Request for confirmation sequences in Czech

4.1 Syntactic design

The overwhelming number of RfCs in Czech are in clausal format (n = 143/204, 70.1%), whereas about a third (n = 61/204, 29.9%) are phrasal. The different types of clausal and phrasal RfCs listed in Table 2 will be illustrated further below.

In Czech, there is no specific format for forming clausal RfCs (see Section 2). This means that most RfCs in the clausal form are declaratives. In our collection of Czech RfCs, the interrogative mode is introduced mainly by prosody (rising final contour), or, e.g. (subordinate) clauses introduced by že ‘that’ (which can also be analysed as an interrogative marker, see Short 2001, 497, Štícha 2015, 48) or by wh-words (ex. 190, see in Table 3). Tags also mark an otherwise declarative structure as possibly interrogative (Cvrček et al. 2010, 296). Only 21 clausal structures (out of 143, i.e., 14.7%) could be identified as clearly interrogative. The remaining cases consist of declarative structures that are more or less simple or complex, and that can include features marking them as RfCs (such as tags, see Excerpt 1). The different types of clausal structures in Czech RfCs are illustrated by examples in Table 3.

Within the group of 143 clausal RfCs, the majority are simple declaratives (n = 95/143, 66.4%), followed by complex declaratives (n = 15/143, 10.5%) and declaratives that are extended (n = 12/143, 8.4%, Table 2). As illustrated by the examples in Table 3, the various types of clausal RfCs in Czech may contain tags or not. Indeed, it should be noted that 61 out of the overall 106 tags (57.5%, Section 4.6) are appended to simple declarative formats (see also Table 2 for a distribution of tags according to syntactic formats). Appended tags such as jo ‘yes’ or ne ‘no’ (n = 82/143, 57.3% of the clausal RfCs containing a tag, see also Section 4.6) can mark the otherwise declarative structure of the confirmable as an RfC (ex. 127, 26, Table 3). Also, plain declarative constructions with falling final intonation can be treated as confirmable, both in the case of simple (ex. 90, ibid.) and complex (ex. 171, ibid.) declaratives, as well as in declaratives that are syntactically extended (ex. 47, ibid.). The relative flexibility of syntactic design in spoken Czech can be seen in cases where the syntactic structure is extended after an inserted tag, in all 12 of these instances these extensions are prosodically non-integrated add-ons (Oloff and Havlík 2018, Ono and Couper-Kuhlen 2007), such as replacements (ex. 47, 123, Table 3) or glue-ons (ex. 28, ibid.). Apart from declarative clausal structures, some clausal RfCs are formatted as subordinates (n = 15/143, 10.5%), typically by the connector že ‘that’ (ex. 98, 172, ibid.). Only six clausal RfCs (n = 6/143, 4.2%) have a clear interrogative syntactic design, for instance, marked by an interrogative pronoun (ex. 190, ibid.) or a post-positioned pronoun (ex. 196, ibid.).

As also shown in Table 2, about 30% (n = 61/204) of the Czech RfCs are phrases. These are of various types, mainly noun phrases (n = 32/61, 52.5%), followed by prepositional (n = 13/61, 21.3%) and adjunctional (n = 10/61, 16.4%) phrases, and some less frequent cases of adverbial (n = 4/61, 6.6%) and adjectival (n = 2/61, 3.3%) phrases. Some examples of phrasal RfCs are shown in Table 4.

Like clausal RfCs, phrasal RfCs can be produced with (n = 24/61, 39.3%) or without (n = 37/61, 60.7%, Table 2) tags, and with various intonation patterns, which mark them as more or less interrogative and thus identifiable as RfCs. Thus, a bare noun phrase with falling intonation such as arašídy; ‘peanuts;’ (ex. 74, Table 4) is more declarative than, for instance, phrasal RfCs carrying a rising final intonation (such as jako vláknina? ‘like fibre?’, ex. 70, ibid.) or a tag (that itself can carry a rising intonation, such as v Německu; jo, ‘in Germany; yes,’, ex. 137, ibid.). Thus, RfCs in Czech do not have a prototypical syntactic design that allows them to be identified as such. Rather, the various syntactic designs can be identified as RfCs by a combination of prosodic patterns, the presence of tags, and, mainly, their position within a sequence and the epistemic asymmetry they express within it.

