Structural Discourse Markers in German Palliative Care Interactions

Abstract

The aim of this study is to provide a systematic account of structural discourse markers operating at a conversational macro-level in German Palliative Care interactions, focusing on their frequency, distribution, co-occurrence, and speaker-group-specific usage. By combining qualitative approaches from conversation analysis and interactional linguistics with quantitative methods from corpus linguistics, discourse markers are analyzed together as a functional category from multiple analytical perspectives to enhance the overall understanding of the use of discourse markers. The analysis reveals a functional distribution across different transition points in conversation: Whereas the German so most frequently appears in openings and transitions to non-verbal activities, gut and okay predominate in topic shifts and conversation closings. However, gut and okay differ in their composition of discursive functions, although discourse structuring emerges as the second most frequent function in both cases, an observation that stands in contrast to the continued neglect of this function in standard dictionary entries. The institutional asymmetries inherent in medical interactions are reflected in the finding that both doctors and caregivers use significantly more structural discourse markers than patients and their relatives. Differences between physicians’ and caregivers’ use of discourse markers can be attributed to their different professional roles and communicative responsibilities.

1. Introduction

Discourse Markers (henceforth: DMs) have attracted the attention of linguists for decades (e.g., Schiffrin, 1987; Jucker, 1993; Maschler, 1994; Lenk, 1998; Fraser, 1999; Günthner, 2000; Auer & Günthner, 2005; Degand & Evers-Vermeul, 2015; Auer & Maschler, 2016; Blühdorn et al., 2017a; Crible, 2018; Cuenca, 2024; Hansen & Visconti, 2024). DMs are highly frequent linguistic elements that are syntactically independent, serve a broad range of pragmatic and discourse-related functions, and predominantly occur in spoken language (for written language see Imo, 2017; Degand & van Bergen, 2018, or, within the context of a neo-standard, Auer, 2018).

While many studies focus on single DMs—such as well in English (cf. Jucker, 1993; Huang, 2019), okay across many languages (cf. Betz et al., 2021), or in German for instance also (cf. Labrenz, 2023), genau (cf. Oloff, 2016, 2017; Auer, 2021), or gut (cf. Willkop, 1988; Arens, 2023)—only a few studies have examined DMs comprehensively as a functional category on a solid quantitative basis (e.g., Crible, 2018; Crible & Degand, 2024). When DMs are compared, this is mostly done cross-linguistically (cf. Oloff, 2016; Lansari, 2020; Zufferey, 2024), for example, for nu/nå (Auer & Maschler, 2016) or for okay (Betz et al., 2021), while studies within a single language or speech community remain scarce. This builds on Betz and Sorjonen (2021, pp. 26–27) stating that there is still potential “in exploring the relation […] to other particles in a language […], further work comparing okay and gut in German could shed light on this matter”.

To address this gap, the aim of this study is to compare competing German DMs that occur in similar functional and contextual environments. German is a language particularly characterized by its extensive use of DMs. As Trotzke et al. (2020, p. 184) put it, “Discourse particles are a signature property of German”.

In order to analyze a sub-class of DMs that share similar functions, the focus is on structural DMs (cf. Maschler, 1994; Cuenca & Marin, 2009), which fulfill “structural functions, such as topic continuity, topic change, reformulation or closing” (Cuenca, 2024, p. 199). In the German-speaking landscape of terminology on DMs (for an overview, see Blühdorn et al., 2017b), the term Gliederungspartikeln (cf. Willkop, 1988) is used specifically for DMs that “structure the distinct conversation segments, mark the beginning and end, […] and organize the turn-taking in the conversation” (Duden, 2016, p. 606, author’s translation).

This study, however, focuses solely on structural DMs that operate at a macro-level of conversation, i.e., in openings, topic shifts, transitions to non-verbal activities, and conversation closings (cf. Schegloff, 1968; Schegloff & Sacks, 1973; Auer, 2020). Besides “comparing okay and gut in German” (Betz & Sorjonen, 2021, p. 27), this includes the DMs so (Meier, 2002; Barske & Golato, 2010; Pfänder et al., 2024) also (Alm, 2007; Labrenz, 2023), and genau (Oloff, 2016, 2017; Auer, 2021), among several others (see Section 2).

Discourse structuring at a micro-level, for instance, within turns, as turn-transition devices (cf. Degand & van Bergen, 2018) or re-formulations (cf. Del Saz Rubio, 2007), is not included in this analysis, since these functions are often more specific and linked to certain DMs, for instance also as a reformulation marker (cf. Deppermann & Helmer, 2013) or turn-initial ja (cf. Nilsson, 2013).

Macro-level structuring for transitions in openings, topic shifts, and closings has several advantages for the purpose of this study.

First, it is a functional category that is clear enough for quantitative annotation, but unspecific enough to be fulfilled by an array of different DMs, which allows for comparison.

Second, the different macro-level conversational transition types (openings, topic shifts, transition to non-verbal activities, closings) enable the specification of the different DMs’ particular usage contexts, thereby highlighting their nuanced functional differences and potentials.

Moreover, there are often unproved comments in the previous literature on single transitional DMs that, for instance, the German DM so accomplishes a caesura and sequence and/or topic closing, “similarly to the forms gut, okay, and bon, which are used in a comparable manner” (Meier, 2002, p. 118, author’s translation, see also Willkop, 1988; Auer, 2021). It is therefore necessary to empirically test and specify these observations.

To ensure comparability and control for contextual factors, the analysis draws on a highly homogeneous, institutional, and unique corpus of 216 interactions at a clinical Palliative Care unit in Southwestern Germany (cf., Buck, 2022; see Section 2).

