Request for confirmation sequences in Low German

4.1 Syntactic design

In LoG, RfCs are routinely deployed in clause structure (n = 144/200; 72%), especially with declarative syntax, i.e., clauses with verb-second position (SVO structure; n = 108/200; 54%, Table 1). In 60% of the cases (n = 65/108), declarative RfCs are equipped with a tag, as in Extract (1). In 40% (n = 43/108), these RfCs are deployed as ‘declarative questions’ without a tag, a typical syntactic RfC design in Germanic languages (Couper-Kuhlen and Selting 2018, 228).

Section 1 outlines that RfCs often have a ‘double-barreled’ nature. Declaratively formatted polar questions are said to convey a bunch of different social actions in interactions, besides the action of questioning (e.g., Seuren and Huiskes 2017). In the LoG data, declarative RfCs with and without a tag are predominantly used as a means to accomplish three different social actions: (1) topic proffers, (2) understanding displays, and (3) understanding checks. With topic proffers “a speaker proposes a particular topic …, but does not actively launch or further develop the proposed topic” (Schegloff 2007, 169–70). Consider (1). Heidi and Annika are talking about Heidi’s training course. After a sequence closure of the just-prior talk (lines 01–02), Heidi marks a departure from the ongoing sequence and manages a topic change by means of the conjunction ower (‘but’) and the following RfC formatted as a tag question (line 03).

Extract 1: LoG-HES02_06 (00:01:24 – 00:01:27), No. 77

In addition to topic proffers, declarative RfCs are also deployed as understanding displays (Heritage 1984, 320–3; see also Zinken and Küttner 2022). With this action type, speakers “wish to show that prior talk has been adequately descriptive and/or they have competently understood its import” (Heritage 1984, 321). Typically, these displays are deployed as confident inferences prefaced with a change-of-state token indexing independent ‘then and there’-understanding “prior to any possible confirmation” (Heritage 1984, 321). Consider (2). Helga refers in line 01 to a recent meeting with a person called Tim, implementing the action of delivering news to Ben. In lines 06–07, he deploys a confident inference prefaced with the change-of-state token ach ja (‘I see/I remember’) to index an independent understanding of the class reunion Helga was referring to in line 01. Because of this index of independence, understanding displays require subsequent (dis)confirmation to a lesser extent compared to understanding checks (Heritage 1984, 323).

Extract 2: LoG-GEL01_02 (00:00:01 – 00:00:09), No. 48

Finally, declarative RfCs are also used as understanding checks in LoG interaction. Following Heritage (1984, 318–20), this article considers understanding checks to be clarificatory repair-like sequences, where a speaker proposes “a candidate understanding of what an earlier speaker had intended, the producer of an understanding check thereby invites that speaker to confirm (or disconfirm) the adequacy of that proposal” (Heritage 1984, 319). Consider (3). Annika deploys a topic proffer in the form of a telling question referring to a meeting that Heidi attended. Heidi embraces the topic and starts to elaborate on the event in lines 03–04. The beginning of Heidi’s elaboration overlaps with an understanding check of Annika in line 05 as a side sequence, which requires clarification of the organizing institution of that meeting.

Extract 3: LoG-HES02_06 (00:00:08 – 00:00:13), No. 75

Besides declarative formats, RfCs in LoG are also produced as phrases in about a quarter of cases (n = 49/200, 24.5%), with the majority being nominal (e.g., dat Grüne ‘the green one’) or prepositional in nature (e.g., mit Auto ‘by car’). 28.5% of all phrasal RfCs are used with a tag (n = 14/49). In 8% (n = 4/49), these phrases are deployed as a candidate answer (Pomerantz 1988), following an information question as in (4). Previous to the extract, Lukas told Tina that he regularly attends swimming courses.

Extract 4: LoG-BÖR01_07 (00:03:45 – 00:03:48), No. 19

In contrast to the functional diversity of declarative RfCs, phrasal RfCs are the predominant format to accomplish understanding checks. Consider Extract (5). Following Anna’s narrative about a guided tour of a German food company (lines 01–02), Heinz launches a “side sequence within an ongoing sequence” (Jefferson 1972, 294) in order to clarify the referential food company by proposing the candidate bi apetito (‘At apetito’s’) in the form of a phrasal RfC. Anna confirms the RfC in line 05 by using a full repeat, thus closing the side sequence and moving back to the main line of action in line 06.

Extract 5: LoG-HES01_07_09 (00:23:12 – 00:23:20), No. 108

RfCs are also formatted with interrogative syntax (VSO structure) in the LoG data (n = 27/200; on interrogative syntax as a typical format of RfCs, refer to Couper-Kuhlen and Selting 2018, 218). In particular, polar interrogative RfCs are predominantly used to accomplish understanding displays, presenting an inference or a highly likely knowledge about a matter at hand. Consider (6) (larger Extract (6) in 4.2). In line 01, Paul deploys a candidate understanding of Ben’s prior turn formatted with interrogative syntax and a negative polarity item. In contrast to declarative understanding displays, interrogatively formatted inferences are not prefaced with change-of-state tokens indexing an independent understanding. Thus, interrogative RfCs require subsequent (dis)confirmation to a greater extent. This is in line with the idea that interrogative questions propose a steeper epistemic gradient than declarative questions (Heritage and Raymond 2012).

Extract 6: LoG-GEL01_02 (00:07:50 – 00:07:52), No. 57

In only 4.5% of the data (n = 9/200), RfCs are designed with an SOV structure (subordinate clause), like in Extract (7). Ben launches a topic change to an old photo of a mutual aunt and uncle’s wedding party. Within a side sequence, Helga starts a remedy of the referred event by deploying an RfC in SOV structure (line 01), which is minimally confirmed by Ben in line 02.

Extract 7: LoG-GEL01_02 (00:16:06 – 00:16:07), No. 61

Elliptical confirmables make up only a small part of the corpus (n = 7/200, 3.5%). Consider (8). Concerning the social actions that elliptical RfCs and RfCs formatted as subordinate clauses accomplish, no clear picture has emerged in the analysis.

Extract 8: LoG-BÖR09_11 (00:01:04 – 00:01:08), No. 125

In sum, in LoG RfC designs are typically formatted with declarative and phrasal syntax, but also – to a lesser extent – with interrogative syntax, subordinate clauses, or ellipticals.