4.2 Polarity

The overwhelming number of RfCs in Czech is realized with positive polarity (n = 178/204, 87.3%). RfCs formatted in negative polarity are therefore rather rare in this data set (n = 26/204, 12.7%). Within the sample of ten languages, Czech therefore figures among the ones having the lowest ratio of RfCs with negative polarity (see Pfeiffer et al. forthcoming). As suggested by Küttner and Szczepek Reed (2024), this comparatively low proportion might reflect the fact that most participants in the data set are well-acquainted, and that there thus is, for instance, less need to formulate more uncertain updates in the form of negative polar RfCs. But to check this more thoroughly, a more heterogeneous data set with respect to setting types and participants would be needed.

While positive polarity is not specifically marked on verbs or other word types in Czech, negative polarity (for negation in Czech in general, see Short 2001, 510–2) is most frequently expressed with a negative verb form (adding the prefix ne- to the inflected verbal form, for example, přijede ‘he/she will come’ vs nepřijede, ‘he/she won’t come’), or by other lexical elements expressing negativity (such as a freestanding ne ‘no(t)’, nikdo ‘no one’, nic ‘nothing’, žádná ‘no(ne)’) (Table 5).

Since in Czech negative polarity is marked not only on inflected verbs but also on concurrent pronouns when expressed in the syntactic construction, some RfCs carry double negation markings (e.g., ježiš to tam vlastně nikdo ↑neumí; ‘jeez no one there actually knows how to do it’, literally ‘jeez that there actually no one doesn’t know how to do’; or tam ↑nic nezůstane ‘nothing will remain there’, literally ‘there nothing doesn’t remain’). This leads to five RfCs in the collection containing double markings for negative polarity (Table 5), corresponding to the standard negation in Czech.

4.3 Modulation

A clear majority of the RfCs in our collection do not contain any modulating elements (n = 152/204, 74.5%). The remaining quarter of the RfCs (n = 52/204, 25.5%) contain mainly approximators (n = 34/58, 58.6%), such as jako ‘like’ (Oloff 2022), or epistemic particles, such as přece or vlastně, which could be translated to something like ‘still’ or ‘actually’ (n = 12/58, 20.7%). The total number of modulation markers (58, Table 6) is higher than 52, as a few RfCs contain several modulating elements. For instance, the RfC (ex. 92) to prostě- to je trošku jako krb; jako kamna; ‘it simply- it’s a bit like a fireplace; like a stove;’ contains four modulating elements, prostě ‘simply’/‘just’, trošku ‘a bit’, and two times jako ‘like’, that are used in the description of a digital picture shown to the speaker of the RfC for the first time. The remaining modulation markers consist of mental predicate phrases that more explicitly display the confirmable as being issued from a subjective perspective (n = 6/58, 10.3%), and some other items that could not be classified as approximators, epistemic particles or mental predicate phrases (n = 6/58, 10.3%; Table 6).

The most frequent modulation marker overall is the approximator jako ‘like’ (for a similar observation for English, see Küttner and Szczepek Reed 2024), corresponding to 36.2% (n = 21/58) of all modulation markers in our sample. Jako is used with both phrasal and clausal RfCs, in the former case being often in initial position (e.g. jako vláknina? ‘like fibre’, jako stará?, ‘like the old (one)’), in the latter being more flexibly positioned within the confirmable TCU (Oloff 2022, e.g., a voni tě znali jako; jo, ‘and they like knew you; yes,’ ona jako dělá- (.) pedikúru manikúru; né, ‘she is like doing- (.) pedicure manicure; no’).

4.4 Inference marking

Regarding inference marking, i.e., confirmables that are linguistically marked as inferable from prior talk, 164 cases out of 204 (80.4%) do not contain any inference markers. The remaining 19.6% (n = 40/204 cases) contain various inference markers, the most frequent ones being tak ‘so’/‘then’ (n = 11/40 cases), takže ‘(and) so’/‘thus’ (n = 10/40 cases), and the imperative form počkej ‘wait’ (n = 8/40 cases). The total number of n = 42 cases (Table 7) is due to the fact that two out of the 40 RfCs contain two different inference-marking items.

4.5 Connectives

A total of 116 out of the 204 cases (56.9%) contain an initial connective, a pre-positioned lexical item that connects the RfC to the co-participant’s prior turn(s). These connectives were double-coded with respect to the inference markers (cf. Section 4.4). The total number of 121 connective markers as illustrated by Table 8 relates to five of the examples including two connectives that were counted separately as they were produced as separate prosodic units and not as a single intonational unit.

