchers speak the same variety as the participants, except for two participating researchers of German origin, who nevertheless have a near-native command of Spanish.
For this study, I collected a total of 320 requests for verification by 61 different participants.Each request has been identified using the next turn proof procedure, i.e., the speaker displays uncertainty about a potential verifiable element and the recipient responds with a verification, a denial or a claim of not knowing, showing their orientation to some element of the previous turn as verifiable (see C. Goodwin, 1981: 149-166, for different ways of responding to requests for verification).Extracts have been transcribed according to the GAT2 conventions (Ehmer et al., 2019) and include multimodal annotations (Mondada, 2001).For quantitative purposes, gaze behaviour was coded using ELAN (2022) by means of a binary annotation (yes/no) of three parameters (Figure 4): 1) the participant on the left looks at the participant on the right; 2) the participant on the right looks at the participant on the left; 3) eye contact between participants (obtained from the automatic crossing of lines 1 and 2).

Analysis
The analysis focuses on how participants make requests for verification recognizable for others across the varieties of Spanish available in the data.
First, I will present the main findings with a qualitative analysis of illustrative excerpts and with quantitative evidence accounting for the complete collection (5.1).The general trend is that, during the verification sequence, speakers 4 The speakers included in this group are originally from Spain (Cádiz, Barcelona and San Sebastián), México (Ciudad de México), Chile (Santiago), Perú (Lima) and Argentina (Buenos Aires).from Bolivia and Ecuador establish mutual gaze less often than speakers from Argentina and Colombia.In turn, speakers from Bolivia and Ecuador seem to design requests for verification more often using tag questions (for example, ¿no? 'right?').Interestingly, the set of recordings of Spanish-speaking immigrants in Germany shows a very similar pattern to the one observed for Argentina and Colombia (more mutual gaze and less tag questions), suggesting that gaze behaviour in the data from Bolivia and Ecuador seems to be rather exceptional among speakers of Spanish.To conclude the analysis, I will return to a qualitative analysis of two deviant cases to further contextualize the implications of quantitative results (5.2).