In a demographically aging society, professional end-of-life care is gaining increasing importance. Over the past decades, Palliative Care has emerged as “a multidisciplinary approach seeking to improve quality of life […] for patients and families facing the problems associated with life-threatening illness” (Pease, 2019, p. 1). Interactions within the Palliative Care unit require patient-centered (cf. de Haes & Koedoot, 2003) and empathetic communication (cf. Ford, 2017), while simultaneously carrying out essential medical procedures and nursing tasks such as washing, wound care, or administering medication. Since “the consultation should not only address symptomatology, but also manage expectations, through open and compassionate communication” (Pease, 2019, p. 1), the conversation itself takes on even greater significance than in other medical contexts. Reflecting this, Gramling and Gramling (2019, p. 14) even describe conversation “as the ‘procedure’ of Palliative Care”.

This field of tension raises the question, how these conversations at a Palliative Care unit are structured and organized by the participants, especially given the institutional asymmetries in terms of linguistic repertoire (voice of medicine vs. voice of lifeworld, cf. Mishler, 1978) and the contrast between the medical staff’s knowledge and the patient’s subjective experience (cf. Heritage, 2011; Heritage & Lindström, 2015). From an applied perspective, the use of discourse-structuring elements is crucial for facilitating patient participation: “Clear markings of the beginning, closing, and transitions of distinct phases in the conversation provide guidance to patients in need of assistance, as well as the necessary situational knowledge to participate autonomously and actively in the interaction” (Lindtner-Rudolph & Bardenheuer, 2015, p. 260, author’s translation).

This makes Palliative Care interactions a particularly interesting subject for addressing the aforementioned desideratum of comparing different, yet similar structural DMs within a single language and a relatively homogeneous communicative institutional setting.

For the analysis of structural DMs on a conversational macro-level in German Palliative Care interactions, this study addresses the following research questions.

The main interest of this study is to figure out how frequent different but in principle functionally similar DMs occur in conversational openings, topic shifts, transitions to non-verbal activities, and closings and how they are distributed across these macro-level transitions in conversation (Section 3.1). To this end, all structural DMs fulfilling such transitional functions were annotated and statistically analyzed (for more detail see Section 2). In order to vividly illustrate and better understand how the DMs actually function and how they are embedded in the interactional context, selected transcript excerpts will be provided and analyzed. These excerpts stick to the principles of conversation analysis (cf. Sacks, 1992; Sidnell, 2010) and interactional linguistics (cf. Couper-Kuhlen & Selting, 2017) and follow the conventions of the Jefferson Transcription System (see Appendix A).

A descriptive heatmap table shall give insights into the possible and recurrent patterns of co-occurrence (cf. Betz & Sorjonen, 2021, p. 28) among structural DMs in conversational transitions (Section 3.2).

Moreover, group-specific differences in the usage of structural DMs are addressed (Section 3.3). On the one hand, the linguistic behavior of the clinical staff is contrasted with those of patients and their relatives. On the other hand, the Palliative Care corpus (cf. Buck, 2022) also allows for comparisons between physicians and professional caregivers in the clinic.

Last, but not least, this study wants to illustrate another analytical pathway of investigating DMs not as single forms but holistically as a functional category. This is achieved by quantifying the proportion of the macro-level structuring function within the multifunctional spectra of the two most frequent DMs in the Palliative Care corpus, namely, okay and gut. Even if it holds true that “[t]ransitional okay and gut […] do not show differences in use (closing and projecting)” (Betz & Sorjonen, 2021, p. 27, see also Meier, 2002, p. 111), it is highly interesting to compare the quantitative distribution of the different functions.

This paper is structured as follows: Section 2 introduces the data and methodology. Section 3 presents the results of the quantitative analyses. Section 4 provides a conclusion and final discussion.

2. Materials and Methods

The corpus used for this study consists of 216 interactions at a clinical Palliative Care unit in Southwestern Germany, comprising over 250,000 tokens. The data were collected between 2018 and 2019 as part of the project Kommunikation auf der Palliativstation: Pflegerisches und ärztliches Sprechen mit PalliativpatientInnen conducted by Martin Bentz, Susanne Günthner, Wolfgang Imo, and Isabella Buck (see Buck, 2022, pp. 62–67 for further details). In total, 16 caregivers, 18 physicians, and 23 patients (and in some cases, their co-present relatives) were included in the study.

Physicians and caregivers were instructed in the audio recording devices and carried out the recordings themselves. All participants provided informed consent, including agreement to the data protection policy and the use of the recordings for scientific purposes. For ethical reasons, only patients who were confirmed to have legal capacity to consent were included in the study.

The 216 recorded interactions are balanced with respect to the number of conversations primarily involving doctors (n = 107) and caregivers (n = 109). However, the number of intonation phrases (i.e., transcript lines) per conversation differs significantly, with more produced in doctor–patient interactions than in those with caregivers (median: 341 vs. 139, mean: 359 vs. 175, t-test: t = 7.5, df = 185.8, p > 0.01), as illustrated in Figure 1. This does not necessarily imply that doctors spend more time with patients, as caregivers devote a considerable portion of their time to non-verbal care activities (e.g., washing, wound care, administering medication, etc.).

This corpus of natural interactions from a German Palliative Care unit was manually annotated to enable the quantitative analysis of the research questions.

To this end, all DMs in the data fulfilling macro-level structuring functions in conversational transitions (openings, topic shifts, transitions to non-verbal activities, closings) were identified and coded. This was done iteratively by initially developing categories in a bottom-up manner in the sense of conversation analysis and interactional linguistics (cf. Couper-Kuhlen & Selting, 2017) and subsequently through a synthesis of the empirical findings in conjunction with a reconciliation with the relevant literature (i.e., Schegloff & Sacks, 1973; Meier, 2002; Harren & Raitaniemi, 2008; Couper-Kuhlen & Selting, 2017; Cuenca, 2024).