Most of the 121 connectives are simple connectives, mainly a ‘and’ (n = 37/121, 30.6%), followed by jako ‘like’ (n = 12/121, 9.9%), tak ‘so’ (n = 9/121, 7.4%), and ale ‘but’ (n = 8/121, 6.6%; Table 8). Together with the combinations of connectives, connectives make up more than two-thirds of the connective items overall (n = 83/121, 68.6%). The second largest group (n = 16/121, 13.2%) are response tokens, which can be either simple (n = 9/121, 7.4%) or combined (n = 7/121, 5.8%), followed by some imperative forms (n = 8/121, 6.6%) and a variety of adverbial and other tokens (n = 14/121, 11.6%). Unsurprisingly, most connectives are used with clausal RfCs (n = 92/116, 79.3%), although about a fifth occur in phrasal RfCs (n = 24/116, 20.7%). Within the phrasal RfCs, however, jako ‘like’ is the most prominent one (9 out of 24 cases). The use of simple connectives shows that participants explicitly tie their RfCs to previous talk, for instance, by singling out a detail that could or should be elaborated on (ex. 39, a tos tam byla tak pět let, °jo;° ‘and so you’ve been there for about five years, yes;’ to a speaker who presented a long storytelling about her student life), or to an aspect related to what the speaker of the RfCs has just heard and needs to revise (ex. 58, ale to se musí přece na↑máčet, ‘but this nevertheless has to soak’, an RfC aiming at questioning a step within a reported process of how to prepare almond milk).

4.6 Tags

In our data, 52% (n = 106/204) of the RfCs contain a tag (‘factual particles’, Cvrček et al. 2010, 296). While there is a large variety of tags in our data (11 different types were identified, see also Müllerová 2007), they distribute unevenly in that two tags are by far the most frequent. The first one is jo (also used as an affirmative response particle, in that sense best translated as ‘yes’ or ‘yeah’), which is used in 43 RfCs (n = 43/106, 40.6% of all tags containing RfCs). In second position with respect to frequency is the tag ne ‘no’, prototypically used as a negative response particle in responsive turns, with 32 cases out of 106 (30.2% of all tag containing RfCs). This means that the most frequent tags in Czech appended to RfCs emphasize (by their concurrent use as (dis)affirming response tokens in spoken Czech) the relevance of a polar response – either confirming or disconfirming – to the RfCs, thereby contributing to disambiguating the declarative syntactic structure of possible RfCs (Section 4.1). The distribution of the tags with respect to the polarity of the RfC (the two right columns in Table 9) illustrates that there is no direct correspondence between the polarity of the RfC and the polarity of the tag. Indeed, 36.5% of the RfCs with positive polarity and tags (n = 35/96) have an appended jo ‘yes’, while 32.3% (n = 31/96) append a ne ‘no’. In the case of RfCs with negative polarity and tags (n = 10), 8 (i.e., 80%) append a jo ‘yes’ while only one case (n = 1/10, 10%) carries a ne ‘no’ tag. While overall rarer (as the remaining 29.2% of the tags are mainly positive with respect to their polarity – apart from nebo ‘or’, 4 cases), the fact that 31 (96.9%) of the negative tags are appended to positive RfCs means that they might take on a specific contrastive function.

The remaining 29.2% of the tags (n = 31/106) consist of clearly less frequently used ones, such as že (literally ‘that’, 8 cases), že jo (literally ‘that yes’, 8 cases, Komrsková 2017), and nebo (literally ‘or’, 5 cases; Table 9). As it is difficult to render a clear pragmatic translation of these tags in English, Table 9 contains only literal translations of their standard (i.e., as defined by dictionaries) uses as response tokens or conjunctions. While most of the tags are single items, some of these can be combined into a prosodic unit to form more complex tags, albeit being rarer (such as že jo ‘that yes’, 8 cases, or nebo co ‘or what’, 1 case).

Looking at the 43 cases containing jo ‘yes’ as a final tag, these RfCs tend to be short in that they are mostly simple declaratives or phrasal ones. The confirmable they offer mainly refers to facts that can be easily guessed or inferred from prior talk or actions. Thus, they are not particularly ‘risky’ or challenging confirmables (and only n = 9/43, 20.9% carry inference markers), but tend to point at a small detail, thereby working more as topic proffers or go-aheads than offering a candidate understanding that truly aims at resolving an epistemic asymmetry between the speakers. In Excerpt 2, Ludmila (LUD) is in the middle of a longer telling (to her daughter Jana, JAN) about her student life, talking, for instance, about the different student accommodations she experienced.