5.1
Requests for Verification across Varieties of Spanish: General Trends Excerpt (1), in the introduction, presented an instance of a request for verification in the data from Buenos Aires (Argentina).In that excerpt, gaze behaviour plays a central role in making the request for verification recognizable for the knowing recipient.On the one hand, María gazes at Juan and establishes mutual gaze with him before initiating her request.This way, she alerts Juan that some sort of response from him (and not Pedro) might become relevant.On the other hand, sharing mutual gaze with Juan allows María to use a facial expression to display uncertainty about a potentially verifiable element.Similar gaze behaviour can be observed in requests for verification in the data from Bogotá (Colombia).
As an illustrative example consider excerpt (2).Ana (ANA) and Mario (MAR) are telling their niece, Oriana (ORI), about a family trip they made to a country house.They arrived at night and it was already dark.This caused Ema, who is introduced in line 01, to fall to the ground right on her belly.Ema was pregnant at the time.As opposed to excerpt (1), we are now at the beginning of the story and the participants are locating the event for the story recipient.
(2) En enero (sorj201803, Colombia) ANA: +*esa misma noche cuando lleGAmos que íbamos con (.) +con#1 Ema, (0.4) that same night when we arrived that we were with with ema ana +gazes up front- ----------------------------------- verified emphatically.Afterwards, the verification sequence is expanded with a second request, now to determine in which pregnancy month Ema was at the moment of the event (ll.07-10).This is done in reference to the birth of the baby (Alma), which took place around April.That way, they attest that Eva was around 7 months pregnant, which is an important fact later in the story (after the transcript), when she fell on her belly.After the verification sequence, Ana continues with the story, withdrawing her gaze from Mario and addressing Oriana again (l.12).
If we compare excerpts (1) and (2), we observe some similarities and some differences.In both cases, the speaker establishes mutual gaze with a knowing recipient and displays uncertainty about some verifiable element, which mobilises a verification from this recipient.Interestingly, establishing mutual gaze happens prior to the display of uncertainty, thus alerting the knowing recipient that coparticipation might be potentially relevant before indicating exactly in what way.Moreover, "uncertainty" is displayed both through linguistic elements, such as creo ('I believe') or tal vez ('maybe'), and facial expressions, frowning the eyebrows and tilting the head.The two excerpts differ mostly in the timing of the response by the recipient.In (1), the verification is provided immediately after the verifiable element is uttered, while in (2) the speaker designs the request including a tag question with an address term and also provides more information until the recipient responds.
Requests for verification by speakers from Bolivia and Ecuador present some differences, especially in terms of gaze behaviour.As opposed to the rest of the participants in the data, these speakers rarely establish mutual gaze with the knowing recipient.In what follows, I present two instances, one from Cochabamba (Bolivia) and one from Chirihuasi (Ecuador), to illustrate the findings regarding these two groups of speakers.
In (3), Eva (EVA) and Carlos (CAR) are telling their friend, Tato (TAT), about the first time Eva got pregnant.The pregnancy occurred right after a young kid named Juani, whom they had temporarily taken care of, had been adopted by another family, which is why Eva was feeling very sad.
(3) A la semana (ssot201708, Bolivia) After explaining that Juani had to be taken to his adoptive family (before the transcript), Eva addresses Tato with her gaze and says that she was very sad (l.01).She continues by animating her thoughts back then (l.02, see Ehmer, 2011) and adds that she cried every time she thought of Juani (l.03).Note that Carlos does not gaze at Eva during the telling, as opposed to Juan in excerpt (2) and Mario in (3).Nevertheless, he still displays hearership by producing a continuer (l.04), which shows that he is monitoring the telling.Subsequently, Eva moves to a new segment of the story, using cuando ('when'), but interrupts her turn to request a verification from Carlos on how much time elapsed after Juani's departure and before she got ill due to the pregnancy (l.06).
Eva makes her request recognizable as such by displaying uncertainty using the epistemic marker creo ('I believe' , l.06), by gazing at Carlos (#2) and by including a tag question at the end of the turn (no, 'right' , l.06).It is important to note that the request occurs prior to Eva introducing the topic of her pregnancy.Thus, the verification of this request is only possible if the recipient already has previous knowledge of the narrated events.This way, the verification sequence is a resource for Eva and Carlos to publicly display their shared knowledge for the story recipient (see also Lerner, 1992).Subsequently, Eva ratifies the verifiable in overlap (l.08) and continues the telling, revealing that she started to feel ill (l.09).She returns her gaze to Tato (l.09, #3), who responds with ya ('alright'), displaying access to the teller's stance by using a smiley voice (l.09).
The design of the request in (3) has some differences to those in (1) and (2).Although Eva does direct her gaze to Carlos, she does so only after displaying uncertainty with creo ('I believe').Furthermore, Eva and Carlos do not engage in mutual gaze during the request, since Carlos is not gazing back.Moreover, the request is designed with an integrated tag question in turn final position, which was only present in excerpt (2), where it was uttered in mid-turn position.These differences are not oriented to as problematic, since Carlos quickly verifies in line 07 and, after ratifying in line 08, Eva continues with the story.