4.2 Polarity

Research on English has shown that polarity influences the preference constraints and the epistemic stance display of polar questions (on reversed polarity questions, refer to the study by Koshik (2002), and on negatively polarized questions [NPQs], refer to the study by Heritage and Raymond (2021)). NPQs in English can be designed with interrogative or declarative syntax and are supposed to be tilted toward polarity-matched ‘no’-responses. However, Heritage (2002) has shown that negative interrogatives beginning with frames such as Isn’t it, Don’t you or modal phrases like Shouldn’t you X are different from NPQs with regard to the level of preference organization as they strongly invite affirmative responses. Epistemically, these assertive negative interrogatives differ from ‘questioning’ NPQs by (a) referring to shared knowledge, thereby not displaying a relative epistemic gradient to a matter at hand, and (b) the following response design, where responses only briefly acknowledge their interrogative packaging. In sum, specific modulations can have an impact on the preference organization of questions with negative polarity items.

In the LoG data, RfCs are predominantly formatted with positive polarity (n = 176/200; 88%); RfCs with negative polarity occur in only a limited number of cases (n = 24/200; 12%). Table 2 outlines the lexical items of negative polarity in LoG. In 87.5% of the cases (n = 21/24), the item nich (‘not’) is used for marking negative polarity in LoG, followed by the negative pronoun/article kein (‘no’) and the negative adverb nix (‘nothing’). The item nich falls into two distinct classes regarding its semantic and syntactic scope: it is used as (1) a negation particle (n = 18) and (2) a modal particle (n = 3). The following analyses will focus on these two usages of nich in RfCs and their preference organization. RfCs with the negation particle nich in LoG are directed to states of affairs that are unlikely by virtue of prior talk and thus inferential in nature (Heritage and Raymond 2021, 46–51, Drew 2018; see also Section 4.4). Consider again Extract (6). Ben and Paul are talking about their acquaintance, Margot. In lines 02–03, Ben reacts to Paul’s information request (line 01) with two epistemic disclaimers (lines 02–03). In line 04, Paul tries again by deploying the RfC kümmst da nich mit tousamen (‘you’re not coming together’). Based on his understanding of the just-prior turn, Paul marks a contact between Ben and Margot as unlikely. Constructing this RfC with the negative polarity item nich (‘not’) displays an orientation toward a ‘no’-type answer. Ben confirms the RfC in line 05 with the response token nä (‘no’). In sum, all NPQs with the negative particle nich (‘not’) are formatted with either declarative or interrogative syntax. When NPQs are confirmed in LoG, they are responded to with a polarity-matched ‘no’-type response token (Extract 14 shows a disconfirming response to a prior NPQ).

Extract 6: LoG-GEL01_02 (00:07:46 – 00:07:54), No. 57

On the other hand, RfCs with the modal particle nich (‘not’) are directed to states of affairs that speakers treat as (highly) likely. In its function as a modal particle, nich (‘not’) has lost its negative semantic scope and therefore its negative polarity in the course of grammaticalization (Thurmair 1989, 160–3). Compared to RfCs with negation particles, these RfCs are different with regard to preference organization, as they are tilted towards a ‘yes’-type response. All of the nich (‘not’) modal particles occur in RfCs formatted with interrogative syntax. Consider (9). In line 01, Lukas launches an RfC regarding the mode of transportation Tina and Theo used on their last vacation. Both confirm the RfC in lines 02 and 03 with a ‘yes’-response token. Thus, RfCs with the modal particle nich (‘not’) are not to be counted as NPQs in the sense of Heritage and Raymond (2021).

Extract 9: LoG-BÖR01_07 (00:14:39 – 00:14:43), No. 32

As can be seen, making a B-event statement by means of negative polarity is applied (a) rarely and (b) with a small spectrum of lexical items in LoG. The item nich is used in two different ways inviting different response types. RfCs with the negation particle nich are directed to states of affairs that are treated as unlikely, meaning they are tilted towards a ‘no’-response, as previously described for NPQs. However, RfCs with the modal particle nich are tilted towards a ‘yes’-type response, comparable with Heritage’s (2002) negative interrogatives. The difference between Heritage’s (2002) negative interrogatives and LoG RfCs is that the first is treated as accomplishing opinion statements to an object of common knowledge, while the latter is still understood as questioning in the information-seeking sense.

4.3 Modulation

Compared to negative polarity, marking epistemic modulation in RfCs with modal particles, modal words or modal phrases is more pronounced in LoG, with modulations occurring in approximately a third of all cases (n = 60/200; 30%). In five of those instances, two modal items occur in one RfC, resulting in a slightly higher total number of markers (n = 65).

Table 3 provides an overview of the lexical elements that serve as modal markers in LoG RfCs. On the one hand, the data suggest a structural convergence with HG lexical material due to variety contact (see also Goss and Salmons (2000) on the loss of modal particles in long-term language contact). The HG modal particles ja (‘yes’) and doch (‘yes’) are the most frequent resources. Moreover, the modal particle nich(t) (‘not’) and the modal phrase ich glaube/ik glöw (‘I believe’) are also typical means of modulations in LoG RfCs. With the modal particle ja (‘yes’), speakers activate a communal or personal common ground in RfCs, indexing that the presented knowledge is shared and highly likely (see Reinecke (2018) on ja (‘yes’) in HG). Consider (10). In line 01, Paul deploys a declarative + tag RfC with the modal particle ja (‘yes’), indexing shared knowledge and a relatively strong assumption about the proposition.

Extract 10: LoG-GEL01_02 (00:07:44 – 00:07:46), No. 56

The modal particle doch (‘yes’) is also deployed as a means of marking common ground in LoG RfCs, strongly committing to a proposition (Pittner 2007, 75–6). Only in a few cases, speakers contradict a prior turn by using doch (Extract 15). Consider (11). With the RfC in lines 01–02, Ben initiates an activity of co-remembering by referring to an event in a club called Eiskümmel (‘ice cumin’), at which Paul is presumed to have met his wife Sandra for the first time. The modal particle doch indexes this situation as part of the common ground.

Extract 11: LoG-GEL01_02 (00:16:34 – 00:16:39), No. 66

On the other hand, Table 3 reveals LoG-specific modal markers which are not applied in HG (see Deppermann et al. 2024). This includes the modal particles wall [LoG]/wohl [HG] (‘probably’) and nu (‘now’). Weber (2020b) shows that with the modal particles wall [LoG]/wohl [HG], LoG speakers can index an epistemic stance towards the assumed high probability of a matter at hand ([K+] stance; Section 2). Consider (12).

Extract 12: LoG-BÖR01_08 (00:02:18 – 00:02:25), No. 02

In lines 01–02, Tina speaks to Jakob about a teacher who offers LoG courses for children, providing his name and his origin. She frames this knowledge as part of the common ground by using the modal particle ja (‘yes’). After a short delay, Theo, who also has knowledge about this teacher, adds an RfC in line 04 with the modal particle wall and the tag nich (‘not’). Theo expects Jakob to know the person in question, thus treating Tina’s action of informing Jakob as inappropriate.^[2] However, Theo’s following pursuit, Jakob’s delay in responding, and his qualified response joa (‘yes (mitigated)’) in line 10 show that Jakob does not have the expected knowledge about the teacher that Theo presupposes.