LUD is talking about a campus she moved to during her studies which had no canteen (lines 01-02), contrasting it to another one which had a ‘great canteen’ (lines 04-05). In that sense, JAN’s RfC (line 06) refers to a fact that is clearly not in her domain (as it refers to her mother’s experience), but that can be easily inferred, as a great canteen is likely to provide ‘good food’. This easily inferable fact is marked by a jo-tag and then responded to by a turn-initial no (line 07) and a partial repeat of the RfC, marking LUD’s higher epistemic position and the RfC’s fittedness (Section 5.6). LUD then expands on the confirmable (lines 07-09), revealing the topic-proffering nature of this RfC.

The RfCs using ne ‘no’ as a tag tend to present less certain or less obvious facts. This can relate, on the one hand, to the fact that they cannot simply be inferred from previous talk, and in that sense present a true guess about a specific fact in the other speaker’s domain. It should be noted that ne-tagged RfCs (with one exception) do not have any inference markers. On the other hand, they seem to offer a new aspect to be confirmed, either as they introduce a new sequence altogether, or by offering a possible deviation from the current topical line of talk. The contrast between the use of jo and ne as tags can be seen in Excerpt 3. Yveta (YVE) and Marta (MAR) have been talking about healthy nutrition, and YVE now confesses that she likes to have a snack while watching movies (lines 01-02). MAR then enquires about the type of snack she has (line 03), leading YVE to elaborate on the unhealthy nature of her snacks (line 04) and MAR to formulate two differently formatted RfCs.

When YVE specifies that she ate a whole bag of nuts, MAR first starts by admitting she also likes eating nuts (line 06). Her RfC ‘but healthy nuts no’ then points to an important distinction from a nutritional point of view, as nuts can be served plain or salted, and thus be more or less healthy. By using a ne-tag, MAR frames the confirmable as something that cannot be inferred from YVE’s previous statements (although the initial ale ‘but’ and YVE subsequent disconfirming response in line 06 illustrate that it might have been more likely that the snacks were not healthy and that therefore the RfC has a possible admonishing flavor to it). When YVE admits that the nuts were ‘not even healthy’, MAR provides a follow-up RfC (line 08) in which she makes their ‘unhealthiness’ explicit. The assumption that the nuts were ‘salted’ can now be safely and easily inferred, and as a consequence, this RfC carries a jo-tag. Only when this is minimally confirmed by YVE (line 09), MAR expands on the possible admonishing nature of her first RfC (line 10), contrasting her nutritional choices with her friend’s. If the difference between jo- and ne-tags lies indeed mainly in the inferrability of the facts they present (and is thus linked to epistemics), or if the difference is a matter of sequential and/or topical management (e.g., if the RfC is in first or second position), remains to be examined more closely. As for the tags že and že jo, these seem to be appended to RfCs that are clearly more in the domain of the speaker, or at least facts with respect to which the speaker claims more certainty. According to the distribution in our data set, the forms že and že jo could be possibly linked to specific regions (že jo being used by the speakers from the capital/central Bohemian region, while že is being used by speakers in the Silesian region). But as their overall numbers (eight cases each, Table 9) are rather low, a larger data set would be needed to check this.

4.7 Prosodic design

Regarding prosodic design, the coded intonation contour (rising, falling or level, see the coding manual, König et al. forthcoming) always refers to the last syllable of the TCU containing the confirmable or the tag. This coding represents a possibly coarse insight into prosodic patterns of Czech RfCs (and future studies should indeed investigate prosodic features in more detail), but the results nevertheless help to reveal some main patterns and trends. Nearly two-thirds of the confirmables (n = 130/204, 63.7%) carry a final falling intonation, compared to 22.6% (n = 46/204) having a rising and 13.7% (n = 28/204) a level final intonation contour. This majority of falling final intonation is maintained in the cases where a tag is appended to the confirmable (see the right column in Table 10, n = 86/106, 81.1%), followed by level (n = 12/106, 11.3%) and rising intonation (n = 8/106, 7.6%). In cases of no appended tag (about half of the RfCs, 98 cases), the falling intonation is still the most frequent one (n = 44/98, 44.9%), but a considerable part of non-tagged confirmables carry a rising final intonation (n = 38/98, 38.8%, the middle column in Table 10).

The majority of the 106 tags are prosodically non-integrated with respect to the confirmable (n = 91/106, 85.9%), one exception being že jo with half of its occurrences being integrated and the other half non-integrated (Table 11, and Table 9 shows a translation of the tags). The most frequent tags used are jo (40.6%) and ne (30.2%), followed by že and že jo (each 7.6%, see also Table 9, Section 4.6). The remaining 10 tags (9.2% of the 106 cases) occur clearly less frequently.