The annotation as a macro-level structural DM was based on both functional and formal criteria such as syntactic and prosodic isolation, co-occurrence with (longer) pauses, audible inbreath, other structuring DMs in the direct environment (see Section 3.3), and other linguistic practices used to accomplish conversational transitions. Additional indicators for the structural DM’s usage as a transition marker were prosodic markedness (e.g., deviations in volume, lengthening, pitch, and falling pitch contours), the logical-semantic exclusion of other possible functions (e.g., excluding gut as an evaluative marker after a sequence about dying), as well as the interactional treatment and participant-orientation within the sequential context of the respective DM.

This resulted in the following list of DMs that were included in the analyses:

Types: aber, äh(m), alles klar, also, dann, genau, gut, ja, jetzt, jo, klar, na, naja, okay, perfekt, prima, schön, so, super, tiptop, und, und (äh) ja, voila, wunderbar

However, DMs are multifunctional and thus also partially functionally ambiguous by nature. Therefore, differentiation was conducted based on whether the DMs solely fulfill a discourse-structuring function or overlap with other functions—such as reformulation (for also, cf. Deppermann & Helmer, 2013), positive evaluations (for gut, schön, super, wunderbar, etc., cf. Arens, 2023), display of understanding (for okay and gut, cf. Betz & Deppermann, 2021), agreement marking (for genau, gut, and okay, cf. Oloff, 2017, 2019), or question tags (okay with rising intonation, cf. König, 2017).

In reality, however, this distinction is often not binary, but rather fluid and continuous. On the German DM gut, for instance, Arens (2023, p. 266, author’s translation) states that “evaluating and structuring aspects are closely and inseparably connected in a complex interplay”.

Nevertheless, the distinction of discourse-structuring gut only, ambiguous gut with overlap in positive evaluation, and evaluative gut without any macro-level structuring is illustrated in the following two excerpts from the Palliative Care data in order to give an impression of what is meant by the annotation categories ‘purely discourse structuring function’ in contrast to ‘ambiguous’ and ‘other’ (in this case ‘positive evaluation’, only relevant for Section 3.4).

The transcript excerpts follow the standards of conversation analysis (cf. Sidnell, 2010) and interactional linguistics (cf. Couper-Kuhlen & Selting, 2017) and are presented in the Jefferson Transcription System (see Appendix A). The original German transcript is provided alongside a rough English translation in italics below. Relevant grammatical categories are added in brackets and capital letters, e.g., [DIMINUTIVE].

The first excerpt (Excerpt 1) begins after the physician and patient have praised the patient’s mother and are talking about her reaction to a patient’s decision to move to a hospice, which is commented on by the patient with “she has come to terms with it” (line 01).

Excerpt 1. Gut as a positive evaluation.
01	PATf	jetzt hat sie sich auch schon damit angefreu[ndet;        ]
		now she has come to terms with it
02	DOCf	[<<p> ja:;>]
		yes
03	PATf	(aber) ich bin ganz zufrieden mit der entscheid[ung;]
		(but) I am quite happy with the decision
04	DOCf	[ja;    ]
		yes
05		<<pp> sehr gut;>
		very good [DM]
06		(0.3)
07		ja ich glaub ihre mutter is da einfach auch bei ihnen so ne?
		yeah I guess your mother is just like that with you too, right [QUESTION TAG]
08	PATf	genau.
		exactly
09		und am ende wird sie merken,
		and in the end she will realize
10		<<smile voice> dass es so viel besser isch;>
		that it is [DIALECTAL] much better this way
11	DOCf	dass das besser is;
		that this is better

When the patient expresses being “quite happy with the decision” (line 03), the physician responds with a quiet sehr gut ‘very good’ as a positive assessment (line 05). It is typical for such evaluations that they come along with an accented intensifier (cf. Arens, 2023, pp. 233, 252) and a follow-up request for (re-)confirmation (line 07), which is given by the patient with genau (‘exactly’, line 08), followed by the optimistic projection (cf. Jefferson, 1984) that “in the end she will realize that it is much better this way” (lines 09–10). This is partly repeated by the physician (line 11) and the sequence is closed. This means that in Excerpt 1 (sehr) gut is used as an evaluative DM only, well embedded in the sequence and not involved in macro-level discourse-structuring.

This is different in Excerpt 2. The conversation is about to end after the physician and the patient have discussed how the transfer to the hospice will take place. After a two-second pause, the patient projects a possible pre-closing by indicating elliptically that there is nothing urgent to talk about anymore (line 01).

Excerpt 2. Ambiguous gut between positive evaluation and discourse-structuring.
01	PATf	ansonschten wüsst ich jetzt nich;
		other than that [DIALECTAL] I would not know
02	DOCf	<<h> gu:t;>
		well [DM]
03	PATf	fühl ich mich in super händen;
		I feel like I’m in excellent hands
04		nach wie [vor;]
		just as much as ever
05	DOCf	[ja;   ]
		yes
06		<<creaky voice> [sehr gut;>]
		very good [DM]
07	PATf	[ganz toll; ]
		really good
08	DOCf	.h ansonsten morgen hab ich ja auch dienst,
		other than that I’m [PARTICLE] working tomorrow as well
09		da wär ich ja auch ansprechbar?
		I’d be [PARTICLE] approachable then as well

The physician interprets the unfinished sentence (line 01) as an indication that there is nothing left to discuss, and thus that the conversation can be closed. The subsequent gut in line 02 clearly serves only as a transition into the pre-closing phase of the conversation. There is no preceding evaluandum to which it could refer in the case of an assessment, its bleached evaluative semantics (cf. Arens, 2023, p. 266) are not applicable in this context (why should it be positive that the patient “would not know”?, line 01). Moreover, it does not elicit a second assessment, which would normally be the case for assessments (cf. Couper-Kuhlen & Selting, 2017, p. 294), and is realized prosodically marked with high pitch and lengthening.