In sum, LoG speakers deploy modal markers frequently in order to index epistemic stances within RfC actions. Thereby, structural resemblances between LoG and HG due to variety contact but also LoG-specific markers in the question design of RfCs are noticeable.

4.4 Inference marking

Section 4.2 reveals that RfCs with negative polarity are mainly used when formulating inferences from prior talk. There are other linguistic resources in LoG, which overtly mark RfCs as inferential actions. Following Drew (2018), this article understands inferences as ubiquitous interactional actions, where a speaker attributes an implication to a prior speaker’s turn. Overall, RfCs in LoG are only rarely marked as inferences (n = 19/200; 9.5%). Table 4 provides the different lexical resources that LoG speakers apply to introduce inferential displays. Most of these resources are also deployed in HG, including more implicit markers (e.g., also (‘so’; n = 3)) as well as explicitly attributing phrases (e.g., du meinst/meenste (‘you mean’; n = 2)).

For the LoG data, it is noticeable that speakers frequently use change-of-state tokens as explicit indexes of inferential actions. Extract (13) gives an example of an achso (‘oh I see’)-plus-RfC format. Ben informs Paul about the professional career of an acquaintance’s daughter (lines 01–03). In line 04, Paul reacts with the overt understanding marker achso (‘oh I see’; Golato 2010, 157–61), equipped with accentuation on the second syllable and falling intonation in a separate TCU. After that, he deploys a candidate understanding in line 05. In the LoG data, achso-plus-RfC formats are always designed as inferences of prior turns-at-talk and therefore as understanding displays (Heritage 1984, 320–3).

Extract 13: LoG-GEL01_02 (00:02:40 – 00:02:49), No. 52

In sum, RfCs in the LoG data are only rarely (linguistically) marked as inferences from a prior speaker’s turn. It is, however, worth highlighting that the speakers use change-of-state tokens frequently as inferential markers. LoG seems to provide almost no variety-specific resources for the explicit marking of inferences. Speakers therefore fall back on HG lexemes.

4.5 Connectives

In 37.5% of all cases, LoG speakers introduce RfC sequences with connectives like conjunctions, subjunctions, discourse markers, or change-of-state tokens (n = 75/200). Table 5 reveals the different connective formats. The conjunctions ower [LoG]/aber [HG] (‘but’; n = 16) and un(d) (‘and’; n = 15) show the highest absolute frequency in the data. Moreover, change-of-state tokens are used frequently as connectives in the LoG data (n = 12). In the following, we focus on the most frequent conjunction ower/awer [LoG] (‘but’) as a turn-initial and TCU-initial practice in LoG.

In the majority of cases, turn-initial ower/awer (‘but’) in LoG is used as a conjunction marking a departure from the ongoing course of action, as Extract (1) shows. In this example, Heidi marks a departure from the ongoing sequence and manages a recipient-oriented topic change (Schegloff 2007, 170) by means of the turn-initial conjunction ower (‘but’). After a [K–]-display of Annika, Extract (14) starts immediately afterward, where Heidi tries a second time, explaining that she saw Annika passing by in the car. With the RfC in line 02, Heidi deploys an awer (‘but’)-prefaced inference in the TCU-initial position. Thereby, Heidi launches a different line of talk that presents her attribution to Annika’s behavior as a lack of mutual perception and thus a presupposed [K–] status of Annika. After this second proffer, Annika embraces the topic by strongly disconfirming Heidi’s presupposition in lines 03–05 with a [K+] display.

Extract 14: LoG-HES02_06 (00:01:31 – 00:01:34), No. 78

In some cases, however, turn-initial ower/awer (‘but’) marks a contradiction to a prior turn-at-talk. Consider (15). Heinz, his sister Anna, and his niece Svenja are talking about a priest from the village. Anna disaffirms Heinz’s request for information in line 03 with the negative response particle nä (‘no’) and a concessive format, leading to a sequence closure. After a short delay, Heinz resumes the topic by deploying an RfC, prefaced by the sequential conjunction aber (‘but’) in line 04. In contrast to turn-initial ower (‘but’) in Extract (1), with aber (‘but’) in (15) Heinz directly ties his RfC to the prior sequence and the ongoing course of action by contrasting his knowledge with Anna’s prior turn. By co-constructing it with the modal particle doch (‘yes’), Heinz displays a strong commitment to the proposition.

Extract 15: LoG-HES01_07_09 (00:05:47 – 00:05:54), No. 98

In sum, conjunctions are part of LoG RfCs formats in about one-third of all cases. Focusing on awer/ower (‘but’), the analysis shows that it can be used both as a turn- and a TCU-initial resource indexing (a) a departure from the ongoing course of action (Extract 1, line 03), (b) a contradiction to a prior speaker’s turn (Extract 15, line 04), or (c) a projection of a contrasting proposition within a speaker’s turn (Extract 14, line 02). All uses index a contrast in one way or another, either to a previous turn-internal proposition, a prior turn, or to a prior topic.

4.6 Tags

In LoG, RfCs are frequently equipped with tags in 45% of the cases (n = 90/200). Table 6 illustrates the different tag formats that occur in RfC question designs in the data.

With regard to the formal resources, only five different lexical items/phrases are discernible: ne (‘not’), nich (‘not’), oder (‘or’), oder wat (‘or what’), oder wann (‘or when’), and kann dat (‘is that possible’). Tags like wa or woll, considered typical Northern German tags (Section 2), do not occur in the data. However, this is in line with Harren’s (2001, 104–5) study on wa, which shows a frequent occurrence of this tag with assessments and not with RfCs.

