Request for confirmation sequences in Castilian Spanish

2.1 Requesting confirmation in spoken Spanish

Spanish RfCs have not been dealt with intensively under an interactional linguistic and/or conversation analytic perspective.^[1] They have rather been studied more generally as a type of ‘polar questions’ or ‘total interrogatives’ (as opposed to ‘partial interrogatives’/question-word questions and alternative questions) that may be used for different purposes (Escandell Vidal 1999, 3934).

Spanish does not have a syntactic device like word order to distinguish between declaratives and total interrogatives as sentence modes. It is commonly assumed that constituent order in declarative sentences and polar interrogatives is flexible and that it is intonation that allows us to distinguish between them. Escandell Vidal (1999, 3951–3) takes the position that inversion (e.g., verb–subject) is the unmarked/neutral constituent order in total interrogatives, which – in the terminology of this article – have to be considered requests for information. Interrogatives without inversion (e.g., subject–verb order), in contrast, have to be interpreted as presenting propositions that are attributed to others, typically the interlocutor. This attribution, depending on the context, gives rise to the interpretation that the proposition needs to be confirmed by that same interlocutor. Such ‘interrogative attributions of a proposition to an interlocutor’ overlap considerably with the notion of RfC in the present article. Please note, however, that (non)inversed constituent order is not recognizable in every clause, since Spanish is a pro-drop language.

Besides syntax prosody has been analyzed to play a crucial role in formally marking the difference between declaratives and total interrogatives. Traditionally, it has been assumed that in Castilian Spanish declaratives are characterized by a final fall contour and total interrogatives by a fall-rise contour, starting on the pitch accent.^[2] Other contours in total interrogatives are however possible, too (Escandell Vidal 1999, 3948–50). Of particular importance is the ‘circumflex’ contour, which is characterized by a continuous rise and a fall to low that starts on the focal accent. While this contour has often been analyzed as being affective, often expressing doubt or disagreement (Fernández Ramírez 1951, 1959), Escandell Vidal (1998, 1999) proposes the (relevance-theoretic) analysis that this contour signals that the speaker attributes the proposition in the question to another person, typically an interlocutor, just as inversed constituent order.^[3] Following the author, typical contexts are turn continuations but also contexts in which an interlocutor has expressed the content before or where it may be arrived at by means of inference. Escandell Vidal (1999, 3947) also states that the circumflex contour is incompatible with requests for information.

Negation in total interrogatives has often been interpreted as the speaker expressing a particular stance toward the proposition, in particular, a deviation from their expectation(s) (e.g., Fernández Ramírez 1959, 245) and has also been analyzed to be merely expressive. Escandell Vidal (1999, 3958), however, distinguishes between internal and external negation. Internal negation forms part of the prepositional content (a negative predication). External negation, in contrast, is not part of the predication but rather introduces an affirmation; e.g., it signals that the speaker presents a proposition (s)he holds to be true. Therefore, external negation is a means to signal that a declarative should be interpreted as a total interrogative or, more specifically, as an RfC.^[4] A typical case of external negation is the negative tag no. Concerning the interaction of word order and negation, Escandell Vidal (1999, 3953) points out that only interrogatives with inversion (e.g., verb–subject order, i.e., RfCs) allow for negative polarity items (otherwise they would have to be considered neutral questions, i.e., requests for information).^[5]

As in other languages, RfCs in Spanish may be introduced by conjunctions such as que ‘that’, si ‘if’, and como (que) ‘as (that)’. For instance, que ‘that’ has also been analyzed by Escandell Vidal (1999, 3967) as a means to attribute the (contents of) the utterance to an interlocutor (thus marking the utterance pragmatically as an RfC). The conjunction si ‘if’ is the Spanish marker for signaling indirect questions. The connectives que and si may also be combined. The conjunction como (que)) ‘as (that)’ also serves to attribute the proposition to an interlocutor, but in addition signals that the speaker is taking a negative stance (‘rejection’ in the sense of Escandell Vidal 1999, 3969).

2.2 Question tags in spoken Spanish

Question tags – or interrogative tags – play an important role in the design of RfCs. A prototypical tag question in Spanish consists of a declaratively formatted confirmable – typically a clause – followed by an interrogative tag. Interrogative tags have been addressed under a variety of terms, among them apéndices comprobativos ‘confirming appendixes’ (Martín Zorraquino and Portolés 1999, 4188, Ortega Olivares 1985, 1986), marcadores discursivos interrogativos ‘interrogative discourse markers’ (Porroche Ballesteros and Laguna Campos 2015), and partículas interactivas con valor epistémico ‘interactive particles with epistemic value’ (Fernández Sanmartín 2009).^[6] Interrogative tags may be derived from a variety of formal structures, among them response tokens (e.g., no ‘no’, sí ‘yes’), verb phrases (e.g., sabes ‘(do you) know’, entiendes ‘(do you) understand’), noun phrases (e.g., verdad ‘truth’), prepositional phrases (e.g., de acuerdo ‘of agreement/do you agree’), and interjections (e.g., eh, ah). The Spanish inventory of tags also includes disjunctive tags (Gómez González 2014, 98–9) that contain the coordinating conjunction o ‘or,’ such as o no ‘or not’ and o qué ‘or what’. In contrast to other languages – for instance, English – tags in Spanish are invariant (Gómez González 2014). It is acknowledged that single tags are polyfunctional and may serve further interactional functions in addition to requesting. For a detailed analysis of no ‘no’, also addressing possible combinations with other linguistic elements, refer to Montañez Mesas (2008). The inventory of tags used to request confirmation differs across varieties of Spanish (Escandell Vidal 1999, 3958). Concerning intonation, Martínez and Domínguez (2006) find a rising contour to be typical for the tag no (in speakers from Venezuela), and Valenzuela Farías (2013) finds this to be the case for question tags in general (in speakers from Latin America).

There are few differential studies focusing on the functional potentials of different interrogative tags (but see for example Ortega Olivares 1985, 1986). Comparing no ‘no’ and eh ‘eh’, García Vizcaíno (2005) finds that the dominating function of no is to request the confirmation of a piece of information (or an opinion), while the dominating function of eh is to reinforce the epistemic stance toward the proposition (see also Rodríguez Muñoz 2009).^[7] In a cross-linguistic comparison, Gómez González (2014) observes that canonical tag questions are less frequent in English than in Peninsular Spanish (and Portuguese).

2.3 Response tokens in spoken Spanish

RfCs are first pair parts that make a (dis-)confirmation relevant as an interlocutor’s next action. Responses may be realized in different ways, typically as type-conforming responses (Heritage and Raymond 2012, Raymond 2003) that give a yes/no-type answer.