Requests for verification by speakers from Chirihuasi (Ecuador) are designed in a similar way to those in the data from Bolivia.Excerpt (4), below, is a case in point.Nelly (NEL) and Fernanda (FER), two friends who used to be neighbours, tell Daniel (DAN) about how they moved houses after marrying.-->*gazes at FER--> nel $, ,,,,,,,,,,,,,,,,, As part of an ongoing telling, Nelly explains that, after marrying, she and Fernanda went to live with their husbands (ll.01-02).She gazes outside and points towards where she used to live (l.03) and where she lives now (l.04).Subsequently, she mentions where Fernanda lives (l.05), pointing at her with the open hand (#1).Note that Nelly holds this deictic gesture during the silence (l.05), possibly alerting Fernanda that further talk about her could follow.Indeed, after allowing a pause to occur, she turns her gaze to Fernanda and requests verification on the fact that Fernanda used to live in the upper part of the village (l.06, #2).She makes her request recognizable by gazing at Fernanda and including a turn final tag question with rising intonation and a pitch jump.As was the case in (3), mutual gaze is not established, since Fernanda is gazing down.However, this is not problematic.Rather, Fernanda responds immediately using the token mhm with rising-falling contour (l.07), which in Andean Spanish does confirmation while displaying an upgraded epistemic stance (Dankel and Soto Rodríguez, 2022, pp. 143-146).Notice that the request refers to a B-event (Labov and Fanshel, 1977), i.e., the information falls into the recipient's knowledge domain.After a silence and a middle-distance gaze towards Nelly (l.08), she continues to relate her side of the story with more -->+gazes at DAN-->> Excerpt (4) shows that requests for verification do not always condition the response to a verification (or a denial).They create an action space where knowing participants negotiate coparticipation in the telling (Dressel and Satti, 2021, pp. 63-68), this way providing an opportunity to claim the floor and render one's own part (Lerner, 1992, pp. 265-266).Still, by initiating her turn with a verification, Fernanda treats the previous action as a request and the fact of living in the upper part of town as a verifiable element.What excerpts ( 3) and (4) have in common is that, although the speakers do gaze at the knowing recipient, they do not establish mutual gaze, since in both excerpts the recipient is gazing down.This is not problematic and in both cases the recipient initiates the response without any delays.One central resource for mobilising a verification in both excerpts is the use of a turn final tag question with rising intonation (no, 'right').
The differences in the design of requests for verification across the data are systematic.In terms of gaze behaviour (figure 8), speakers usually gaze at recipients during the request, which is a resource to alert the current addressee of the telling of a change in the participation framework.However, participants from Bolivia and Ecuador establish mutual gaze significantly less frequently during these requests (x2 (2, N = 320) = 103.1,p < .001).5 Furthermore, by including the data recorded in Germany, which comprises speakers from different 5 For quantitative purposes, I have grouped Argentina and Colombia, on the one hand, and Bolivia and Ecuador, on the other.A larger sample size would be needed to determine whether there are significant differences within each group.varieties, we see that the gaze behaviour in the data from Bolivia and Ecuador seems to be rather exceptional among speakers of Spanish.Instead, this sort of gaze behaviour is similar to the gaze behaviour of Quechua speakers (Satti and Soto Rodríguez, 2021;Muz, 2023) and other indigenous communities in America, like Tzeltal speakers (Rossano et al., 2009).
The lower frequency of mutual gaze can be attributed to different practices for monitoring the current speaker (figure 9).While in our recordings from Argentina, Colombia and Germany the knowing recipient frequently gazes at the speaker as a way of monitoring, as we would expect based on previous studies on gaze behaviour (see Kendon, 1967;C. Goodwin, 1980), in our data from Bolivia and Ecuador, the knowing recipient frequently gazes elsewhere and, instead, may display hearership by using other practices, such as continuers.This has a consequence on the availability of recipient's gaze.In the first case, when the speakers turn their gaze to the knowing recipients, recipient gaze is readily available for establishing mutual gaze, which also allows for the use of facial expressions for displaying uncertainty.In our examples from Bolivia and Ecuador, the speakers direct their gaze to the knowing recipients, but since they are not gazing back, mutual gaze is not established.
Another interesting difference is that speakers from Bolivia and Ecuador design requests for verification more frequently using tag questions (figure 10, x2 (2, N = 320) = 63.2, p < .001).6This type of resource, which is also a response mobilising feature of turn design (Stivers and Rossano, 2010), allows 6 In Spanish, the most frequent lexical element deployed as a tag question is "no" (Uclés Ramada, 2018; Quartaro, 2021).Beyond this element, which is considered to be present across all varieties, other frequent elements are ¿Sabes?, ¿Verdad?, ¿Cierto?, among others, depending on the regional variety (see Uclés Ramada, 2018).for the inclusion of rising intonation and does not require the establishment of mutual gaze to be perceived by the recipient (as opposed to facial expressions).Furthermore, tag questions are typically included in turn final position, which contributes to making the end of the speakers' turn recognizable.Therefore, in combination with gaze direction and recipient tilted epistemic asymmetry, they allow knowing recipients to start responding without delays.In sum, data shows that there is a connection between the establishment of mutual gaze and the inclusion of tag questions: speakers who establish mutual gaze during the verification sequence tend to design the request without tag questions (and vice versa).