Of the 91 tags carrying their own prosodic contour (i.e., those being prosodically non-integrated, Table 12), the monosyllabic ones mostly carry a rising final intonation contour (see also Müllerová 2007) and rarely a level contour (in case of jo and ne). The disyllabic ones (being composed of one or several tags, in these cases, the intonation contour of the final syllable was coded), albeit being less frequent overall, show a different pattern, in that they show level or falling intonation contours (že jo, že jo no, nebo, nebo co, Table 12, for a translation of the tags see Table 9). Due to their infrequent occurrence in our data set, we cannot, however, make a statement with respect to the systematicity of this pattern.

Combining the quantitative findings from Tables 10–12, the most frequent intonation contours for confirmables in Czech without tags are falling (n = 44/98, 44.9%) and rising (n = 38/98, 38.8%, Table 10), whereas the main intonation pattern for a combination of a confirmable with appended, non-integrated tag is fall-rise (n = 65/82, 79.3%), followed in frequency by a fall-level pattern (n = 14/82, 17.1%), and a rise-fall pattern (n = 4/8, 50%, see Table 13. Integrated tags (n = 15/106, 14.2%) most often combine with level-intonation confirmables (n = 11/15, 73.3%, check the right column of Table 13).

5.1 Responsive actions

Of the 204 RfCs in our collection, 196 (96.1%) received a response, with 8 cases being not responded to (3.9%). Responses to RfCs can be either verbal (audible) only, embodied only, or both verbal and embodied (Table 14). While audible responses, regardless of their precise length and formatting (cf. Sections 5.2–5.6), represent a clear majority (n = 179/196, 91.3%) of all responses, embodied resources are also an important part of responsive action to RfCs in Czech, with 83.2% of all responses containing embodied features. Among these, 8.7% (n = 17/196) are realized solely by visible means. Notice also that in 11 cases of the category ‘verbal response only’, the responding participants are not (sufficiently) visible. Consequently, some of these 11 cases might contain embodied responsive features as well, meaning in return that only 22 responses (11.2% out of the 196 responses to RfCs) do demonstrably contain no embodied features.

The embodied elements accompanying or providing the response to the RfC mostly consist of head nods (92 cases with multiple head nods, 34 with single nods, i.e., n = 126/163, corresponding to 77.3%). Head shakes are less frequent, with 18 cases of multiple headshakes and 6 single ones (i.e., n = 24/163, 14.7%). Subtracting the 4 cases in which both a type of headshake and a nod occur, this means that nearly 90% of all the responses to RfCs rely on (supplementary) head movements. As the majority of responses to RfCs are clear confirmations (80.1%, Table 15, compared to 14.8% of clear disconfirmations), the higher occurrence of head nods is not surprising. Regarding this distribution, it should however also be considered that the polarity of the RfC might also be consequential for the type of embodied resource that can be used to (dis)confirm. Indeed, 15 confirmations do not contain any head nods, of which 9, however, contain single or multiple headshakes (typically linked to an RfC in negative polarity). Also, during 7 disconfirming responses, head nods are carried out. Other frequent embodied resources used for responding consist of facial gestures such as raising or flashing the eyebrows, frowning, blinking or closing the eyes (cf. De Stefani 2021, Dix and Groß 2024, Hömke et al. 2017), which co-occur in about 67 cases (34.2% of all 196 responses to RfCs).

The data set reveals a clear preference for the confirmation of RfCs, with less than a fifth being disconfirmations (80.1% compared to 14.8%). Only 5.1% (n = 10/196) of the responses cannot be said to clearly confirm or disconfirm.

5.2 Response tokens

Of the 179 responses containing lexical or other audible material (Table 16), 152 contain one or more response tokens. In multi-party settings, more than one participant can respond to the RfC, leading to some cases of double responses to a single RfC. In these 13 cases (7.3% of the 179 audible responses overall), only the response of the mainly addressed recipient of the RfC was considered for counting the types and frequency of the response tokens used (see the coding manual, König et al. forthcoming). Nearly 85% of the responses contain at least one (n = 120/179, 67.0%) or more (n = 32/179, 17.9%) response tokens. In the latter case, the tokens can be produced separately or as multiple or complex clusters of response tokens. Only 27 responses (15.1%) do not make use of any response tokens, illustrating their significance for (dis)confirming responsive actions in Czech.