By producing a yes/no-type response to an RfC, responders accept and comply with the grammatical and pragmatic constraints set by the request. In contrast, in the case of non-type conforming responses, responders do not abide by the relevant constraints but rather take issue with some aspect of the RfC (e.g., embedded presuppositions, assumed epistemic asymmetries, etc.) (Enfield et al. 2019, Heritage and Raymond 2012, Stivers 2022).

Using a response token is the prototypical way of realizing a type-conforming response. Unsurprisingly, in Spanish, the positive polarity token sí ‘yes’ and the negative polarity token no ‘no’ are the most common forms, which are traditionally analyzed as adverbs.^[8] The general usage of response tokens to (dis)confirm an RfC is quite straightforward. In RfCs with positive polarity, tokens with positive polarity are used for confirmation, and negative polarity tokens are used for disconfirmation (Sadock and Zwicky 1985). In RfCs with negative polarity, the reverse is true: confirmation is realized with a negative polarity token, while disconfirmation is realized with a negative polarity token. Table 1 presents a schematization (Section 5.2).

Response tokens in Spanish are traditionally subsumed under different linguistic categories such as adverbs (e.g., no ‘no’, sí ‘yes’), discourse markers (e.g., claro ‘sure’), and non-lexical vocalizations (e.g., hm ‘hm’, ɁhmɁhm ‘hm hm’). In analyzing sí ‘yes’, Dumitrescu (2007) finds that when used as a response particle, it is often combined with other discourse markers rather than occurring as a single item, and Gambetta (2018) shows that Spanish speakers also reduplicate sí in responses. There is, however, surprisingly little research comparing the different functional profiles of response tokens^[9] and other conventionalized ways of responding to RfCs (such as no sé ‘I don’t know’, ni idea ‘no idea’, etc.). A notable exception is the work of Martín Zorraquino and Portolés (1999), who treat expressions such as por supuesto ‘of course’, naturalmente ‘naturally’, claro ‘sure’ and sin duda ‘without doubt’ as evidential discourse markers that convey more than a simple confirmation, since their main function is to reinforce the assertion and thus realize an upgraded confirmation in the sense of Stivers (2019, 2022).^[10]

4.1 Syntactic resources

4.1.2 Clausal RfCs

As detailed in Section 2.1, Spanish does not offer a specific syntactic format for polar interrogatives, as for example English or German, where the finite verb is placed in first position. This means that there is no principled syntactic distinction between declaratives and polar interrogatives in Spanish.

Clausal RfCs are typically formatted as single clauses. Extract 1 illustrates a very basic RfC. The participants are talking about going out to a particular bar. One of the participants, R, requests confirmation from participant L that she will be going out, too.

The RfC in l. 02 is realized by a single clause, consisting of a personal pronoun and a finite verb, followed by a tag.

Among all clausal RfCs only 9.5% consist of clause combinations (n = 15). Besides only two coordinated clause combinations (all with y ‘and’), all others are subordinate clause combinations (n = 13 of 15). Among subordinated clause combinations the most frequent type are adverbial clauses (n = 7) followed by object clauses (n = 4, with verba dicendi or pensandi).

Besides clause combinations involving a connective, asyndetic clause combinations in RfC turns are also attested (n = 7). In those instances, however, the first clause does not belong to the confirmable as such, but rather only establishes an epistemic framing (e.g., no me acuerdo cómo era, ‘I don’t remember how it was’) for the ensuing clause that actually presents the confirmable.

A common function of clausal RfCs is to formulate upshots and make explicit inferences from prior talk. In Extract 2, R talks about her new crush, on whom she wants to make a good impression.

In l. 01-04, R states that she is in (emotional) trouble and that she even told her new crush that she had a driving license. The use of encima ‘even’ as a mirative/modal particle (Fuentes Rodríguez 1987, 109, Garachana 2013, 157) in combination with the ensuing adverbial purpose clause with por ‘in order to’ licenses the inference that R actually does not have a driving license. In l. 05, C requests confirmation for this inference, which is confirmed by R in l. 06.

While this extract illustrates an inference about a discourse participant, Extract 3 illustrates an inference about a third party, namely fictional characters in the television series ‘Los hombres de Paco’. R narrates a quarrel that takes place in one episode between two characters, Sara y Lucas.

Based on R’s narration of the quarrel in l. 01-06, participant C arguably draws the inference that the two fictional characters are engaged again, since she formulates this inference in l. 07, prefaced with the change of state token ah (Vázquez Carranza 2014, 137–69, 2016; for English oh see Heritage 1984, 1998, 2002b, 2018, Vázquez Carranza 2014, 2016) (for English oh cf. Heritage 1984, 1998, 2002b, 2018).

4.1.3 Phrasal RfCs

After clausal formats, the second most frequent format for realizing an RfC is phrases (22.8%, n = 57 of 250). Table 3 gives an overview of different phrase types in RfCs in the collection.

Table 3

Types of phrases in RfCs

Type of phrase	Absolute frequency	Relative frequency (total n = 57) (%)
Noun phrase	32	56.1
Prepositional phrase	19	33.3
Adverbial phrase	5	8.8
Adjective phrase	1	1.8
Σ	57	100

The most frequent type of phrases that occur as RfCs are noun phrases (56.1%, n = 32 of 57) followed by prepositional phrases (33.3%, n = 19 of 57). Phrasal RfCs are often used for the clarification of (temporal, locational, and personal) reference. In Extract 4, the RfC is used for clarifying a personal reference. The participants are talking about personal relations among acquaintances.

In l. 01-02, C refers to two men, Jonás and Axel. In l. 04, R requests clarification of personal reference by using a pero-prefaced noun phrase (consisting of the proper name Axel) and the immediately juxtaposed complex noun phrase el de maRIsa ‘Marisa’s’, that together form a left dislocated structure (Lambrecht 1994, Pekarek Doehler et al. 2015). The RfC is answered by C with a correction, by reusing the syntactic structure and replacing the proper name Marisa by Clara (l. 05).

Phrasal RfCs are also very often used to specify quantities, as in Extract 5. Participant X is talking about the working hours and salary of a job that he has. L and R are commenting on and requesting information about the job.

In l. 02, L requests the information on how much X is being paid, to which X responds in l. 03 mentioning the amount of sixty Euros (per day). In l. 06 R requests confirmation about the working time per day by producing the noun phrase tres HOras; ‘three hours’. This temporal quantity is confirmed by X in l. 07.

4.1.4 (In)subordinate RfCs

It is noteworthy that 10.4% (n = 26 of 250) of the RfCs in the data set are freestanding clauses that are introduced by a (originally) subordinating connective. Table 4 gives a quantitative overview of the connectives used.