5.2
The Use of Tag Questions While Sharing Mutual Gaze In this section, I present two instances of requests for verification where participants establish mutual gaze, but also design the request including a tag question.These can be considered deviant cases of the general trend presented in the previous section.This way, I further contextualize the implications of the quantitative results and discuss some differences in the use of tag questions across the groups of speakers included in the study.
it was very nice (we had a very good time?) ain: -->*ga.at PED*gazes at Pablo-->> pab: +middle distance gaze-->> At the beginning of the excerpt, Ainara provides Pedro with background information (ll.01-04) and he responds with continuers (l.03 and l.05).Subsequently, Ainara introduces the topic of Pablo's first trip to Europe (l.06,#1) and goes on to mention different places they visited (l.07).Note that Ainara initiates repair, allows a pause to occur and produces and audible click before introducing these places.Furthermore, she gazes at Pablo (#2), although he does not provide any type of visible response.After mentioning two cities by name (Madrid and Barcelona), she allows another pause to occur and utters a hesitation marker (eh) before introducing another place, although this time referring a country and not a city (Italia).After another short silence, and while sharing mutual gaze with Pablo (#3), she utters a tag question with rising intonation (no, 'right').Pablo confirms (l.09), but goes on to add another city they visited in Spain (San Sebastián, l.10).He also explains that they met with a friend there (l.11-12), and that they had a good time (l.14).This way, he shows his orientation to this visit as an omitted element in Ainara's list, which is understood by her as such by using a change of state token and an affirmative particle (l.13), displaying that now she does remember.
The use of the tag question by Ainara in l.07 is interesting for two reasons.On the one hand, it can only be explained by taking into account the emergent character of requests for verification.Ainara encounters problems while listing the places they visited, which is evidenced by multiple hesitation markers and pauses.By gazing at Pablo in two occasions and establishing mutual gaze with him, she creates an opportunity space for Pablo to claim the floor and verify or collaborate with the list.However, Pablo does not respond in visible ways and, after a silence, Ainara includes the tag question, this way mobilising a response from Pablo.In other words, the tag question is included in an emergent way to deal with a possible lack of response by Pablo and to make the request for verification explicit.On the other hand, we notice that the use of the tag question in excerpts (3) and (4), from Bolivia and Ecuador respectively, is different in terms of its prosodic realisation.In these cases, the tag question is prosodically integrated into the design of the request.Figure 12, below, provides quantitative evidence that this aspect is actually systematic in the data.
Figure 12 shows all instances of requests for verification that include a tag question (n = 108).In Bolivia and Ecuador, tag questions are more frequently prosodically integrated into the turn, while in the rest of the data they are more frequently produced as an independent intonation unit (x2 (1, N = 108) = 12.5, p < .001).These instances, like in (5), are usually related to pursuits of response (Pomerantz, 1984).Moreover, the fact that tag questions are so frequently prosodically integrated in requests in Bolivia and Ecuador could be interpreted as a sign of sedimentation of tag questions in those varieties, although more research is needed to confirm this.
Lastly, I present a request for verification that stems from the Colombian data, where we observe the use of a prosodically integrated tag question.In (6), Diana (DIA) and Paula (PAU) are telling their friend Oriana (ORI) about the bachelorette party of a mutual friend, which started with a breakfast in a town called Subachoque, outside of Bogotá.If we focus more closely on the design of the request, we notice that line 04 may be considered a first attempt at requesting verification.Diana already hesitates and turns her gaze to her friend before introducing the verifiable element.However, in that moment, Paula is visibly not looking at her (#2).Subsequently, she repairs the design of the request integrating a tag question.This way, she mobilises a response from Paula and establishes mutual gaze with her, which also allows for the use of a facial expression, namely frowning the eyebrows while producing the verifiable element (#3, l.05).In other words, the excerpt shows that participants use resources that are available to them in the local ecology where the interaction takes place.As we can see in this example, gaze can also be visibly unavailable due to different local contingencies, even among participants who establish mutual gaze more often, as is the case in the data from Colombia.When this happens, tag questions are still a useful resource for mobilising a response after introducing a verifiable element.
In this section, I presented two instances that contradicted the trend "if mutual gaze, then no tag question".On the one hand, these excerpts show that tag questions seem to be more sedimented in the design of requests in Bolivia and Ecuador, while in the rest of the data, they are used more often as a way of dealing with the emergent character of the request and a possible lack of response by the recipient.On the other hand, they show how linguistic and embodied resources are sensible to the interactional contingencies where the request takes place.This way, the comparative approach proposed here attempts to deal with the challenge of accounting both for local ecologies and sedimented routines in the deployment of multimodal resources.