However, the patient reconsiders that there is nothing more to say and gives the compliment that she still feels like she was “in excellent hands” (lines 03–04). The now in line 06 following sehr gut (as in Excerpt 1 preceded by an overlapping ja, line 05) is ambiguous: On the hand one it is a positive, evaluative response to the fact that the patient feels comfortable at the clinic. In overlap, the patient even reformulates her compliment with a second assessment (ganz toll ‘really good’, line 07). On the other hand, it closes the sequence and opens the floor for a topic shift or the closing of the conversation. This becomes apparent in the physician’s subsequent announcement (lines 08–09) that she would be “approachable” (line 09) the next day, too, implicating that any further (or upcoming) issues could be discussed then and hence the conversation can be closed for now.

Other functions besides macro-level transitions were only annotated for the two most frequent DMs gut and okay in order to compare the role of discourse-structuring within their multifunctional spectra (see Section 3.4). For all other DMs, the annotation included just the categories of ‘discourse-structuring function only’ or ‘ambiguous’, which could be any other function but transitioning.

Besides the functional categories of the DMs, which have been illustrated in Excerpt 1 and 2, several metadata variables were also annotated and are indicated in the transcript excerpts, namely, the speaker’s institutional role (doctor = DOC, caregiver = CG, patient = PAT, relative of the patient = REL), gender (f/m), and an individual identification number. Phonetic and non-standard variants, such as dialectal, were normalized during annotation (e.g., subber, supi, etc. → super). Question tags were excluded, although they may in some cases serve a structuring function (cf. König, 2017).

Statistical analyses (descriptive statistics, chi-square tests) and data visualizations were carried out using the software R and R Studio including the packages ggplot2 and dplyr.

3. Results

3.1. The Distribution of Concurring DMs in Similar Contextual Environments

In order to answer the main research question concerning the distribution and comparison of different structural DMs in German interactions at a Palliative Care unit, the identified transition points in conversation (i.e., openings, topic shifts, transition to non-verbal activities, and closings) are analyzed separately, beginning with opening sequences.

3.1.1. Openings

Openings have been extensively described in conversation analytic literature (cf. Schegloff, 1967, 1968; Heath, 1981; Lindström, 1994; Meier, 2002; Auer, 2020). The participants of the conversation need co-presence and joint attention to start a focused interaction. Structural DMs can assist in establishing this shared focus (cf. Meier, 2002, p. 66).

Figure 2 presents the distribution of all DMs occurring in opening sequences within the Palliative Care corpus involving both physicians and caregivers, as well as the patients and their relatives.

The darker parts of the bars (above) in Figure 2 indicate clear cases of discourse-structuring, whereas the lighter parts of the bars (below) include ambiguous instances that overlap with other functions (see Section 2) such as positive evaluation for gut (cf. Arens, 2023; see Excerpts 1 & 2) or display of understanding for okay (cf. Betz & Deppermann, 2021). This color scheme applies to all upcoming figures as well.

In the majority of the 216 interactions in the corpus, the opening takes place without any DMs. However, if the opening sequence includes a DM, the German so (for a comparison with English so see Barske & Golato, 2010, pp. 247–248) is by far most frequently used in this position of conversation (n = 28). This is in line with the previous literature on German so marking “the beginning of the actual activity” (Schwitalla, 2002, p. 246, author’s translation) due to its “focusing quality” (Meier, 2002, p. 74, author’s translation) and is already documented for opening sequences in classroom interactions (cf. Mazeland, 1984), business meetings, and in the courtroom (cf. Meier, 2002; see also Barske & Golato, 2010, p. 246).

The following Excerpts 3 and 4 show examples of so in the very beginning of the conversation (Excerpt 3) and after the greeting (cf. Schegloff, 1968) in order to get to the first topic of the conversation (Excerpt 4).

Excerpt 3. So in an opening at the very beginning of the conversation.
01	CGm	’`so,
		so [DM]
02		was macht die kunst,
		how are you doing
03		(0.3)
04	PATm	alles gut;
		everything fine
05	CGm	alles gut.
		everything fine
06		(0.4)
07		.h sind sie bettfein,
		are you ready for bed

In Excerpt 3, the caregiver initiates the conversation in line 01 with a falling-rising intonated so with open vowel [zoː]. Different phonetic realizations, e.g., [zɔʔ] or [zɔʊ̯], were included in the analysis, although they might (and probably will) have nuanced functional potentials. The caregiver follows with an open how-are-you-question (line 02) and acknowledges the response “everything’s good” (line 04) by repeating it in line 05 (cf. Rossi, 2020). After a brief pause and an inbreath, the conversation shifts toward the actual agenda, as the caregiver asks whether the patient is “ready for bed” (line 07).

Excerpt 4. So after the greeting and before the first topic of talk.
01	PATf	hallo?
		hello
02		(0.4)
03	CGm	hellaule;
		hello [DIALECTAL] [DIMINUTIVE]
04		(3.8)
05		.h so;
		so [DM]
06		ich bring noch ä bissl
		I bring a bit [DIALECTAL] of
07		(0.3)
08	PATf	<<pp> hm_hm;>
		uh_uh
09	CGm	infusiö::nchen,
		infusion [DIMINUTIVE]

In contrast, in Excerpt 4 the DM so appears following the greeting with a dialectal diminutive (line 03), a longer pause (line 04) and an inbreath (line 05) and preceding the caregiver’s announcement that he will “bring a bit of infusion” (lines 06–09).

However, other DMs like na, okay, also, genau, and gut also occur in openings, though less frequently than so (Figure 2). This fits the description in the previous literature of gut having a “weak opening function” (Willkop, 1988, p. 152, author’s translation).