Table 4

Connectives introducing subordinate clauses in RfCs

Connective	Absolute frequency	Relative frequency (total n = 26) (%)
que ‘that’	18	69.2
si ‘if’	3	11.5
cuando ‘when’	2	7.7
como ‘like’	1	3.8
porque ‘because’	1	3.8
por ‘because (of)’	1	3.8
Σ	26	100

Table 4 shows that the most frequently used connective is que ‘that’ (n = 18), amounting to more instances than all other subordinating connectives combined (n = 8), which are all adverbial.

Concerning que-introduced (in)subordinate constructions, Gras and Sansiñena (2015) show that their central function is to indexically refer to contextually relevant information that can be inferred or retrieved from previous turns or discourse situations.^[11] Extract 6 gives an example of a que-introduced RfC. The participants are talking about a night out with a common friend, Norbi.

In l. 01, L states his assumption that Norbi would be going to sleep at an early moment of the night out (l. 01-02). R emphatically rejects this assumption (l. 03) and adds the account that going to sleep would have meant ‘ending’ something (l. 05-06), but without stating explicitly what event or opportunity would have ended. In overlap with the account, L produces the que-introduced clause, expressing that Norbi is flirting a lot. Arguably, with this turn L is presenting an inference, based on the prior turn and their shared knowledge about their friend. In addition to introducing the clause with que, the tag no serves to mark l. 07 as a RfC. Importantly, L immediately modalizes and downgrades the epistemic stance taken by adding =por lo VISto° ‘so it seems’ in l. 08. Nevertheless, the RfC is responded to directly by the other participants and L further downgrades his assumption by o NO; ‘or not’ (l. 10). At that point, X responds with a humorously framed counter question, with whom Norbi is flirting (l. 11). R then responds too, confirming L’s RfC with [ < < pp > SÍ. >] (.) con [TOdas;] ‘yes, with all (girls)’, thereby also responding to X’s humorous question. Note that although R and X have started responding to his RfC, L still backs down from his epistemic stance with o tanto NO; ‘or not so much’ (l. 13). Thus, this extract also illustrates how participants pursue a response and yet continue to modalize their epistemic stance after a first realization of their RfC.^[12] In addition, the example illustrates the use of RfCs to proffer a new conversational topic.

4.1.5 RfCs in as candidate answers after open questions

This section deals with a specific sequential context in which RfCs frequently occur in the collection, namely immediately following an open question produced by the same speaker. In this sequential environment, the RfC presents a candidate answer (Pomerantz 1988) to the preceding question. This is the case in 16.4% of all RfCs in the collection, as detailed in Table 5.

Table 5

RfCs (not) preceded by an open question

	Absolute frequency	Relative frequency (total n = 250) (%)
RfC…
… preceded by an open question (as a candidate answer)	41	16.4
… not preceded by an open question	209	83.6
Σ	250	100

Extract 7 includes a pertinent example of an RfC in the position of a candidate answer. The participants are talking about the date when the date when classes resume after some upcoming holidays.

The extract starts with a request for information from R involving a partial interrogative (l. 01-03), which C answers by giving the date of the next class (l. 04). R then continues to ask for the corresponding day of the week by producing another partial interrogative qué ES; = ‘what is (it)’ in l. 06, immediately followed by the formulation of the weekday = MARtes; ‘Tuesday’ (l. 07). The mentioning of ‘Tuesday’ can be interpreted as an inference or guess of the weekday of the next class, based on the previously mentioned date and their (presumably) shared knowledge about the upcoming holidays. Sequentially, l. 07 presents a candidate answer to the immediately preceding question, itself still being confirmable. In fact, l. 07 is treated as an RfC by the other participants, both confirming the request. C is first to answer in an epistemically downgraded fashion (l. 08-09) followed by L, who gives an unmitigated response (l. 11) combined with an account of the weekdays that comprise the upcoming holidays (l. 12–15).

Regarding the syntax of the RfC (the candidate answer) in l. 07, it is important to note that it comes in the form of a noun phrase. In fact, RfCs after open questions exhibit a strong tendency to come in a phrasal format. This is illustrated by Table 6, which shows the frequencies of different syntactic formats according to the (non)presence of an open question immediately before the RfC.

Table 6

Syntactic design of RfCs that are (not) preceded by an open question

	Absolute frequency			Relative frequency (total n = 41)
RfC […]	[…] preceded by an open question	[…] not preceded by an open question	Σ	[…] preceded by an open question (%)	[…] not preceded by an open question (%)	Σ (%)
Clause (or clause combinations)	14	142	156	34.1	67.9	62.4
Phrase	23	34	57	56.1	16.3	22.8
(In)subordinate clause	3	23	26	7.3	11.0	10.4
Other	1	10	11	2.5	4.8	4.4
Σ	41	209	250	100	100	100

Table 6 shows that after an open question RfCs are typically produced in a phrasal format (56.1%, n = 23 of 41), whereas if there is no immediately preceding open question, only 16.3% (n = 34 of 209) of RfCs are phrasal. In contrast, clausal formats in RfCs are rarer after an open question (34.1%, n = 14 of 41), whereas clausal formats are the preferred form (67.9% n = 142 of 209) when the RfC is not preceded by an open question. Put differently, whereas clausal formats are generally the preferred format for RfCs, after an open question phrasal formats are preferred. Other formats do not seem to vary much depending on the (non)presence of an open question immediately before the RfC.

4.2 Polarity

The vast majority of RfCs in the collection are formatted with positive polarity (83.2%, n = 208 of 250). Consequently, negative polarity is less frequent (16.8%, n = 42). Table 7 gives an overview of the linguistic resources used for marking negative polarity in the collection.

Overwhelmingly, the negative particle no ‘no’ is used for marking negative polarity. Other resources, such as the adverb tampoco ‘neither’ and the negative indefinite pronoun nada (de) ‘none/nothing (of)’, are used very infrequently (each only in 0.4% of the cases, n = 1 of 250).

An example of a negatively polarized RfC is given in Extract 2, repeated here for convenience: <<:-)> no tienes [tú el carNET;] ‘you don’t have the license’ (l. 05). While this is a clausal RfC, other formats may also be polarized negatively. See the following extracts.

In Extract 8, the confirmable viaje ‘journey’ is realized as a bare noun phrase, prefaced by the connective o sea ‘well’ and negatively polarized using the negative particle NO. The incremented seGUro modalizes the epistemic stance and serves to mobilize a response. In Extract 9, the proper name/noun phrase sofía is preceded by the connective y ‘and’ negatively polarized by the adverb tamPOco ‘neither/also not’. It is important to note, however, that negatively polarized RfCs are overwhelmingly realized in a clausal format (90.5%, n = 38 of 42 negative RfCs).