Concerning the frequency ratios, the supraregional ne-tag prevails in the LoG data (n = 69/90; 77%) followed by the lexically similar, more regional tag nich (n = 11/90; 12%). While Harren (2001) shows a frequent co-occurrence of ne-turns with modal particles in HG, modulations in tag-attached RfCs in LoG are found in only 39% of cases (n = 27/69). In line with Harren (2001, 128), the most frequent co-occurring modal particles with ne-tagged RfCs are ja (‘yes’; n = 12) and doch (‘yes’; n = 8). Both serve as markers of an established common ground (Section 4.3). With respect to nich, the relative frequency ratios look similar: 36% (n = 4/11) of all nich-tagged RfCs are constructed with modal particles, while 64% are not (n = 7/11). Sequential analyses of RfCs with ne- and nich-tags in the LoG data show that both exhibit a wide range of similar social actions. Thus, the deployment of ne and nich in RfCs is not driven by the accomplishment of different social actions in LoG. In addition, ne and nich are deployed with similar final intonation that is predominantly rise-to-mid and rise-to-high intonation, as indicated in Table 10. In sum, the analyses could not reveal a robust link between final intonation contour and social action. However, for the second most frequent tag, nich (‘not’; n = 11), an age effect is evident in the data, as only older speakers use nich, while younger speakers exclusively use the tag ne. This apparent-time comparison^[3] supports the hypothesis of a phonetic erosion from nich to ne, outlined in Section 2, at least for LoG RfC actions. Accordingly, variety contact might cause a decline of the northern regional variant nich in favor of the supraregional variant ne.

Compared to ne (‘not’) and nich (‘not’), the tag oder (‘or’) with only 7% of the cases (n = 6/90) is not strongly conventionalized in LoG interaction. Almost all oder-RfCs in LoG are responded to with minimal responses. In only one instance, the oder-tagged RfC solicits topic talk and thus a non-minimal answer. Moreover, compared to HG, there is less of a connection between the final intonation of oder-RfCs and the complexity of the following actions in LoG (Section 4.7). Again, an age effect in the usage of the oder-tag is observable, as it is deployed only by younger speakers. It is therefore likely that the oder-tag is an interference from HG and a rather recent tag in LoG.

In sum, while tags are frequently used resources in LoG RfCs, the spectrum of items that serve as tags is noticeably small. The deployment of the two most frequent tags, ne and nich, is not driven by action or prosody. As noted, an age effect in the use of the tags nich and oder is evident. The apparent-time comparison for nich supports the hypothesis of a phonetic erosion from nich to ne. Accordingly, variety contact might cause a decline of the more regional variant nich in favor of the supraregional variant ne. The tag oder is mainly used by younger speakers and does not play a decisive role in the LoG RfC design (yet).

4.7 Prosodic design

In addition to the different tag-types, the attachment of tags also impacts the prosodic design of RfCs. Table 7 reveals that the majority of RfC confirmables^[4] with or without a tag are equipped with a final falling intonation contour in LoG. However, confirmables without a tag show a higher relative frequency ratio of rising contours. In addition, Table 8 shows the distribution of tags with regard to prosodic integration into the anchor sentence (core proposition). In 98% of the cases, tags are not prosodically integrated. Moreover, Table 9 reveals that tags mainly occur with a final rising contour in 90% of the cases. A more fine-grained prosodic analysis of the tags ne and nich, analyzed in Section 4.6, reveals that they are predominantly deployed with similar final intonation, namely rise-to-mid and – to a lesser degree – rise-to-high intonation (Table 10). One exception in the data is the tag oder was (‘or what’), which is always produced as a prosodically integrated tag. In LoG, all non-integrated tags are latched to the confirmable such that no notable pause between the confirmable and the tag occurs. Overall, the data do not contain any cases of tags as resources to accomplish response pursuits after silences.

Concerning RfCs with a tag, a stable pattern of final falling confirmables with an appended final rising is documented in the data. Notable exceptions concern oder (‘or’), and oder was (‘or what’), for which level and falling contours can be observed. Contrary to the study on HG oder by Drake (2016) and König (2020), the only oder-tagged RfC in the LoG data, which solicits topic talk, is not deployed with a level contour – as assumed by both studies – but with falling intonation. Thus, a robust link between the final intonation of oder (‘or’) and subsequent actions cannot be proclaimed for LoG RfCs. This may also be due to the small number of instances and would have to be investigated in more detail in a follow-up study.

5.1 Responsive actions

When speakers ask polar questions like RfCs in interaction, they make confirmation or disconfirmation relevant as a next action (e.g., Raymond 2003, Enfield et al. 2019). In LoG interactions, RfCs predominantly receive confirmatory responses (n = 141/200; 70.5%), with 60% being response tokens^[5] (n = 85/141), 26% full(-expanded) repeats (n = 36/141), and 14% token + repeat combinations (n = 20/141). Disconfirming responses occur rather rarely in the data (n = 19/200; 9.5%), with 79% being response tokens (n = 15/19), 10.5% full-expanded repeats (n = 2/19), and 10.5% token + repeat combinations (n = 2/19). Only four RfCs do not receive a response in the data (n = 4/200; 2%). Since the data could only be analyzed in audio form (see Section 3), these four instances might include pure embodied responses like nods. This finding may therefore be biased due to the data format. Moreover, in 18% of the cases, responses neither confirm, nor disconfirm the previous RfC (n = 36/200; n = 8 response tokens, n = 2 full(-expanded) repeats, n = 1 token + repeat combination, n = 25 other types). These consist of epistemic disclaimers (e.g., ik weet nich ‘I don’t know’), conditional confirmations (e.g., joa; Gipper and Groß 2024), additional information (e.g., un die dochter ‘and the daughter’), concessive response formats (e.g., ausgeschlossen is dat nich, aber ‘this is not excluded, but’), ja aber (‘yes but’)-constructions (Szczepek Reed 2015), and corrections (e.g., A: sechs ‘six’?; B: sieben ‘seven’). The following analyses in Sections 5.2–5.6 will be based on the 196 instances where a response follows the RfC action. Consequently, the denominator will be 196 instead of 200.

5.2 Response tokens

Literature on answering systems distinguishes two semantic systems, a truth-value and a polarity system, as well as two-formal systems, a non-echo (‘interjection type^[6]’)-system and an echo (‘repetition’)-system (Jones 1999, 42). However, Enfield et al. (2019, 280) criticize this rigid division into interjection vs repeat systems, since most languages use both formats in interaction, just with varying degrees of frequency. Following Raymond (2003), RfCs as a subtype of polar questions prefer equivalents of ‘yes’- or ‘no’-response types. To this point, research distinguishes between marked and unmarked interjection options (Heritage and Raymond 2005, 2012, Stivers 2011, 2022, 91–121). With unmarked (minimal) response tokens, speakers just do the action of (plain) confirmation to a prior question (Raymond 2003). Marked response tokens can be subdivided into upgraded, downgraded, and acquiescent interjections (Stivers 2022, 93), challenging the asking itself, the answerability or the action agenda of a prior question. In this section, we will analyze different response tokens and the interactional work they do as part of the ‘answer possibility space’ (Stivers 2022, 3) in LoG interaction. We will learn that both response formats play a strong role in the data.