In the Palliative Care data na is produced in openings only and serves as a “contact signal” (Schwitalla, 2002, p. 262, author’s translation; see also Willkop, 1988; Rensch, 2018). Excerpt 5 illustrates how na (n = 5) functions within the opening sequence.

Excerpt 5. The use of na to open the floor for the first topic of the conversation.
01	CGf	hallöchen;
		hello [DIMINUTIVE]
02		(0.4)
03	PATf	hallöchen;
		hello [DIMINUTIVE]
04	CGf	n:a,
		[DM]
05		(1.1)
06		wie war der nachmittag;
		how was the afternoon
07		(0.3)
08	PATf	och:;
		[INTERJECTION]
09		(0.6)
10		furchtbar viel los;
		terribly much going on
11	CGf	furchtbar viel los;
		terribly much going on

Excerpt 5 shows a reciprocal greeting sequence (cf. Auer, 2020) with a diminutive form (lines 01–02) followed by the DM na (line 04) produced by the caregiver, and a 1.1-second pause (line 05). The lengthening and rising intonation of na already projects interest in the interlocutor (cf. Schwitalla, 2002), comparable to the connective und (‘and’), which is included under the category “other DMs” in Figure 2. The patient’s response (lines 08-10) to the subsequent how-are-you question in line 06 (“how was the afternoon”) is again repeated by the caregiver in line 11 (cf. Rossi, 2020).

With regard to co-occurrences of DMs within opening sequences, the observation of “the frequent occurrence of temporal-deictic elements […] following so” (Meier, 2002, p. 77, author’s translation, see also Barske & Golato, 2010, p. 256) can be confirmed in the Palliative Care interactions. However, this applies more frequently to jetzt (‘now’) rather than to dann (‘then’), as illustrated in Excerpt 6. The addition of jetzt (‘now’) often follows problems or interruptions that needed to be resolved before the interaction could properly begin. Similarly to so, jetzt marks “the beginning or end (or both) of conversational segments […] [and] in such uses, it can serve as an attention-guiding element” (Staffeldt, 2018, p. 282, author’s translation).

Excerpt 6. Co-occurrence of so and jetzt in openings.
01	CGm	so: jetzat;
		so [DM] now [DIALECTAL]
02		(.)
03	PATf	hm_[hm?]
		uh_uh
04	CGm	[   m:]erke sie scho d_besserung von de schmerze,
		have you already [DIALECTAL] noticed improvement in the pain [DIALECTAL]

After problems have occurred beforehand, the caregiver begins the conversation with the lengthened DM so combined with a dialectal realization of jetzt (‘now’, line 01), which is operating between temporal and conversational deixis (cf. Imo, 2010). In this instance, both are produced within a single intonation phrase. However, they can also be realized in two separate intonation phrases. The patient responds minimally (line 03), signaling his attention and willingness to start the conversation. In overlap, the caregiver inquires whether there has already been some “improvement in the pain” (line 04) and moves on to the first topic of the conversation.

3.1.2. Transitions to Non-Verbal Activities

In the transition to non-verbal activities, so is, like in openings, by far the dominant DM (n = 64) for both introducing and closing the non-verbal activity, followed by gut, okay, also, and other DMs. Again, this is in line with the discourse-structuring function of so to end an action and “signal readiness to move on to a new, next task” (Meier, 2002, p. 74, author’s translation). Similarly, Barske and Golato (2010, p. 254) state that so “seems to indicate that one interactional business has been completed and the speaker is ready to continue with a next action.”

Pfänder et al. (2024, p. 3) analyze the German DM so as “transition marker” and cite Brünner (2005, p. 104) describing so in the context of corporate instructions as the “by far most frequently used resource […] [to indicate] segmentation of activities”.

Figure 3 depicts the proportion of the different DMs in transitions to non-verbal activities, such as wound care, washing, or a medical investigation, as in the subsequent Excerpt 7 with the doctor listening to the patient’s belly.

Excerpt 7. Transition to a non-verbal activity with the DM so.
01	DOCf	.h so;
		so [DM]
02		ich will einmal noch kurz auf den bauch draufhören;
		I just want to listen to the belly once more
03		(27.4)
04		so;
		so [DM]
05		(0.3)
06		.h also darmgeräusche sind gut da,
		so [DM] bowel sounds are there
07	PATm	<<p> ja;>
		yes

In Excerpt 7, the doctor uses the German DM so (line 01, similar to Excerpt 2, following an inbreath) to get the patient’s attention (cf. Schwitalla, 2002) for his announcement of the upcoming non-verbal task (“I just want to listen to the belly once more”, line 02). Without a response from the patient, the doctor proceeds with the examination, which lasts almost half a minute (line 03). When the non-verbal activity is completed, the doctor produces another falling so (line 04) to announce the results of the examination (“bowel sounds are there”, line 06) beginning with a short pause (line 05), an inbreath, and the conclusive DM also (‘so’, cf. Deppermann & Helmer, 2013; Labrenz, 2023).

In transitions to non-verbal activities, as in openings, a co-occurrence with temporal deictic expressions in turn-initial position can be observed (cf. Meier, 2002; Barske & Golato, 2010; Pfänder et al., 2024) as the following Excerpts 8 (so + jetzt) and 9 (gut + dann) demonstrate. In both excerpts, the caregiver uses dialectal variants of the 1st person plural pronoun, which has been shown to be a typical pronoun use in German medical interactions (cf. Günthner, 2021), to transition to the non-verbal joint activity (cf. Pfänder et al., 2024), namely, getting the patient in a standing position (Excerpt 8) and taking off the patient’s shoes (Excerpt 9).