In general, negation has been shown to play an important role in common ground management (e.g., Deppermann 2014). More specifically, in RfCs polarity has an impact on the preference organization (Pomerantz and Heritage 2013) for the response turn. In RfC sequences, a confirmation is generally the preferred response. In this view, negation can be considered to be a means to increase the likelihood of a preferred response (see also Heritage and Raymond 2021).That is, while in principle it would be possible to produce only RfCs with positive polarity, a negative polarization of the confirmable may give the responder the opportunity to realize a socially preferred action (confirmation) more easily, rather than a dispreferred one (disconfirmation). Table 8 shows that positively and negatively formatted RfCs roughly get the same percentage of (dis)confirmations.

A confirming response is produced in response to an RfC in 61.1% of the cases (60.1% for positively and 65.9% for negatively formatted RfCs) and a disconfirming response is produced in 27.4% (26.9% for positively and 29.3% for negatively formatted RfCs). This supports the hypothesis that requesters sensitively use negative polarization as a design feature of RfCs to increase the likelihood of a preferred response (confirmation).

4.3 Epistemic modulation

RfCs are characterized by a shallow epistemic gradient, the requester ascribing to theirself a less knowledgeable position than the interlocutor(s) addressed. Yet, as shown in Table 9, epistemic devices are used in only 7.6% of all RfCs in the collection.

Before addressing the surprisingly low use of epistemic markers, an overview of the different markers used in the collection will be presented. In Spanish, both morphological and lexico-syntactic devices may be used for epistemic modulation. Table 10 gives an overview of their frequency of use. Please note that none of the RfCs features more than one modulation marker, which also means that morphological and lexico-syntactic devices are not combined in the collection.

Lexico-syntactic resources are most frequently used for modulation. They can be grouped into predicate phrases, expressing mental states and verbal actions (42.1%, n = 8 of 19), and approximators and epistemic adverbs (31.6% n = 6 of 19), amounting to 73.7% (n = 14 of 19) of all epistemic modulation markers. Tenses as morphological resources are used in 26.3 (n = 5 of 19) of the cases.

The most frequently used morphological resource is the Pretérito imperfecto ‘Imperfect’. Its use for epistemic modulation has been described as imperfecto de cita ‘quotative imperfect’ (Bosque 2010, 23.11h), indexing that the proposition has been reported (possibly by the interlocutor). A high affinity with negation in questions has been pointed out too (Bosque 2010, 23.11i). Extract 10 is an example of the Pretérito imperfecto in a positively polarized RfC. The participants are talking about different kinds of sports.

In l. 01, R addresses C and requests confirmation for the proposition that C is an expert ping-pong player. In this context, the Pretérito imperfecto does not anchor the proposition to the past, which is evidenced by C’s response in l. 03–04, where she confirms the request using the present tense. The Pretérito imperfecto is rather used to index shared knowledge among the participants and thus serves to modalize the speaker’s epistemic stance.

In the sample of ten languages analyzed, Castilian Spanish has the fewest modulation markers (Pfeiffer et al. forthcoming). It is therefore reasonable to assume that apart from modulation markers Spanish uses other resources to signal a reduced degree of epistemic access. In this respect, it is important to note that inference markers (Section 4.4) have not been coded as epistemic markers. It can be argued, however, that inference markers are in fact a means to signal a lower degree of epistemic access (namely by inference). Table 11 shows inference marking and epistemic marking combined.

Only in very few cases are epistemic modulation and inference marking actually combined (2.8%, n = 7 of 250), which may count as an argument that their functional properties to some degree overlap, and that inference markers also modalize epistemic access. Following this perspective, only 76.8% (n = 192 of 250) of the cases come unmarked of epistemic modulation.^[13] A further observation concerns the aforementioned sequential context of RfCs produced after an open question as a candidate answer (16.4%, n = 41 of 250, Section 4.1). In this context, it can be argued that the open question already signals a low degree of knowledge. This interpretation is corroborated by the data: none of the RfCs that occur after an open question are produced with epistemic markers or inference markers. Taking open questions immediately preceding an RfC as an index of epistemic access reduces the number of RfCs that are unmarked for epistemic access (either by lexico-syntactic design or by sequential position) to 62.2% (n = 155 of 250).

4.4 Inference marking

RfCs may be used to confirm conclusions or inferences based on other participants’ prior talk (e.g., Extracts 2 and 3). Consequently, participants may, but do not need to, use inference markers in the design of their RfCs to present the confirmable as an inference. This is the case in 18.4% (n = 46 = of 250) of instances in the collection, as shown in Table 12.

Table 13 gives an overview of the frequencies of the different inference marking devices in the collection. Note that some RfCs feature more than one inference marker. Therefore, the total number of inference markers (n = 56) is higher than the number of RfCs with inference marking (n = 46).

The most frequent means to mark a confirmable as an inference is the use of an (in)subordinated que-clause (Section 4.1). The second most frequent device is o sea (que) ‘that is/so (that)’, which has been shown to have reformulative and conclusive functions (Pons Bordería 2014b).^[14] Next in frequency is the change-of-state token ah ‘oh’ (Vázquez Carranza 2014, 137–69, 2016), which – following the coding manual project (König et al. forthcoming) – is doubly coded as inference marking device and connective (see also Section 4.5). It is interesting to note that the connective entonces ‘so’ – typically treated as an inference marking device in languages such as German (cf. Deppermann and Helmer 2013) – and es que ‘it’s that’ – which has been analyzed as inferential construction (Declerck 1992, Delahunty 1995) – are used with lower frequencies.

Although different inference markers may be combined in RfCs, Table 14 shows that only one inference marking device is typically used. Only in 3.2% of the RfCs in the collection are two markers used.

A typical combination of inference marking devices is the use of the change-of-state token ah ‘oh’ as a preface, followed by other devices, as the following extracts illustrate.

Inference markers are typically placed at the beginning of the RfC. It is, however, also possible to place inference markers differently, as the following extract illustrates.

In l. 09, L incrementally produces the inference marker entonces ‘so/then’ when the RfC is already potentially completed (Auer 2007, Couper-Kuhlen and Ono 2007, Luke et al. 2012). Please note that the inference marker even follows the address term tío ‘dude’, although it could also have been incremented right after the first TCU in l. 09.

4.5 Connectives

RfCs may be introduced by connectives that link them to the prior discourse and mark their relation to the common ground. This is the case in about half of the cases in the collection (49.2%, n = 123 of 250), as shown in Table 15.