In the 152 responsive turns containing response tokens, we distinguished the production of single tokens, of multiple single tokens (i.e., either with other lexical material in between or with prosodic breaks), and of complex forms containing several tokens uttered without any prosodic break. This led to an overall number of 168 response tokens in the 152 turns considered. The majority are produced as single tokens (n = 137/168, 81.5%), with the remaining 18.5% (n = 31/168) being clusters of two or more tokens within a prosodic unit. The most frequently used response token is no ‘yes’ (n = 65/168, 38.7%) or complex forms starting with it (n = 15/168, 8.9%), making up nearly half of all response tokens used (n = 80/168, 47.6%). The second most frequent token is jo ‘yes’, with overall 31 cases or 18.5% of the responses containing response tokens (n = 24/168, 14.3%, as single tokens and n = 7/168, 4.2%, as complex forms). The negation token ne ‘no’ is in the third position, with overall n = 25/168 (14.9%), of which 17 are used for disconfirming (n = 17/25, 68%). This is followed by non-lexical tokens, mainly the affirmative and disyllabic token mh_hm, typically produced with a fall on the first and a rise on the second syllable (n = 19/168, 11.3%), its negative counterpart occurring only twice (n = 2/168, 1.2%), being once produced as monosyllabic, once as a disyllabic token. These four tokens correspond to the four most frequent ones in spoken Czech in general (ne, hm, jo, no, according to Cvrček et al. 2010, 299). The rest of the lexical tokens used for responding are other variants of ‘yes’ (n = 3/168, 1.8%) and a variety of confirmation tokens (n = 8/168, 4.8%), such as právě ‘right’/’exactly’, tak ‘so’, or přesně tak ‘exactly so’ (for an overview on response particles in Czech see Čermáková et al. 2019). Table 17 gives a detailed list of all response tokens in our data.

The list in Table 17 illustrates that more than two thirds of all the tokens used for responding (n = 114/168, 67.9%) correspond to affirmative response tokens. The fact that no is the most frequent, not only among these, but also overall, emphasizes its potential to display a higher or independent epistemic stance of the speaker with respect to the confirmable (Müllerová 2022b (1996), 233, Čermáková et al. 2019, 296). This can be correlated with its epistemically weaker counterpart jo (which has been described to express weak agreement or to function as a continuer, Čermáková et al. 2019, 286), which is also the most frequently used tag appended to RfCs (Table 9, Section 4.6). However, and parallel to this, jo does mainly respond to RfCs that do not contain any tag (n = 22/31, i.e., 71%). Indeed, jo tends to be used when confirming epistemically low facts, or issues that are not treated as being highly implicative for further elaboration, which is the case in Excerpt 4, in which Nora (NOR), right after a possible closing of a previous larger telling by Jana (JAN) (lines 01-06), opts out of the conversation by telling that she has to take care of some chores. This is receipted by a go-ahead from JAN and a follow-up RfC (line 12).

While JAN’s first response to NOR (line 10) provides a go-ahead and permission for leaving the table (note also the use of jo as an affirmative, but epistemically neutral receipt of the prior turn, line 08), her following RfC (line 12) requests confirmation with respect to the type of chore, namely doing laundry. However, this is already clear from NOR’s previous turn in which she refers to the washing machine (pračku) and picks up on the temporal indication ještě zas ‘once again’, thereby further demonstrating JAN’s acceptance of NOR’s account for getting up from the table. Therefore, this confirmable is in no need of a true epistemic confirmation or further elaboration, and NOR simply minimally confirms this RfC with a double jo (line 13), then leaving the room. The double jo (within one intonation unit, contrary to the two separate jo, line 10) here possibly refers to the evident nature of the RfC and the definitive closing of the sequence, cf. Stivers 2004, or to the fact that the response is given by NOR while already leaving the room, thus being possibly adjusted to speaking-while-moving (cf. Mondada 2014). The exact difference between the production of single vs multiple response tokens (Table 17 indicates 19 cases of double or more jo, no or ne) would merit a separate investigation on a larger corpus and in other types of question–answer sequences.

In case the RfC alludes to a sequentially more implicative epistemic asymmetry, the preferred response token is no, as illustrated by the following simple example (Excerpt 5), in which Milan (MIL) introduces a new conversational topic, enquiring about how ‘this development’ went (line 01). Karel (KAR) initiates repair with an RfC (line 02), suggesting a specification of the noun ‘development’, namely, if it indeed refers to the soccer match he and his friends just played. This is confirmed by a minimal response using no (line 03):