Excerpt 8. So + jetzt in a transition to a non-verbal activity.
01	CGf	’`so:;
		so [DM]
02		(0.5)
03		jetzt gucke_mer mal dass sie in de stand komme?
		now let’s look [DIALECTAL] that you come [DIALECTAL] in a standing position
Excerpt 9. Gut + dann in a transition to a non-verbal activity.
01	CGf	<<p> gut dann müss_mer (.) kurz (.) die schuh ausziehe,>
		well [DM] then we have to [DIALECTAL] quickly take off [DIALECTAL] the shoes

3.1.3. Topic Shift

Structural DMs are used not only for transition to non-verbal actions (Section 3.1.2), but also in transitions from one topic to another. The distinction between the more content-oriented notion of ‘topic’ and the more interactionally organizational notion of ‘sequence’ has been the subject of extensive discussion within Interactional Linguistics (see for an overview: Couper-Kuhlen & Selting, 2017, pp. 312–354). However, the authors state that “managing topicality in conversation is always a collaborative, interactive phenomenon” (Couper-Kuhlen & Selting, 2017, p. 315). In the context of structural DMs, both terms are used given the DMs’ transitional function “as a device to close down a prior topic and/or sequence while simultaneously signaling the beginning of [a new one]” (Barske & Golato, 2010, p. 246). Each topic shift also includes a closure of the previous sequence, but obviously not vice versa.

Unlike in more informal interactions, where a more gradual form of “topic shading” (Schegloff & Sacks, 1973, p. 305) is often observed, institutional settings such as those in the Palliative Care unit tend to exhibit clearer and more marked topic boundaries, or, as Betz and Sorjonen (2021, p. 20) put it, “where turn-taking organization is constrained, ‘topic’ may be a stronger and more important source of coherence than in everyday interaction, and topical organization become more clearly traceable”.

Of course, topic shifts can occur without the use of DMs. In some cases, there are even explicit formulations on the topic structure or underlying ‘agenda’ the doctor is following, as illustrated in Excerpt 10.

Excerpt 10. Topic shift without DMs.
01	DOCf	s_zweide das ist die patientenverfügung,
		the second [DIALECTAL] this is the living will

After the initial topic of the conversation appears to have been sufficiently addressed, the doctor introduces “the second” (line 01) one, which is picked up anaphorically and elaborated upon: “this is the living will” (line 01).

If, however, structural DMs are used in topic shifts, they often introduce “a next activity or discourse segment and close at the same time the preceding one” (Auer, 2021, p. 18) and are thus both prospective and retrospective (see also Pfänder et al., 2024). For this type of transition, the German DM so may also be used, though less frequently than okay, gut, genau, and also, as illustrated in Figure 4.

In topic shifts, a considerable proportion of overlapping functional uses can be observed, as indicated by the lighter-colored section in the bar chart below (Figure 4).

Excerpt 11 demonstrates a typical topic shift with the DMs okay, so, and gut.

Excerpt 11. Topic shift with okay, so, and gut.
01	DOCf	des werden se sehen_
		you yill see this
02		=also (0.3) .h die kümmern sich da;
		so [DM] they are taking care of you
03		.h und wenn die darmgeräusche da sind isses das alles nicht so schlimm;
		and if the sounds of the intestine are there is it that everything not so terrible
04	PATf	okay;
		okay [DM]
05	DOCf	.h äh:m
		uhm
06		(0.7)
07		so;
		so [DM]
08		gut;
		well [DM]
09		also von den schmerzen hab ich das gefühl wir sind aufm richtigen [weg?      ]
		so [DM] with the pain I have got the feeling that we are on the right track
10	PATf	[hm_hm,]
		mhm
11		hm_hm,
		mhm
12		(0.6)
13	DOCf	und äh:m_
		and uhm
14		(0.4)
15		was als nächsten schritt dann folgen (.) würde
		what would follow then as a next step

The doctor closes the sequence (lines 01-03) about the upcoming intervention on the patient’s intestine with optimistic projections (cf. Jefferson, 1984), suggesting to the patient that “everything [is] not so terrible” (line 03). Interestingly, the preceding also in line 02 functions as a reformulation marker (cf. Del Saz Rubio, 2007), but at the same time, due to its conclusive nature (cf. Deppermann & Helmer, 2013), it already projects the potential sequence closure.

The slightly falling intonated and emphasized okay by the patient (line 04) displays understanding (cf. Betz & Deppermann, 2021) and simultaneously indicates that there are apparently no further questions from the patient on that matter, suggesting that the topic could potentially be closed (cf. Mondada & Sorjonen, 2021). Therefore, it was annotated as functional ambiguous (see Section 2).

Due to the doctor’s long inbreath, the filled and unfilled pause are not being treated by the patient as an opportunity to re-enter and expand the sequence (lines 05–06), the doctor is thus able to close the topic of the patient’s intestine and move on to the next topic, that of pain management (line 09). This is accomplished by the DMs so and gut, both with falling intonation (lines 07–08). The use of also serves to shift the focus on the previously discussed topic of pain, while simultaneously initiating a reassurance sequence (line 09), which is acknowledged (cf. Jefferson, 1984) by the patient in the subsequent turns (lines 10–11). As a result, the topic does not require further elaboration, and the doctor can proceed to the “next step” (line 15).

In Excerpt 12, discourse structuring uses of genau and prima are analyzed. Prima thereby represents a broader category of positive evaluation expressions beyond gut (cf. Arens, 2023), which can also be used as a DM (i.e., super ‘superb’, wunderbar ‘wonderful’, perfekt ‘perfect’). A female caregiver is talking with the wife of a patient, whose medication has been changed from dropwise delivery to an intravenous bolus injection. At the beginning of the excerpt, they are discussing the consequences of this change including the possibilities of subsequent adjustments such as to “increase the flow rate” (line 02) of the injection.