According to the coding manual (König et al. forthcoming), the category of the connective is broadly defined to include conjunctions, subjunctions, adverbials, and discourse markers. It is important to note that for Spanish there is considerable overlap between the categories ‘connective’ and ‘inference marking device’ (Section 4.4). In fact, all but clausal inference markers are connectives (n = 55). The second column in Table 16 indicates which of the connectives have also been coded as inference marking devices. Please note that the total number of connectives (n = 154, Table 16) in RfCs is higher than the number of RfCs containing one or several connectives (n = 123, Table 15), since connectives may co-occur.

The most frequently used connectives in the collection are pero ‘but’ (n = 42) and y ‘and’ (n = 39). They never co-occur in a single RfC. Next in terms of frequency are several inference marking connectives.^[15]

Extract 15 presents a typical case of a pero-prefaced RfC. It is taken from a conversation where the participants are talking about going on a hiking trip together. R suggests a place called Garganta del Cares, a canyon with a river. R knows about the landscape and is the first participant to speak about the Canyon. C is apparently confounded by the fact that R has started talking about a river. At the beginning of the extract, R clarifies that in fact there is a Canyon that has been created by the river.

In l. 05-06, C requests confirmation for the assumption that they are intending to go on a hiking trip. By prefacing the RfC in l. 05 with pero, C signals that the proposition of the RfC contrasts with some assumption in the personal or common ground, in this case, the possible inference that R is planning to do something else than hiking. This is corroborated by C’s explicit reference to the possible alternative, doing river rafting (l. 6), using the negative polarity item nada de (‘nothing of/no’). R then confirms that they will be going hiking (l. 07). The reduplicated n]o_NO. relates to the negated part of the RfC (nada de rafting) and is thus a confirmation. This example is a typical instance of prefacing an RfC with pero in order to signal a (possible) deviation of the confirmable from the speakers’ prior assumptions.

The connective y ‘and’ is the second most used preface in RfCs after pero ‘but’ (e.g., Extracts 9 and 19). A function often found in y ‘and’ prefaces in the collection is to connect the RfC to a preceding question or answer and to establish an orientation to a course of action, as has been shown for English by Heritage and Sorjonen (1994). For an in-depth analysis of Extract 9, which is part of a series of challenging questions, refer to the study by Küttner and Ehmer (2024). As described for English by Bolden (2010), and-prefaces may be used in Castilian Spanish too to introduce a formulation of prior talk that is claimably inferable.^[16]

4.6 Tags

A prominent function of tag questions in general is to request confirmation (see for example Drake 2015, Kimps et al. 2014, Tottie and Hoffmann 2006). Indeed, tags are used in approximately one-third of RfCs in the collection (32.8%, n = 82 of 250), as shown in Table 17.

While tags are thus an important resource for designing RfCs, they are not obligatory in Castilian Spanish.^[17] Table 18 gives an overview of the tags used in the collection.

The tags have been categorized into two groups. In one group are tags that only make relevant a (dis)confirmation from the interlocutor(s). By far the most frequently used tag is no, amounting to 76.8% (n = 63 of 250) of all tags in the collection. The tag eh ‘eh’ and the phrase sabes lo que digo ‘you know what I am saying’ are each used only once in the collection. Apart from those ‘proper’ tags, there are tags that are characterized by the presence of the coordinating connective o ‘or’. Formally, those tags make relevant the existence of alternatives to the confirmable.

Table 19 shows the frequencies with which a tag is used depending on the syntactic format of the RfC.

Overall, the most frequent RfC formats are clauses without a tag (38.8%), followed by clauses with a tag (23.6%) and phrases without a tag (17.6%). There is a difference in the likelihood of specific formats having a tag. Clausal RfCs show the strongest tendency to be realized with a tag (37.8%), followed by (in)subordinate clauses (30.8%) and phrasal RfCs (22.8%). Other formats show the lowest presence of a tag (18.2%).

In general, tags function as a response mobilizing feature (Stivers and Rossano 2010) in RfCs. In addition, for other languages such as German (Deppermann et al. 2024) it has been argued that tags also modalize the epistemic stance the requesters take. This seems to hold true for Spanish no and question tags in general. An argument supporting this analysis is the fact that RfCs with tags receive more disconfirmation (34.1%, n = 28 of 82 RfCs) than requests without a tag (21.4%, n = 36 of 168 RfCs). Since disconfirmation is the dispreferred response type, these numbers point to the fact that requesters have less access to the requested information – by in fact being mistaken more often – when using a tag in the design of the RfC.^[18]

While no ‘no’ can be said to be the default tag in Castilian Spanish (used in 76.8%, n = 62 of the 82 cases where a tag is present), the following analysis will focus on the second most frequent tag o qué ‘or what’ (11.0%, n = 9 of 82). It will be argued that the tag o qué is typically implicative of critique and conveys a negative deontic stance. Extract 7 is taken from a conversation in which participants share the common ground that R is going to go parachuting for the first time. Immediately before the extract C stated the assumption that R has not been parachuting yet. Counter to this assumption, R has actually been parachuting. Upon receiving this information C reacts surprised. In response, C asks if R did not know this. This course of interaction points to some obligation between the participants to inform each other about such events. The extract starts with the RfC.

In l. 01, C formulates the assumption that C has already been parachuting several times. Having been parachuting several times would arguably be an even more severe breach of the assumed obligation between the participants to inform each other about such events. By using the tag o qué ‘or what’ C indexes that she would disapprove of R not informing her about this. R responds to this by stating several times that she has been parachuting only once. With her non-minimal response, R clearly displays that she is aware of being accountable for informing C about such actions. In sum, this example illustrates that RfCs with the tag o qué ‘or what’ typically operate not only on the epistemic domain, but also on the deontic domain, often being implicative of critique or implying a negative assessment.

4.7 Prosodic design

4.7.1 Prosodic design of the confirmable

In the collection of RfCs, the confirmable is typically realized with a falling intonation (84.4%, n = 211 of 250), independently of the presence of an ensuing tag. Table 20 shows the frequencies of intonation of the confirmable, differentiated by the (non)presence of a tag.

Table 20

Final intonation of the confirmable by presence of a tag

	Absolute frequency			Relative frequency (n = 250)
Final intonation on the confirmable	Without tag	With tag	Σ	Without tag (%)	With tag (%)	Σ (%)
Rise	16	7	23	9.5	8.5	9.2
Level	7	9	16	4.2	11.0	6.4
Fall	145	66	211	86.3	80.5	84.4
Σ	168	82	250	100	100	100

In 86.3% of all cases, RfCs without a tag are realized with a falling intonation contour. For RfCs that come with a tag, the percentage is a bit lower (80.5%) and level intonation is a bit more frequent (11.0%). Table 21 shows that the preference for falling intonation is largely independent of the syntactic format.