In 67% of the cases in the LoG data, the responses include at least one response token of a ‘yes’- or ‘no’-type (n = 132/196 [with repeat combinations; for details on the minimality of response tokens see Section 5.5]; n = 105 confirming; n = 17 disconfirming; n = 10 neither confirming nor disconfirming; n = 109/196 [56% without repeat combinations]). In the following, we will focus on confirming and disconfirming response tokens (n = 122). Research on answers to polar questions stresses the importance of a question’s polarity for the preference organization and the response constraints (e.g., Jones 1999, 8–14). Thus, Table 11 first provides the confirming and disconfirming response tokens and their variants in response to RfCs with positive polarity (n = 104). The quantitative assignment is based on the first response token of the responsive action.

The data show that RfCs with positive polarity are predominantly confirmed with variants of ‘yes’-response tokens. When studying variety contact of response particles, the assignment of response tokens to one or another variety is methodologically challenging. The historical development and the regional distribution of response tokens are important factors to consider. Regarding the variety contact between HG and LoG, Table 11 reveals that the HG token ja (‘yes’) prevails in the data, followed by the LoG variant jo (‘yes’; Damme [2015, column 463] for the classification of jo ‘yes’ as a LoG response token). Similar to modal markers (Section 4.3), a growing structural convergence of LoG toward HG is observable, especially when it comes to combinations of response tokens (e.g., ja genau ‘yes exactly’ or ja eben ‘yes as I said’). In contrast to these forms of structural convergence, the unmarked response token hm_hm, which Deppermann et al. (2024) describe as very common for HG, is not frequently deployed in the LoG data. Regarding disconfirming responses in Table 11, the tokens nee (‘no’) and nä (‘no’) prevail. Historically, nee [ne:] (‘no’) has its roots in LoG (nēn in Middle LoG; see Kluge and Mitzka 1960, 506). However, today, it is used as a supraregional response token (Elspaß and Möller 2004; see also Imo 2017 and Deppermann et al. 2024 on nee in HG data). As for nä (‘no’), there are no empirical studies on its regional distribution or variety-based specifics. In the study by Damme (2015, column 463), nä (‘no’) appears in a lexical entry in the Westphalian LoG dictionary. Its origin, however, seems to be in the Upper Saxon area (Adelung 1798, column 463). Today, nä [nɛ] (‘no’) also seems to be used both in LoG and in HG (Imo 2017 on nä in HG). In the following, we will not treat the use of nee (‘no’) and nä (‘no’) as variety specific. The HG variant nein (‘no’) occurs infrequently in the data, which is in line with existing research (Imo 2017, Deppermann et al. 2024).

In the LoG data, the HG variant ja (‘yes’) is deployed as both a minimal response and a non-minimal response, predominantly in combination with a repeat (see Section 5.5 for a discussion of token + repeat combinations). Minimal HG ja (‘yes’) occurs as an unmarked closure-implicative token in two sequential environments. First, it is used in series of little questions, as Extract (16) shows. Jakob and Theo are looking at an old picture of Theo’s parental home. In lines 02 and 04, HG ja (‘yes’) is used as a minimal response token to the previous RfC that prefers a minimal response. These occurrences follow the maxim described by Heritage and Raymond (2012, 184): ‘little questions get little answers’ (see also Heritage and Sorjonen 1994). In the majority of the cases, RfCs in this environment are formatted as tag questions.

Extract 16: LoG-BÖR01_08 (00:05:59 – 00:06:08), No. 5/6

Second, HG ja (‘yes’) as an unmarked response is deployed in understanding checks, realized as side sequences, or what Enfield et al. (2019, 284) call subordinate action sequences. Consider (17). In line 01, Tina manages a recipient-oriented topic change by deploying a telling question to Lukas on a surgery of his acquaintance Samuel. In lines 03–04, Lukas launches a side sequence to clarify person reference by proposing the origin of Samuel with an RfC formatted as a tag question. Tina minimally confirms Lukas’ RfC with HG ja (‘yes’), thus accepting the terms of the prior questioning and closing the side sequence. Lukas moves “quickly back to the main line of action” (Enfield et al. 2019, 291) and starts giving information about the surgery. RfCs prior to HG ja (‘yes’) are predominantly formatted as tag questions or – though less frequently – as phrases.

Extract 17: LoG-BÖR01_07 (00:01:55 – 00:02:03), No. 18

In contrast to HG ja (‘yes’), the LoG response token jo (‘yes’) is for the most part not used as a minimal response token, but in combination with other components – mostly repeats or marked tokens like in (18) (see also Extract 25 in Section 5.5). Prior to the extract, Helga tells Jakob about an LoG teacher from the neighboring village. In line 01, Jakob launches an RfC to clarify person reference, which Theo confirms with the LoG response token jo (‘yes’) and the marked token richtig (‘correct’) in line 03. In contrast to HG ja (‘yes’), LoG jo (‘yes’) is mainly deployed within marked interjection responses to RfCs, claiming agency and epistemic authority over a matter at hand.

Extract 18: LoG-BÖR01_08 (00:02:48 – 00:02:50), No. 4

LoG jau (‘yes’) is deployed as an infrequent minimal response token. In contrast to LoG jo (‘yes’), jau (‘yes’) is more sequence closure implicative – mostly after an extended sequence as in (19). Gisela informs Sebastian in line 01 that the recording device is running. After a first possible sequence closure in line 04, Sebastian resumes the issue in line 06 with the RfC dann dö_t dat nau (‘so it works’), requesting confirmation regarding the proper functioning of the equipment. In line 07, Gisela confirms the RfC with the LoG response token jau (‘yes’), thereby (re-)confirming her prior proposition and closing the sequence. In comparison to HG ja (‘yes’), jau (‘yes’) closes extended sequences and is therefore used as a more marked token in its function as a sequence closure device.^[7]

Extract 19: LoG-STEV02 (00:00:01 – 00:00:09), No. 132

Concerning disconfirming responses to RfCs, the supraregional token nee (‘no’) occurs as the most frequent unmarked response token in the LoG data. In the majority of cases, nee (‘no’) is deployed as a minimal response token prosodically integrated into the following TCU with no particular prosodic mark-up, as in (20). Prior to the extract, Anna tells Heinz that she will visit a mutual acquaintance for her birthday. In lines 01–02, Heinz deploys an RfC, insinuating that she is having a big party on the day of her actual birthday. Anna responds with nee (‘no’) that is prosodically integrated with the following disconfirming action. It is thus used in preface-position (Heritage 1998, 292), projecting same-speaker turn continuation.