Excerpt 12. Topic shift with prima and genau.
01	CGf	dann darf man auch immer (0.8) auf den bolus drücken,
		then one may always press on the bolus
02		oder halt in der laufrate hochgehen;
		or [MODAL PARTICLE] increase the flow rate
03		(.)
04	RELf	<<p> hm_hm;>
		mhm
05		(1.2)
06		<<pp> prima;>
		excellent [DM]
07		(2.1)
08	CGf	genau;
		exactly [DM]
09		(1.5)
10		jetzt wollt ich grad schauen <<acc> wie man die am besten festmacht,>
		now I wanted to just look how to best secure/fasten this

The caregiver’s reflection on possible adjustments (lines 01–02) are commented on by the patient’s wife with a display of understanding (line 04, on “hm_hm” see Jefferson, 1984). The subsequent prima (line 06) may be interpreted not only as a positive evaluation of the entire sequence but also, and perhaps more importantly, as a general signal of agreement marking the closure of the previous topic (cf. Arens, 2023). Further evidence supporting this interpretation is provided by the sequential positioning of prima as well as the prolonged pauses preceding and following it (lines 05–07). Particularly noteworthy is the prosodic distinctiveness of this utterance, specifically its notably low volume, a feature that can generally be observed as a common tendency among structural DMs within the Palliative Care corpus.

Although the following genau (line 08) could be seen as an expression of agreement with the topic closure indicated by prima, it is more plausibly interpreted as a marker of topic transition. In this respect, the refocusing function of genau is important, emphasized by its normal volume and the following utterance containing jetzt as an element between temporal and conversational deixis (cf. Imo, 2010). This use of genau has been described as “self-responsive” (Willkop, 1988, p. 145, author’s translation) or “auto-dialogic” (cf. Auer, 2021), meaning that “for the other participants, the reason for this confirmation is no longer accessible. The confirmable lies outside the current interactional space” (Auer, 2021, p. 14, author’s translation).

After the transition, the caregiver accelerates her speaking speed when starting the next topic of “how to best secure/fasten” (line 10) the device.

3.1.4. Closing of the Conversation

The distribution of DMs in closing sequences (cf. Schegloff & Sacks, 1973; Harren & Raitaniemi, 2008; Sidnell, 2010; Auer, 2020) is quite similar to that observed in topic shifts (Figure 4), both in absolute numbers and in relative proportions (Figure 5).

This highlights the relevance of DMs in closings, especially considering that a conversation may contain several topic shifts, but typically only one closing sequence, which can, however, include an entire cluster of structural DMs (cf. Cuenca, 2024).

Like in topic shifts (Section 3.1.3), okay and gut are the most frequently used DMs in closing sequences. However, a tendency towards even more tokens of gut in closings can be observed. While genau is significantly more frequently used in topic shifts, also, alles klar, and dann are more common in closings, which might be explained by the conclusive semantics of also and dann (cf. Deppermann & Helmer, 2013). This is in line with previous research listing for closings in this order gut, okay, ja, alles klar, also, and dann as “an offer to come to an end in the conversation” (Schwitalla, 2002, pp. 262–263).

A prototypical closing sequence featuring a cluster of different DMs is illustrated in Excerpt 13. In the preceding interaction, the physician and the patient had discussed the patient’s discharge from the hospital.

Excerpt 13. Closing sequence with a cluster of structural DMs.
01	DOCf	<<smile voice> gä,>
		[DIALECTAL QUESTION TAG]
02		sie erfahren_s als [erste wenn_s heimgeht   ] nach uns;
		we will let you know at first after us when it’s going home
03	PATf	[ja <<laughing> hehehehe;>]
		yes ((laughing))
04		und dann,
		and then
05	DOCf	gä,
		[DIALECTAL QUESTION TAG]
06		[.h ja?      ]
		yes [QUESTION TAG]
07	PATf	[geh_mer] heim [na;    ]
		we go home [PARTICLE]
08	DOCf	[okay;]
		okay [DM]
09		prima;
		excellent [DM]
10	PATf	hm_hm,
		mhm
11	DOCf	alles klar,
		everything clear [DM]
12		gu[t,]
		fine [DM]
13	PATf	[o][kay;    ]
		okay [DM]
14	DOCf	[dann,]
		then [DM]
15		(.)
16		bis morgen;
		see you tomorrow
17		[gä,]
		[DIALECTAL QUESTION TAG]
18	PATf	[ja;]
		yes
19		dan[ke;  ]
		thanks
20	DOCf	[tsch][ü:ss;]
		goodbye
21	PATf	[tschü]ss;
		goodbye

The doctor assures the patient that she will be informed immediately in the event that discharge is possible (line 02), accompanied by several question tags that seek affiliation and understanding (lines 01, 05–06). This is confirmed by the patient (line 03), and in line 07 the exchange is concluded with an optimistic projection (cf. Jefferson, 1984). Several DMs are produced by the doctor (okay, prima, alles klar, gut, lines 08–12) as well as by the patient (agreement marker mhm, line 10; okay, line 13). Those DMs function as pre-closing devices (cf. Schegloff & Sacks, 1973) and constitute the transition to the actual closing of the conversation, initiated by dann (line 14) and bis morgen (‘see you tomorrow’, line 16). In line 17, the doctor produces another question tag. In partial overlap with the patient’s danke (‘thank you’, line 19), the doctor initiates the final, mirrored goodbye-adjacency pair.

The order in which these DMs appear could provide insights into an implicit hierarchy among them. Therefore, analyzing the most frequent co-occurrences of these DMs is highly relevant and will be performed in the next section (Section 3.2).

3.2. Co-Occurrences

The co-occurrence of different DMs remains an under-researched area within the field of DM research (but see Cuenca & Marin, 2009; Sorjonen & Vepsäläinen, 2016; Pons Bordería, 2018; Cuenca, 2024). However, it is often stated that research on DMs “needs to consider which co-occurrences […] are possible and recurrent” (Betz & Sorjonen, 2021, p. 28).