Table 21

Final intonation of the confirmable by syntactic format

	Absolute frequency					Relative frequency (n = 250)
	Syntactic format					Syntactic format
Final intonation on the confirmable	Clause	Phrase	Subordinate clause	Other	Σ	Clause (%)	Phrase (%)	Subordinate clause (%)	Other (%)	Σ (%)
Rise	14	5	2	2	23	9.0	8.8	7.7	18.2	9.2
Level	11	1	1	3	16	7.1	1.8	3.8	27.3	6.4
Fall	131	51	23	6	211	84.0	89.5	88.5	54.5	84.4
Σ	156	57	26	11	250	100	100	100	100	100

The predominance of falling intonation on the confirmable holds for different types of clausal as well as phrasal confirmables (percentages from 84.0 to 88.5%). Other syntactic formats are also produced predominantly with a falling intonation (54.4% of all cases), but show more level and rising intonation than other formats.

Prior research on Spanish total questions (Section 2.1) has shown that in total questions a ‘circumflex’ contour – characterized by a continuous rise until the focal accent and a subsequent fall to low – is used to attribute the contents of the question to the interlocutor and thus request its confirmation. Figure 1 shows the pitch contour of one such example.

Figure 1

Falling intonation on clausal confirmable, (in)subordinate clause (°t_ah; = que FUISte., rfc037, Extract 11).

After the ah-preface, the intonation contour starts to rise on the connective que until reaching the focal accent on fuiste and then falls to low.

In the collection, however, other types of falling contours are attested too. In Extract 17, L asks R to confirm that he has decided to be in a relationship with Ana (for a sequential analysis, cf. Küttner and Ehmer 2024). Since it is given in the common ground that R had to decide between two possible partners, the focal accent is clearly on the name Ana.

Figure 2 shows the pitch contour of the RfC. The pitch contour is already starting to fall on the minor accent of the participle decidido.

Figure 2

Falling intonation on clausal confirmable (rfc026, Extract 17).

Another instance of a different falling intonation contour in the RfC y_ha coGIdo su abrigo y tOdo. is shown in Figure 3. After the focal accent on coGIdo, the intonation stays on a plateau and only starts to fall after the minor accent on TOdo. Further research about the different kinds of falling contours in RfCs is required.

Figure 3

Falling intonation on clausal confirmable introduced with a connective (y_ha coGIdo su abrigo y tOdo., rfc123, Extract 19).

4.7.2 Prosodic design of tags

When a tag is present in the RfC (32.8%, n = 82 of 250), it is typically realized in a separate contour from the confirmable (95.1%, n = 78 of 82 tags). Only rarely, in 4.9% (n = 4 of 82 tags) it is integrated into the contour of the confirmable. Table 22 shows the prosodic (non)integration differentiated by tag.

Table 22

Tags by prosodic (non)integration with the confirmable

	Absolute frequency			Relative frequency (total n = 82)
	Prosodic realization of tag			Prosodic realization of tag
Tag	Separate contour	Integrated	Σ	Separate contour (%)	Integrated (%)	Σ (%)
no ‘no’	62	1	63	98.4	1.6	100
o qué ‘or what’	6	3	9	66.7	33.3	100
o no ‘or not’	7		7	100		100
eh ‘eh’	1		1	100		100
sabes lo que te digo ‘you know what I am saying’	1		1	100		100
sí o no ‘yes or no’	1		1	100		100
Σ	78	4	82	95.1	4.9	100

Note that prosodic integration only occurs with the two most frequent tags (no, n = 1) and o qué (n = 3). Regarding realization in a separate contour, it is important to note that the tag is typically produced in immediate succession to the intonation unit of the confirmable. In very few instances (3.7%), there is a pause between the confirmable and the tag (n = 2 instances of o no and n = 1 instance of no out of n = 82 tagged RfCs). Extract 18 shows a case with a long pause between confirmable and tag.

R and L are sitting in the studio, waiting for the interviewer to arrive. R is singing a song while L is commenting on the surroundings. L formulates an assumption about a (possibly blinded) window in the room (l. 03), requesting confirmation from R. R, however, keeps on singing and does not react. L then produces the tag no, (l. 05) to mobilize a response from R. As R still does not respond, L verbally addresses R’s lack of response (l. 07). Such non-integrated uses are illustrative cases of the response mobilizing potential of tags.

Table 23 shows the prosodic realization of tags that have a separate contour from the confirmable (n = 78 of 82 tags).

Table 23

Final intonation on tags that are not prosodically integrated with the confirmable (i.e., tags that have their own pitch contour)

Final intonation on Tags	Absolute frequency			Relative frequency (total n = 78)
Tag	Rise	Fall	Σ	Rise (%)	Fall (%)	Σ (%)
no ‘no’	62		62	100	0	100
o qué ‘or what’	3	3	6	50.0	50.0	100
o no ‘or not’		7	7		100	100
eh ‘eh’	1		1	100	100	100
sabes lo que te digo ‘you know what I am saying’		1	1		100	100
sí o no ‘yes or no’		1	1	50.0	50.0	100
Σ	66	12	78	84.6	15.4	100

The most frequent tag in the collection, no ‘no?’, is realized exclusively with rising intonation. Falling intonation is only found on tags that contain the connector o ‘or’ and make alternatives relevant (i.e., o qué, o no, sí o no), as well as the complex tag sabes lo que te digo. Level intonation is not found in any of the tags in the collection.

5.1 Responsive actions

RfCs are first pair parts and thus make a second action relevant. In fact, RfCs in the collection overwhelmingly do receive a response (93.6%, n = 234 of 250), as shown in Table 25.

In the 16 cases without a response, the lack of response is due to the request being produced in overlap with the main speaker not suspending their current project (Extract 18), the addressee asking the requester to restate the RfC (presumably because of understanding issues) or the requester him/herself issuing another request, either an RfC or a request for information. Within the sample of languages examined, Spanish has the highest percentage of requests without response (see Pfeiffer et al. forthcoming). This result is most likely due to the dynamic nature of the recorded interactions.

Table 26 shows the frequencies of the types of responses that RfCs in the collection receive.

Responses to the RfCs analyzed are in their vast majority confirming (61.1%, n = 143 of 234). Disconfirmation is only found in 27.4% (n = 64 of 234) of the cases. In 15.6% of the responses, the addressee(s) neither confirm nor disconfirm. The most frequent way of doing this in the collection is by expressing a lack of knowledge (not knowing, not remembering). The second most frequent method is the production of a transformative answer (Stivers and Hayashi 2010).