Extract 20: LoG-HES01_09_07 (00:00:26 – 00:00:29), No. 86

In contrast to nee (‘no’), the response token nä (‘no’) is frequently used as a marked response, challenging the presupposition or the askability of a prior question. Consider (21). The two siblings, Clara and Benno, discuss whether their uncle needs more care. In lines 01–03, Clara asks Benno if he knew whether their cousin would soon visit their uncle. After displaying resistance to the question’s agenda in line 04, Benno responds to the RfC of Clara in line 05 with the marked interjection nä (‘no’), rejecting the presupposition of the question that he could have knowledge about the matter at hand. A second attempt by Clara in lines 07–08 is also highly rejected by Benno with the HG response token nein (‘no’) in line 09.

Extract 21: LoG-LEV01 (00:15:39 – 00:23:06), No. 168

Extract (21) indicates that speakers use HG nein (‘no’) as a marked response token highly rejecting a prior proposition. Consider also (22). In the preceding context, Tina, Theo’s wife, got up from the table to turn up the heat. Theo then addresses an RfC to Jakob, asking if he is cold. Jakob emphatically disconfirms the RfC with HG nein (‘no’), deployed with prosodic mark-up (increased loudness, rise-fall-contour). He adds an assessment (line 03) and a request for halting the action (lines 04–05).

Extract 22: LoG-BÖR01_08 (00:34:23 – 00:34:25), No. 14

In contrast to RfCs with positive polarity, RfCs with negative polarity are confirmed with variants of ‘no’-response tokens. Table 12 provides the different response tokens following RfCs with negative polarity in the LoG data.

Concerning the different negative polarity items outlined in Section 4.2, RfCs with mere syntactic (e.g., nich) or lexical/derivational negation markers (e.g., kein, nix) show no difference in the preference for upcoming response tokens. With 61% of the cases, the supraregional variant nee (‘no’) is the most common way of confirming an RfC with negative polarity (n = 11/18), followed by nä (‘no’) in 28% of the cases (n = 5/18). Regarding nee (‘no’), considerable differences occur in responses to RfCs with negative and positive polarity. Consider (23). Heidi and Annika talk about their farsighted glasses. In lines 01–02, Annika invokes the common ground that Heidi is not as shortsighted as she is by using an RfC equipped with the modal particle ja (‘yes’). In line 03, Heidi confirms the RfC with the minimal response token nee (‘no’) with falling intonation. While nee (‘no’) following RfCs with positive polarity is predominantly deployed as a preface to other disconfirming actions, nee (‘no’) following RfCs with negative polarity is applied as an unmarked confirming response token in a TCU in its own right.

Extract 23: LoG-HES02_06 (00:11:32 – 00:11:35), No. 80

Nä (‘no’) following RfCs with negative polarity is commonly deployed with an affective stance toward the RfC proposition and at the same time contests the askability of the prior question, that is the question’s presupposition “that both yes and no are possible answers” (Stivers 2011, 86). Consider (24). Svenja talks about her struggle to stop smoking by naming typical situations in which she threatens to relapse from abstinence. In line 04, Svenja confirms Anna’s RfC with the response token nä (‘no’). She thus treats the alternative of the previous RfC that she could suffer a relapse as inconceivable. In sum, nä (‘no’) following RfCs with negative and positive polarity belong to the group of marked interjections (Stivers 2011). Prosodically, nä (‘no’) can be used with more (e.g., Extract 6) or less prosodic mark-up (e.g., Extracts 21 and 24) in LoG interaction.

Extract 24: LoG-HES01_07_09 (00:13:44 – 00:13:49), No. 105

Disconfirming responses to negative polar questions are designed with HG doch (e.g., Pittner 2007, Couper-Kuhlen and Selting 2018, 247–8). Wohl, which is also described as a typical disconfirming response token in LoG and North HG (Turco et al. 2014), was not found in the LoG data.^[8] As for doch in HG, consider again (14). Heidi launches an RfC in line 02, presupposing that there are different knowledge states about her last encounter with Annika. Annika disconfirms the proposition with doch in line 03. She thus claims to have seen Heidi and counters the negative polarity of the prior RfC.

Extract 14: LoG-HES02_06 (00:01:31 – 00:01:32), No. 78 (shortened)

Besides confirming ‘no’-response tokens and the disconfirming item doch, Table 11 also shows one instance of a ‘yes’-type response token used to respond to RfCs with negative polarity. As analyzed in Section 4.2 (Extract 9), ‘yes’-response tokens mainly occur after RfCs with nich as a modal particle and therefore cannot be defined as proper NPQs.

Overall, the LoG data suggest a polarity response system, where the use of response token types is determined by the polarity of the prior RfC. In particular, the LoG response token system is a three-form system, which contains ‘yes’- and ‘no’-type tokens as well as additional forms like HG doch, which “counter the bias of a negative question” (Jones 1999, 42). Due to variety contact, LoG exhibits a large basic positive and negative response system, containing tokens that can be unequivocally characterized as HG (e.g., ja ‘yes’ or doch ‘yes (objecting)’) or as LoG (e.g., jo ‘yes’ and jau ‘yes’). Yet, other tokens are not variety specific (e.g., nee ‘no’ or nä ‘no’). As for non-variety-specific tokens, nee (‘no’) seems to be used as an unmarked token, while nä (‘no’) challenges the askability or the presuppositions of the previous RfC. Regarding variety-specific tokens, it is noticeable that HG response tokens serve as more unmarked ways of responding to a prior RfC, while LoG tokens are used in a more marked way (for an exception, see the use of nein [‘no’], which is also a marked response token in HG). However, we also see that LoG tokens persist more strongly in confirmations than in disconfirmations. Given that confirmations are preferred over disconfirmations (Stivers et al. 2009), a preference-induced effect is observable as the higher frequency of confirmations might enhance persistence with regard to the LoG ‘yes’-token types.^[9]

5.3 Clusters of response tokens

In the LoG data, clusters show a low-frequency ratio (n = 20/132; 15%). They include forms like ja siker (‘yes sure’; n = 5), ja eben (n = 2), ja richtig (‘yes correct’; n = 1), ja genau (‘yes exactly’; n = 2), ja klar (‘yes sure’; n = 1), double sayings like ja ja (‘yes yes’; n = 7), and multiple sayings like nein nein nein (‘no no no’; n = 1) or nä nä nä (‘no no no’; n = 1).^[10] All clusters do more than simply confirm the RfC. They reinforce the (dis-)confirmatory action, display epistemic primacy, or serve as devices to “properly halt course of action” (Stivers 2004, 260).

5.4 Position of the first response token

Section 5.2 reveals that in only 67% of the cases, the responses include one or more response token(s) in the LoG data (n = 132/196; including combinations). In the vast majority of cases, these response tokens occur in turn-initial position (n = 120/132; 91%). Eleven instances are used in the turn-final position, occurring mainly after repeats and being closure-implicative. Only in one instance is the response token deployed in mid-TCU-position.