If the RfC presents an understanding check or initiates repair (such as in the previous example), the response token no therefore confirms and positions its speaker in an epistemically higher position than that of the speaker of the RfC. For this reason, no can also respond to possibly challenging RfCs (and might, in that sense, be better translated as ‘yes indeed’). In Excerpt 6, Jana (JAN) writes a text message in front of her mother Ludmila (LUD). While JAN voices her writing throughout her typing activity, LUD displays her disapproval of this individual activity by commenting on the address term JAN uses for her friend, ‘little bug’ (line 02):

While the RfC (line 06) itself has a typical, epistemically unmarked format (see the use of the jo-tag), its larger sequential context and laughter particles frame it as being a challenge (cf. Küttner and Ehmer 2023). Shortly after JAN has voiced the beginning of her text message (lines 01-02), LUD initiates an accusative turn referring to something her daughter should possibly account for (line 04). Note that no can also be used for projecting a possible disalignment or modification (and would thus in this case be better translated as ‘well’). The pre- and post-positioned laughter particles (beginning of lines 6 and 10) also convey LUD’s stance toward this address term as being potentially inappropriate. Her RfC indeed alludes to something that is already known to her (as it has just been voiced by JAN, line 02), therefore clearly not pointing to an epistemic asymmetry. This RfC is responded to by a plain no (line 07), and, once the text message has been finalized, an extended account re-using the no (lines 12-13), claiming JAN’s independent decision regarding the address term and displaying her rejection of LUD’s opinion.

Prefacing elements to the responsive turns to RfCs are very rare (n = 9/179, 5%; Table 18), these elements being all non-lexical, such as tongue clicks (Ogden 2013, see also Ben-Moshe and Maschler 2024 for Hebrew) or hesitation particles. It should be noted that the heterogeneous audio quality of the recordings and responsive turns produced in overlap might mean that a few more responsive turns make use of non-lexical prefacing elements, especially tongue clicks. Nevertheless, prefacing responsive turns can be said to be not very important for (dis)confirming in Czech.

5.3 Clusters of response tokens

As has been already shown in the previous section (Table 17), clusters of response tokens are less frequent than single tokens. Among the 152 responses to RfCs containing response tokens, only 32 (21.1%) contain two or more tokens or complex forms (Table 16, Section 5.2). The complex forms or clusters refer to a series of two or more response tokens that are uttered without a pause or clear prosodic break; i.e., they are produced as one intonational unit.

In our data set, the 24 cases of cluster responses (15.8%, Table 19) consist of either multiple versions of the same token (Stivers 2004), such as jo_jo; or no_no_no; (15 cases), or of series of different tokens, such as no:_jasně; or no:_právě; (9 cases). These seem to be related either to RfCs containing more reformulated and self-repaired elements (and thus syntactically less smooth RfCs), or to a series of RfCs related to a similar topic, in which subsequent RfCs are responded to with increasingly multiple tokens (Stivers 2004).

5.4 Position of the first response token

The position of the first response token is overwhelmingly initial (132 cases or 86.8% out of the 152 responses containing response tokens), followed by 15 cases with tokens in final position (9.9%) and only 5 with tokens in mid-position (3.3%, Table 20). It should be noted that among the 132 responsive turns with an initial token, 45 are minimal responses in that they contain only one or more response tokens, but no other lexical material (cf. Section 5.5). In these 45 cases (29.6%), the initial position is therefore a default one, as opposed to the 87 (57.2%) remaining occurrences representing non-minimal responses.

Of the 15 cases with final response tokens, 11 contain a full or partial repeat of the RfC, such as illustrated by Excerpt 7 (line 09). This holds also for three out of the five cases with a response token in mid-position. It should, however, be noted that the combination of an initial response token and a (partial or full) repeat is overall more frequent (27 cases). In general, response tokens can be positioned freely around partial and full repeats of the RfCs (overall 41 cases, i.e., 27.0% out of 152 containing various forms of repeats, see also Section 5.6).

The position of the response token indeed relates to sequence management. An initial response token treats the confirmation as the first and thus most (or solely, in case of minimal responses) relevant responsive action. In Excerpt 7, within a longer update that Jana (JAN) gives about her family, Nora (NOR) enquires about her brother Martin’s whereabouts (line 01), to which JAN provides an answer (lines 02-03).

NOR’s RfRC (Gipper et al. 2024) is only minimally responded to (lines 05–06), not leading JAN to elaborate on this apparent news for NOR. NOR subsequently offers a topic-proffering RfC (‘and what- he has a job there yeah’, line 08). Instead of immediately confirming this with a response token, JAN provides a partial repeat of the RfC and appends the response token no (line 09). It cannot be said if JAN treats NOR’s RfC as simply being rather obvious (work being one of the most frequent reasons for moving) and therefore in no need of elaboration, or if she rather resists elaborating on this topic at this moment. In any case, a turn-final response token displays an orientation to sequence closing. This is also visible in the fact that in the next position, NOR offers yet another topic-proffering turn, trying to elicit more information about the brother beyond the preceding sequence closing (line 11).