Of the n = 234 responses in the collection, only n = 4 are realized in a purely nonverbal fashion, confirming the request by nodding. This results in n = 230 RfCs with a verbal response that will be scrutinized in the subsequent analysis.

Several resources can be used for designing responses, among them response particles (5.2), full and partial repeats of the confirmable (Section 5.7) that can also be combined with different kinds of expansions. Before scrutinizing some of the formal variations in detail in subsequent sections, the following examples illustrate some typical cases, first confirmations then disconfirmations. Extract 17 is an example of a minimal confirming response using a response token only. Responses with response tokens may also be expanded, for example by a partial repeat of the RfC (e.g., Extract 19) or by adding an account or explanation (e.g., Extract 1). Confirming responses may also be realized without response tokens (e.g., Extract 14). In disconfirmations, minimal responses are possible, too (e.g., Extract 20). More typical, however, are expanded responses that also include a correction (e.g., Extract 21). Disconfirming responses without response tokens, providing a correction only are also possible (e.g., Extract 22).

5.2 Response tokens

5.2.1 Inventory of response tokens

The relevance of response tokens as a resource for responding not only to RfCs but more generally to polar questions has been shown for different languages (Enfield et al. 2019). In Spanish, too, response tokens are a central resource for designing responses to RfCs. In 67.8% (n = 156 of 230) of the verbal responses, a response token is used, typically placed at the beginning of the response turn.

Compared to other languages – for instance, German (Deppermann et al. 2024) or English (Küttner and Szczepek Reed 2024) – the Spanish collection shows a relatively limited inventory of only five response tokens. Table 27 presents the frequencies of the response tokens.

Table 27

Response tokens used in responses to RfCs^a

Response token(s)	Absolute frequency	Relative frequency (total n = 156) (%)
Positive polarity	97	62.2
sí ‘yes’	83	53.2
claro ‘of course’	12	7.7
hm ‘hm’	2	1.3
Negative polarity	59	37.8
no ‘no’	58	37.2
ʔhmʔhm	1	0.6
Σ	156	100

^aClusters of response tokens have been counted as a single instance in this table.

The most frequent response tokens are sí ‘yes’ (positive polarity, 53.2% n = 83 of 156) and no ‘no’ (negative polarity, 37.2% n = 58 of 156). Regarding positive polarity items, the inventory also includes claro ‘of course’, which is typically used for emphatic/stressed confirmation, sometimes in different combinations with sí (Freites Barros 2006, Hummel 2012, Pons Bordería 2003).

Apart from lexical response tokens, the inventory also contains hm as an interjection (Dingemanse 2017). In its monosyllabic realization, hm is used for confirmation, and in its bi-syllabic variant ʔhmʔhm – with preceding glottal stops – it is used for disconfirmation. In contrast to claro, the interjection hm is often used for weak confirmation, that is, when the epistemic gradient is quite shallow and the confirmable is easily accessible in the common ground. The same holds for ʔhmʔhm, which is used for weaker disconfirmation than no.

The following extract illustrates the use of hm for weak confirmation.

In this sequence, R states that she will be going to her village (l. 01) during the weekend. C first reacts surprised, repeating R’s statement (l. 02), but then recollects the reason for R’s trip, namely that she forgot her mobile phone there (l. 04-05). This reason is marked as an RfC by the tag no. Since the reason is part of the participant’s common ground, a weak confirmation using hm (l. 06) is sufficient.

Apart from response tokens, three other conventionalized ways of responding to RfCs are found in the collection, all of them expressing no epistemic access: no (lo) sé ‘I don’t know (it)’ (n = 6), ni idea ‘no idea’ (n = 2), and no me acuerdo ‘I don’t remember’ (n = 1).

5.2.2 Modalization of epistemic stance

In addition to choosing among different response tokens to modalize their epistemic stance, responders may also use different modalizing resources in combination with the default response tokens sí and no. Tokens are often embedded in matrix structures such as (yo) creo que… ‘I think that…’ (e.g., Extract 7) and puede que… ‘(it) may be that…’ (e.g., Extract 24) to downgrade their epistemic stance. In addition, expressions of reduced epistemic access, such as no sé “I don’t know” (e.g., Extract 7), no sé decirte “I can’t tell you” (e.g., Extract 24) and me parece “it seems to me” (e.g., Extract 25), are frequently included in the response turn for modalization, often in a concessive fashion (e.g., Extract 24).

5.2.4 Confirming and disconfirming RfCs using response tokens

While the generalities of how response tokens are used to (dis)confirm an RfC have been described in Section 2.3, this section will provide concrete examples and frequencies. As mentioned above, the usage of response tokens for (dis)confirmation depends on the polarity of the RfC. Positively formatted RfCs are confirmed with as positive polarity token (e.g., Extracts 1, 3, 5, 6, and 7) and disconfirmed with a negative polarity token (e.g., Extracts 15 and 16). With a negatively formatted RfC, the polarity of the response items is reversed: confirmation is realized with a negative polarity token (Extract 26), and disconfirmation is realized with a positive polarity token (Extract 27).

In Extract 26, C only uses the response token no ‘no’ to confirm, while L uses no in combination with a repetition. In Extract 27, C uses sí ‘yes’ for disconfirmation in combination with a repetition, here used for correction.

Table 29 presents the absolute and relative frequencies of the response tokens used for (dis)confirming positively and negatively formatted RfCs.

Table 29

Polarity of RfCs and response type by response tokens used

		Absolute frequency
		Polarity of response token
Polarity of RfC		Positive			Negative		Σ
		sí ‘yes’	claro ‘of course’	hm ‘hm’	no ‘no’	ʔhmʔhm
Positive	Type of response	75	12	2	36		125
	Confirmation	75	12	2			89
	Disconfirmation				36		36
Negative		8			22	1	31
	Confirmation	2			21	1	24
	Disconfirmation	6			1		7
	Σ	83	12	2	58	1	156

		Relative frequency (total n = 156)
		Polarity of response token
Polarity of RfC		Positive			Negative		Σ (%)
		sí ‘yes’ (%)	claro ‘of course’ (%)	hm ‘hm’ (%)	no ‘no’ (%)	ʔhmʔhm (%)
Positive	Type of response	60.0	9.6	1.6	28.8		100
	Confirmation	84.3	13.5	2.2			100
	Disconfirmation				100		100
Negative		25.8			71.0	3.2	100
	Confirmation	8.3			87.5	4.2	100
	Disconfirmation	85.7			14.3		100
Σ		53.2	7.7	1.3	37.2	0.6	100