Discussion and Conclusion
This article has shown how participants make requests for verification recognizable for others across different varieties of Spanish.What all instances have in common is that speakers shift their gaze direction from the unknowing to the knowing recipient.This way, they make a change in the addressee visible for both recipients and alert the knowing recipient that coparticipation from them might become relevant.Furthermore, they establish a momentary recipient-tilted epistemic asymmetry with this recipient, which is achieved by displaying uncertainty towards a verifiable element.For this purpose, speakers deploy different resources to downgrade their epistemic stance, such as epistemic markers, facial expressions, tag questions or rising intonation.Data revealed differences in gaze behaviour across the collection.Although all speakers frequently gaze at the knowing recipient, speakers from Bolivia and Ecuador rarely establish mutual gaze with them.This can be attributed to different practices for monitoring the current speaker and the availability of the recipients' gaze.While in Argentina and Colombia the knowing recipient frequently gazes at the current speaker, in Bolivia and Ecuador they frequently gaze elsewhere during requests for verification.Interestingly, participants from Bolivia and Ecuador use their gaze in a way that is comparable to Tzeltal communities (Rossano et al., 2009) and some aboriginal communities in Australia (Blythe et al., 2018).The main contribution here is that differences in gaze behaviour are fairly independent of the language of talk, thus providing evidence that gaze behaviour has a cultural component (Rossano et al., 2009), since all participants in the data interact in Spanish during the recordings.Furthermore, the analysis showed that differences in gaze behaviour are intertwined with other response-mobilising features of turn design.Eliciting a response by a particular recipient is arguably a universal interactional task.However, there are different resources for doing so, and there seem to exist some systematic differences among the groups of speakers under study.In this article, I focused on the practices for establishing epistemic asymmetry and for dealing with lacks of response.Establishing an epistemic asymmetry among two knowing participants requires a display of uncertainty by the current speaker.The analysis showed that, while speakers from Argentina and Colombia use facial expressions, speakers from Bolivia and Ecuador more frequently use prosodically integrated tag questions with rising intonation.Moreover, tag questions in Argentina and Colombia are more frequently used as a resource to deal with a lack of response, while in Bolivia and Ecuador they seem to be specialised for epistemic downgrading and turn taking.Still, linguistic and embodied resources are sensible to local interactional contingencies and they are deployed by all speakers in a situated manner.In other words, the comparative approach proposed here tried to address the challenge of accounting both for situated uses and sedimented routines in the deployment of linguistic and embodied resources.
In terms of the method used for the comparison of gaze behaviour, the article argued in favour of an action-based approach.Gaze patterns vary greatly depending on the course of action within which gaze is deployed (Rossano, 2013) and according to the participation framework of the encounter (Auer and Zima, 2021).Therefore, if we aim to compare gaze behaviour across certain groups of speakers, an action-based approach allows to deal with these challenges.In fact, although experimental research has contributed greatly to our knowledge of variation in gaze behaviour, recent studies suggest that, beyond more reliable experimental designs, we also need more studies that approach cultural differences in gaze behaviour that take into account "the dynamic nature of social interactions" (Haensel et al., 2022: 110).
There is a series of open questions and a lot of potential for future research.First, the quantitative analysis was based on a binary annotation of whether participants gaze at each other's faces or not.However, there are multiple ways of gazing at other copresent individuals.For instance, glances, middle-distance gazes and peripheral vision could play a central role in Bolivia and Ecuador.Future research could therefore focus more on the different ways of gazing and what implications those may have for interaction and language.Second, the differences observed in the analysis cover a pre-arranged setting and a rather specific action.More research on other participation frameworks and different seating (or mobile) positions is needed to confirm the results presented here.
Third, alerting a non-addressed recipient of potential coparticipation is a complex multimodal and sequential accomplishment.The management of silence in Bolivia and Ecuador could be of central importance, although further analysis is needed (see Mushkin and Gardner, 2009).Lastly, prosody is another resource available to speakers that should be considered in more detail.
In conclusion, this article tackled differences in gaze behaviour in Spanish with a multimodal and action-based approach.Gaze is deployed in the local ecology where social interaction takes place and, therefore, it is best considered from a holistic point of view that takes into account other aspects of conversational organisation.Requests for verification, an action type where gaze behaviour plays a central role, offered a perspicuous locus for pursuing a comparative perspective.By showing how differences in gaze behaviour are intertwined with linguistic elements, such as tag questions, the article contributed to our knowledge of the grammar-body interface (Keevalik, 2018) and questioned verbal-only views of language.

Figure 2 Figure 3
Figure 2 View of the data from the main camera

Figure 4
Figure 4 Illustration of coding parameters Figure 5 Ana gazes at Oriana (#1), who is sitting on her right (out of frame).Subsequently, she gazes at Mario and requests a verification regarding the date of the trip (#2 and #3).

Figure 7
Figure 7 Nelly (right) gazes at Fernanda (left) and requests verification from her (#2).After verifying, Fernanda continues the telling towards Daniel (#3), who is sitting in front of them (out of frame).

Figure 8
Figure 8 Gaze behaviour during requests for verification (aggregation of cases)

Figure 9
Figure9Abstract representation of knowing recipient's gaze behaviour."A" refers to current speaker, "B" refers to knowing recipient and "C" refers to the researcher (unknowing recipient).

Figure
Figure 10 Requests for verification including a tag question

Figure 12
Figure 12 Prosodic integration of tag in the design of the request for verification

Figure 13
Figure13Diana gazes at Oriana (#1), who is sitting on her left (out of frame).Subsequently, she shifts her gaze towards Paula, who is visibly not gazing back (#2).