5.5 Minimal and non-minimal responses

Responses to RfCs can be minimal, meaning that they consist of a single (or more, cf. Section 5.2) response token, which is the case for 25.1% of the 179 RfCs having received an audible response. A larger part of the RfCs receive a more elaborate response, in that the response token(s) is/are accompanied by more lexical material providing a possible answer to the RfC. These non-minimal responses (n = 107/179, 59.8%) are mostly formatted with a response token in the initial position, followed by a more elaborate response (n = 87 out of 107 cases, cf. Section 5.4, Table 20). The remaining 27 examples (15.1%) contain elaborate responses without any response token, corresponding to the coding category ‘other’ (Table 21).

More than half of these ‘other’ examples consist of partial or full repeats of the RfCs (n = 16/27), the other 11 cases provide different types of responses that contain lexical material that is fully different from the one provided in the RfC. Looking at the distribution of all response types with respect to their confirming potential, all minimal responses are confirmatory (n = 45/45, 100%, Table 22). For the other response types, we can notice more variation, although confirmation is also the most frequent action carried out both by non-minimal (n = 83/107, 77.6%) and by other responses (n = 17/27, 63.0%).

Table 22 shows that disconfirmations are typically in need of more lexical material, although a first disconfirming move seems to be usually tied to a response token (21 non-minimal disconfirming responses as compared to 5 other-formatted ones). The responses providing neither a confirmation nor a disconfirmation (‘neither’) always contain more lexical material than just a response token (i.e., they belong to the categories ‘non-minimal’ or ‘other’).

While a confirmation can be fully implemented by a minimal response (n = 45 cases), the more frequent way of confirming an RfC in Czech is by a non-minimal response (n = 83 cases). This is not surprising if we consider that RfCs can be part of larger sequences (big packages, Sacks 1992) in which they act as, e.g., topic proffers or requests for information that make an elaboration in the next turn relevant. If an RfC is clearly framed as clarifying a specific and small fact that is relevant, for instance, for following the larger line of talk, a minimal answer to the RfC might be sufficient. In the following example (Excerpt 8), the RfC mainly serves to identify the correct person. While looking at pictures on a smartphone, Marta (MAR) enquires about one of the women who are visible in one of the pictures (line 01). The noun phrase that initiates a repair regarding the referenced person obtains a minimal confirmation, then Yveta (YVE) self-selects again in order to answer MAR’s initial question from line 01 (lines 05–06).

In this case, the minimal response confirms and closes the side sequence (Jefferson 1972), allowing to get back to the main (or a different) line of talk right afterward, or, more generally, to lead the talk on that precise topic toward a closing (see also Excerpt 7 regarding final response tokens). If the confirmable itself is more important to the larger sequence, or if it introduces an up-to-then new aspect (which can be either repaired and/or which considerably contributes to the development of the topic or the overall progressivity of the sequence), the confirmatory response will be more than just minimal (cf. Gubina et al. 2024 for a similar observation on German data).

5.6 Full and expanded repeats

Compared to other languages, for instance, English (Küttner and Szczepek Reed 2024), in Czech, repeated elements in responses to RfCs are more frequent. After all, out of the 179 audible responses to RfCs, 57 (31.8%) of the responses re-use at least some of the lexical material from the RfC, while 122, i.e., 68.2%, do not contain any repeated elements (Table 23).

According to research on English, repeating lexical material from a co-participant’s turn conveys a stronger epistemic position (Heritage and Raymond 2005). As a consequence, repeats in responses to RfCs in English have been suggested to be rare as they might lead to an ‘epistemic overkill’ in the otherwise already epistemically asymmetrical RfC sequence (Küttner and Szczepek Reed 2024, 29). In co-constructional sequences (Lerner 2004), repeating a co-participant’s pre-emptive completion reclaims, on the one hand, authorship and independent continuation of the turn, but, on the other, it also treats the pre-emptive completion as indeed fitting to the emerging and pre-emptively completed turn (Oloff 2018). While RfCs and pre-emptive completions are not the same type of action, (partial) repeats as responses both to a pre-emptive completion and to a specific type of RfC might work in a comparable way. Again, in order to investigate this more thoroughly, a larger number of full and expanded repeats in responses to RfCs in Czech would be needed.