While the overall distribution conforms to the schematization presented in Section 2.3, there are n = 3 responses to negatively formatted RfC that at first sight do not conform to the schema: in n = 2 instances, sí is unexpectedly used for confirmation, and in n = 1 cases, no is used for disconfirmation. This ‘deviation’ can be explained by the fact that in these instances the negation does not have scope over the confirmable but is rather part of a modalizing strategy. The seemingly atypical cases are the negated use of a Pretérito imperfecto in no ibas a […] “weren’t you going to” (Extract 28, see also Section 4.3) and the negated reportative clause no me dijiste que […] “didn’t you tell me that” (Extract 29). Since the negation in these examples does not have scope over the confirmable, these instances should actually be interpreted as positively formatted RfCs. For English, Heritage (2002a) has claimed that certain negative interrogatives function as vehicles for assertion (compare also Bolinger 1957, Heritage 2012). This seems to hold for Spanish too.

5.3 Clusters of response tokens

Response tokens may be used not only as a single occurrence but also with reduplication. This is the case in 5.8% of all verbal responses (n = 9 of 230). Table 30 gives an overview.

Only the main response tokens sí and no (but not claro) are found in reduplicated form in responses to RfCs within the collection, typically signaling emphasis (e.g., Extract 15). For this reason, reduplication also shows an affinity with (partial) repetition of the RfC (Section 5.7), both for confirmation and for correction in disconfirmations.

Reduplicated uses of no only occur in non-minimal responses in the collection. In contrast, reduplicated sí also occurs in minimal responses (40% of all reduplicated instances of sí).

5.4 Position of the first response token

If present, response tokens are overwhelmingly placed in turn-initial position in the response 92.3% (n = 144 of 156). Table 31 gives an overview of the frequencies of different response token positionings.

A non-turn-initial placement of the response token (medial 2.6% n = 4 of 156, final 5.1% n = 8 of 156) is found in cases of modalization (e.g., creo que sí, ‘I think that yes’), partial repetition of the RfC followed by the token (Section 5.7), and preceding emphatic markers other than discourse markers (e.g., por fin ‘finally’).

Please note that response tokens that have been coded as initial also include n = 37 cases in which the response is minimal (cf. Section 5.5).

5.5 Minimal and non-minimal responses

Responses to an RfC can consist of response tokens only, in which case they have been coded as minimal. In non-minimal responses, a response token is accompanied by further talk. Table 32 shows how (non-)minimal responses are distributed over response types.

In total 25.6% (n = 40 of 156) of the token responses are minimal, and 74.7% (n = 116 of 156) are non-minimal. Confirmations are realized in a minimal form almost twice as often as disconfirmations (29.2% vs 16.3%). This can be accounted for by the tendency to not only disconfirm with a particle but also provide a correction (Extract 28), an account, or a partial repetition of the confirmable for emphatic negation (Extract 27).

5.6 Response prefacing

Responses are realized with a preface in 15.6% (n = 36 of 230) of the cases. There are differences in frequency, however, depending on the response type, as represented in Table 33.

While confirming responses are prefaced in only 10.0% of the instances, prefacing is much more frequent in the case of non-confirming responses: disconfirmations are prefaced in 14.1% of the instances and the percentage rises to 50.0% in the case of responses where neither confirmation nor disconfirmation is realized.

This tendency for prefacing non-confirming responses is relevant to the crosslinguistic finding that Castilian Spanish shows the highest percentage of prefacing in the sample of languages analyzed (see Pfeiffer et al. forthcoming). This high rate of prefacing is most likely a consequence of Castilian Spanish exhibiting the second-highest percentage of non-confirming responses in the sample of languages (see Pfeiffer et al. forthcoming).^[19]

Table 34 gives an overview of the different tokens used to preface responses in the collection by response type.

Due to the overall low frequencies, no strong claims can be made about tendencies in the use of particular prefaces for certain response types. Nevertheless, it is apparent that many of the prefacing elements have the potential to mitigate dispreferred (e.g., non-confirming) responses.

In the collection, prefaces to responses typically contain only one element, although there are two cases of combinations of elements. One of those cases is presented in Extract 30.

5.7 Repeats

Responses to RfCs may contain or even solely consist of a full or a partial repeat of the request. This is the case in 21.7% of the instances in the collection, as shown in Table 35.

Full repeats may contain the entire verbal realization of the confirmable, with possible morphological adaptation or omission of discourse markers and tags (e.g., Extracts 8 and 26). This is the case in 5.7% of all verbal responses (n = 13 of 230). Full repeats may also be expanded, which is the case in n = 6 of the n = 13 full repeats in the collection. In partial repeats, only part of the confirmable is recycled (e.g., Extract 19), sometimes in combination with an expansion. Table 36 shows the frequencies of full and partial repeats by response type.

Full repeats (including full expanded repeats) are only used for confirmation in the collection (n = 13 of 13). In contrast, partial repeats are also used in disconfirming responses (32.4%), albeit with a lower frequency than in confirming responses (62.2%).

Responses with repeats may and often do contain a response token, which holds for both full repeats (e.g., Extracts 8 and 26) and partial repeats (e.g., Extract 19). Responses may, however, consist solely of a repeat without a response token (e.g., Extract 31 with a full repeat).

Table 37 shows the frequencies of responses with/without repeats according to the presence or absence of a response token, differentiated by response type.

Concerning confirmations, it should be remembered that, overall, 80.7% of the responses do contain a response token (Section 5.2, Table 28). The above table shows, however, that there are strong differences depending on whether the confirmation contains a repeat or not. In 90.4% of the instances without a repeat, the response contains a response token. In contrast, only 52.8% of the confirmations with a repeat also contain a response token. This can be accounted for by the fact that repeats may be used as devices for confirmation. The presence of both a response token and a repeat may be accounted for by two factors. Token and repeat may respond to different interactional problems presented in the RfC (cf. Enfield et al. 2019), as in Extract 15, where the RfC presents a confirmable and a contrasting item. In addition, token and repeat may be combined to achieve a strong confirmation (Ehmer and Rosemeyer in prep., Rosemeyer and Schwenter 2019).

In disconfirmations the difference in the use of a response token in responses with repeats (58.3%) and without repeats (69.2%) is not as marked as in confirmations. The frequency of occurrence of response tokens in combination with partial repeats is similar in disconfirmations (58.2%) and confirmations (56.5%).

To summarize briefly, in the collection full repeats are only used for confirmation (with and without response tokens), and there is a tendency to use less response tokens in responses with repeats (as compared to responses without repeats).