5.5 Minimal and non-minimal responses

Answers containing a response token in the LoG data are designed either minimally, i.e., solo with a stand-alone response token (n = 43/132, 33%; e.g., Extracts 16 and 17), or non-minimally, i.e., a response token combined with other components (n = 89/132; 67%; e.g., Extract 9). Second response slots (Raymond 2013) in responses to RfCs accomplish many different actions in LoG interaction, e.g., assessments (e.g., Extract 22), corrections, further topic-related informings, accounts, or repeats. In the following, an in-depth analysis of response token + repeat combinations is provided (n = 23/132; 17%). In both combinations, response tokens are almost exclusively limited to confirming ‘yes’-type responses. By contrast, confirming ‘no’-response tokens or disconfirming ‘no’- response tokens are found rarely in combination with repeats. Thus, repeat + response token combinations predominantly accomplish confirmations in LoG interaction.

[‘yes’-token] + [repeat]

Extract (25) illustrates a ‘yes’-token + repeat combination in LoG interaction. In the preceding conversation, Svenja elaborates on an encounter with the local priest. In line 01, Heinz launches a phrasal RfC focusing on the origin of this priest. Svenja confirms the RfC with a [yes] + [repeat]-format. Using jo (‘yes’), Svenja first seems to accept all terms of the prior RfC. However, by combining jo (‘yes’) with a full repeat in the second response slot in line 03, Svenja partially retracts her acceptance by displaying agency over the proposition. She confirms only this spatial alternative compared to other spatially close alternatives that might be projected due to the use of the QT oder (‘or’; Drake 2016). The co-occurrence between LoG jo (‘yes’) and a full repeat reveals that jo (‘yes’) is rather used as a more marked confirmation token in the answer possibility space of LoG speakers (see also Section 5.2).

Extract 25: LoG-HES01_09_07 (00:05:45 – 00:05:47), No. 97

[repeat] + [yes-token]

Heritage and Raymond (2012) describe the deferred use of type-conforming response tokens as a practice to register qualifications. In LoG, repeats in this format do not display qualifications, but highlight epistemic agency in order to manage sequence closure and secure expansion of a previous turn. Consider again (13). By using a repeat, Ben asserts his epistemic rights over the matter at hand, which seem to be endangered by the preceding independent understanding display of Paul. In this excerpt, the speaker not only confirms the RfC, he also confirms that he has alluded to the categorization of these people as Bankmenschen (‘banking people’) with his prior turn in line 03 (Schegloff 1996). However, confirming allusion practices do not prevail in [repeat] + [token] combinations (see also Section 5.6 on confirming allusion practices). The response token in line 08 has a double-barreled nature: it confirms the previous understanding display together with the repeat as a marker of agency, but it also serves as a sequence closure-implicative device (Imo 2013, 188–9).

Extract 13: LoG-GEL01_02 (00:02:41 – 00:02:49), No. 52 (shortened)

5.6 Full and expanded repeats

Yes/no-response tokens are described as the most frequent and type-conforming way of responding to polar questions like RfCs (Raymond 2003, Enfield et al. 2019). However, certain languages tend to be so-called ‘echo systems’, in which repeats serve as the most conventionalized way of confirming a polar question (Sadock and Zwicky (1985) and Jones (1999) on Welsh; Enfield et al. (2019) on Tzeltal). Existing research has described response tokens both as unmarked and as marked types of the answer possibility space (Section 5.2). Repeats, however, are almost exclusively described as marked responses in interaction (Schegloff 1996, Raymond 2003, Stivers 2005, Heritage and Raymond 2012).

‘Markedness’ of repeats is defined at very different levels: (1) on a semantic basis, as repeats assert a prior proposition rather than simply assent to it (Enfield et al. 2019, 284–5); (2) overall frequency, as repeats are often less frequent than response tokens (Enfield et al. 2019, 281), (3) epistemic, sequential, and thematic agency, as recipients claim “epistemic rights over the information required than the original polar question conceded” (Heritage and Raymond 2012, 187; see also Enfield et al. 2019, 292), and (4) sequential positioning, as repeats predominantly do not occur in minimal side sequences, but in ‘primary actions’ that bring new information to the talk (Enfield et al. 2019, 283–4). However, some studies on languages exhibiting a high number of repeats show that repeats are used as unmarked responses in subsidiary sequences (e.g., Keevallik 2010, Brown et al. 2021, Harjunpää and Ostermann 2023). Moreover, these languages also challenge the semantic argument with a minimality–maximality continuum of repeats, where verb-type repeats reassert the prior proposition only minimally (e.g., Hakulinen 2001, Harjunpää and Ostermann 2023). Moreover, Harjunpää and Ostermann (2023, 104–5) argue against the idea of repeats being markers of epistemic agency that in RfCs “the recipient’s greater agentive/epistemic role is sequentially implicated…, which would seem to reduce the need to specifically claim epistemic or agentive rights.”

LoG as a non-standard language exhibits the highest frequencies of repeat responses across the ten languages investigated in the research network (32% of the data; n = 63/196; Pfeiffer et al. Forthcoming) – together with the indigenous language Yurakaré (Gipper, Forthcoming). In the network’s framework only full and full-expanded repeats were included in the study; partial repeats were excluded due to inter-coder reliability issues (König et al. Forthcoming). Concerning the syntactic format of repeats in LoG, phrasal and clausal types prevail in the data. Verb-type repeats like in Finnish or Brazilian Portuguese are not grammaticalized in LoG. In 63% of the cases, repeats are used as stand-alone features (n = 40/63), while in 37% of the data, the repeat is combined with a response token (n = 23/63). 29% of these cases are full-expanded repeats (n = 18/63).^[11] Full-expanded repeats are deployed as both confirming and disconfirming responsive actions in the LoG data, always displaying some form of epistemic agency to a prior proposition. In over 71% of the cases, the repeat is a full repeat of the prior RfC (n = 45/63 (n = 27 repeat only, n = 18 repeat + token combinations)). Using overall frequency as a criterion for markedness (Enfield et al. 2019), repeats seem to be marked response types (roughly 67% response tokens vs 32% repeats). Frequency, however, can be measured in different ways: (a) as an overall frequency, or (b) as an action-based frequency, counting frequencies in a specific sequential environment. The latter approach will be the basis of the following analyses. Since the results on full-expanded repeats do not contribute any new insights, the focus of the article is on full repeats without combinations.

Full repeats in the LoG data always confirm the prior RfC action.^[12] The data feature that stand alone full repeats predominantly follow understanding checks (n = 14/27; 52%), understanding displays (n = 12/27; 44%) or allusions (Schegloff 1996; n = 1/27; 4%) in the LoG data. RfCs as topic-proffers or as ‘little questions’, however, are almost exclusively responded to with response tokens. Understanding check sequences as subordinate lines of actions are backward-oriented and prefer minimal responses. However, understanding checks do not only attract (minimal) response tokens (Section 5.2), but also full repeats in LoG interaction. In the majority of cases, these understanding checks are formatted as phrasal RfCs. Consider again Extract (5). Anna is telling about an invitation to a guided tour of a food company called Apetito. In lines 03–04, Heinz launches an understanding check in a side sequence by first deploying the open-class repair initiator wie (‘what do you mean’), followed by the candidate bi apetito (‘At apetito’s’). Anna confirms the RfC in line 05 by using a full repeat with a final falling intonation, thus closing the sequence and moving back to the main line of action in line 06. In line 07, Heinz adds the sequence-closing third hm_hm.

Extract 5: LoG-HES01_07_09 (00:23:12 – 00:23:20), No. 108

Regarding full repeats as marked options, frequency does not serve as a criterion for markedness in understanding check sequences. Minimal response tokens and repeats do not considerably differ based on the action-based frequency ratio in the LoG data (n = 16 minimal response tokens; n = 14 repeats). With regard to action ascription on the basis of sequential positioning, repeats can be analyzed as unmarked minimal ways of confirming a prior understanding check. Full repeats “seem to claim just the appropriate amount of rights for continuing the action in a way that the question warrants” (Harjunpää and Ostermann 2023, 104). This ascription as an unmarked response is also supported on a prosodic level. The repeat is not equipped with a marked focus accent shift, as discussed by Stivers (2005), but it is constructed similarly to the prior RfC design. Repeats, however, seem to compete with HG ja (‘yes’) in this environment (Section 5.2). In both cases, speakers confirm the prior RfC and then return to a course of action that was temporarily suspended. Thus, both formats are used as closure-implicative confirmations. The only difference between these response types in this context is that RfCs prior to repeat responses display a relatively flatter epistemic stance due to their formatting as phrases with falling intonation. By contrast, RfCs prior to HG ja (‘yes’) are mainly formatted as tag questions, displaying a slightly steeper epistemic gradient (Heritage and Raymond 2012, 180, and Section 5.2). Due to the differences in the epistemic displays set by the prior RfC, full repeats can be analyzed as marking epistemic agency to a slightly higher extent in this context. However, as the epistemic gradient is sequentially implicated in RfCs, the display of agency is rather minute.

In addition, repeats occur after RfCs as understanding displays (Heritage 1984), which index independent access to a matter at hand. They thus require subsequent (dis)confirmation to a lesser extent compared to understanding checks. Because of the flatter epistemic gradient, repeats following understanding displays index a higher level of epistemic agency. Consider the extended Extract (2). With the RfC in lines 06–07, Ben deploys a confident inference, which Paul responds to with a repeat in line 08, thus upgrading his own epistemic agency or primacy to the matter at hand. In Extract (5), the speaker first closes the sequence with a repeat and then comes back to the main line of action. In Extract (2), however, the repeat leverages^[13] into a continuation, allowing for a fast resumption. The repeat in line 08 thus has a dual function: (1) it works as a confirmatory response to the prior RfC (as a prosodic unit in its own) and (2) it is syntactically integrated into the following continuation, thus indexing epistemic and thematic agency. Prosodically, these repeats are always equipped with a final rising intonation. In the majority of cases, understanding displays do not occur as subordinate lines of actions in Enfield et al.’s (2019) sense. Finally, repeats in this environment compete predominantly with non-minimal response token designs or repeat + response token combinations (n = 16 non-minimal response tokens; n = 12 repeats), which supports the index of ‘non-minimality’.

Extract 2: LoG-GEL01_02 (00:00:03 – 00:00:12), No. 48

Schegloff (1996) describes full repeats as a practice for accomplishing confirming allusions. In the LoG data, there are only three instances of full repeats that can be analyzed in this way (n = 1 repeat only; n = 2 repeat + token combination as in Extract 13). Consider (26). In the preceding context, Walter has started a longer sequence, requesting his daughter Petra’s help to get wood for his farm. In line 04, Petra provides a candidate RfC, asking if the wood should be collected from the forest. This information was not mentioned explicitly in the previous talk, but there might be an inexplicit conveyance in Walter’s use of the verb inhaln (‘to collect’). Walter confirms the RfC with a full repeat in line 05, reclaiming authorship of what was stated allusively by his prior turn. Confirming allusion practices are rare in the LoG data. In addition, all of these cases would not count as ‘core phenomena’, because none of them occur with follow-up remarks making the allusion obvious.

Extract 26: LoG-BÖR03_09_10 (00:05:24 – 00:05:31), No. 42

Section 5.6 discusses different criteria of markedness and how these criteria can be applied to the action ascription of full repeats as (un-)marked response types in LoG interaction. As for the criterion of sequential positioning, the analysis indicates that full repeats sufficiently confirm previous RfCs in understanding check sequences. These subsidiary actions prefer minimal responses, which leads to a reconstruction of repeats as more unmarked responses in this context. As for the criterion of agency, the analyses reveal different degrees of agency that a speaker can index with a repeat in LoG interaction. In understanding check sequences, agency might be displayed due to the flatter epistemic gradient by the prior phrasal RfC, however, this display is rather minute compared to the index of agency with full repeats after understanding displays or as confirming allusions practices. As for the criterion of frequency, the action-based frequency of full repeats vs response tokens in understanding check sequences feature no significant difference between both response options. Thus, frequency does not serve as a criterion for ‘markedness’ in this environment. As for the semantic criterion, the research literature has discussed verb-type repeats as rather minimal ways of asserting a prior RfC. In LoG, verb-type repeats are not grammaticalized as responsive means of interaction. However, the analysis in Section 5.6 revealed that full repeats in understanding check sequences seem to be attracted by phrasal RfCs – a syntactically reduced variant of repeats compared to clausal repeats. RfCs formatted as tag questions in understanding check sequences seem to attract the minimal HG response token ja (‘yes’). Thus, the limitation of full repeats as responses to syntactically reduced phrasal RfCs leads to a categorization more on the minimal end of the minimality-maximality continuum of repeats (e.g., Hakulinen 2001). In sum, the sequential positioning of repeats in subordinate lines of actions, its prosodic design, the limitation to phrasal RfCs, and the action-based relative frequencies with minimal response tokens support the analysis of full repeats as being relatively unmarked response types in LoG.