To address this, the co-occurrences of the five most frequent structural DMs (okay, gut, so, genau, also) within the corpus of German Palliative Care interactions were counted and visualized in a heatmap table (Figure 6).

Two DMs were considered to co-occur when they were uttered adjacently, either within the same or across two consecutive intonation phrases. Pauses and hesitation markers in between were ignored.

The table in Figure 6 is to be read from left to right, quantifying the occurrences of DMs listed on the left in relation to those listed at the top. The colors indicate the frequency of the co-occurrences as illustrated in the legend (white <1%, light grey 1–5%, light green 5–10%, dark green >10%).

On the right side of the heatmap table, the co-occurrences with temporal deictic elements, namely, dann (‘then’) and jetzt (‘now’), are also listed.

Several conclusions can be drawn from Figure 6. First, the majority of the DMs can, at least in principle, co-occur with one another, as evidenced by the relatively few empty cells in the table (for example, the combinations *also + okay or *so + so are not found in a discourse-structuring function in the data). This shows that for structural DMs in German many “co-occurrences […] are possible.” (Betz & Sorjonen, 2021, p. 28).

Second, there are striking differences in the frequency of certain co-occurrences as indicated by the different colors in the heatmap of Figure 6. For instance, more than 12% of all uses of okay as a structural DM on a conversational macro-level are followed by gut (n = 37). This might indicate an implicit hierarchy of the DMs, since, in the other direction, gut + okay is a less frequent combination (n = 9; 4% of all tokens of gut). This observation is in line with the findings of Arens (2023, p. 153) with more than double the amount of okay + gut compared to gut + okay.

Also is barely followed by other structural DMs, if so, however, almost only by gut (n = 8, 6.3% of all instances of also as a macro-level DM). In contrast, also is able to follow other transitional DMs, as the co-occurrences of genau + also, okay + also, and gut + also occur quite frequently in the data (around 7% of all instances for each DM).

So seems to co-occur less frequently with other DMs, if so, similar to also, so + gut is the most likely combination (n = 9; 6.2%, see Excerpt 11; for so + also see Excerpt 7)

As mentioned beforehand, combinations of DMs with temporal-deictic elements such as dann (‘then’) and jetzt (‘now’) are frequently observed across all the different DMs (in line with Meier, 2002; Barske & Golato, 2010). Particularly, in combination with dann, many DMs such as also, super, gut, or okay reach high co-occurrence percentages between 10 and 15% (see Excerpts 9 and 13). Although so is also co-occurring with dann (n = 10, 6.8%), it seems to be more closely connected with jetzt (n = 20; 13.7%; see Excerpts 6 and 8), which is barely the case for all other DMs.

Although combinations between two or more different DMs are more frequent, the doubling of a single DM is also quite interesting, as it can indicate a heightened necessity or willingness to close the sequence, representing a functional deviation compared to the combination of two different DMs (cf. Barth-Weingarten, 2011). However, this does not seem to work for *also + also and *so + so, which might give us an insight into the difference of how certain DMs function.

A case of DM doubling becomes apparent in Excerpt 14 (okay + okay, n = 5). Prior to this sequence of interaction, the patient had reported having a bowel movement again, which the doctor attributed to the laxative effect of a new medication the patient had received the previous day. When the patient asks whether the new medication was crucial in the end, the doctor responds, “that’s what I assume” (line 01), suggesting that, from the perspective of the “voice of medicine” (cf. Mishler, 1978), this knowledge is probabilistic rather than one hundred percent certain.

Excerpt 14. Co-occurrence of the same DM (repetition): okay + okay.
01	DOCf	davon geh ich aus;
		that’s what I assume
02		(.)
03		ja?
		yes [QUESTION TAG]
04	PATm	ok[ay;]
05	DOCf	[als]o beweisen kann ich das auch nich_
		well I cannot really prove that
06		ab[er_]
		but
07	PATm	[ ja]jaja;
		yes yes yes
08		[jaja;               ]
09	DOCf	[das sagt halt] die erfahrung;
		this is what the experience tells us
10	PATm	o[kay] okay
11	DOCf	[ja? ]
		yes [QUESTION TAG]
12		und (.) n_das der nächste punkt ist,
		and (.) uh the the next point is

With the question tag ja realized with rising intonation (line 03), the doctor makes a confirmation relevant, which is then provided by the patient with a falling okay (cf. Betz et al., 2021). However, this is interpreted by the doctor as a request for more information (Betz & Deppermann, 2021), since she initiates an expansion with the focus particle also to state explicitly that she cannot “really prove” (line 05) the causality suggested by the patient. In overlap, the patient responds with repeated ja, signaling that the information provided was already sufficient for understanding (cf. Barth-Weingarten, 2011) and indicating that the expanded sequence can now be closed (lines 07–08). The doctor adds that this knowledge is, at least, based on professional experience (line 09), which is again brushed off by the patient through a repeated okay as a display of (over-)understanding (line 10). The doctor closes the sequence with another question tag (line 11), which can serve a similar transitional function to structural DMs (cf. König, 2017), although not included in this study. The transition to the next conversational topic is initiated by the physician through a turn-initial und, also commonly used in transitions at a macro-level (see Section 3.1.1).

3.3. Distribution of Structural DMs Across Participation Groups

Finally, the research question of how DM usage differs across different interactive roles or participation groups. The Palliative Care corpus enables not only a comparison between medical professionals and patients, but also between physicians and caregivers, a relevant group frequently overlooked in studies on medical communication (see for qualitative differences in several communicative practices, (Buck, 2022)).

As the Figure 7 and Figure 8 visualize, the institutional asymmetries of medical staff (visualized in blue) and patients as well as their relatives (visualized in red) are also reflected in the distribution of structural DMs, which are produced significantly more frequently by doctors and caregivers than by patients and their relatives (χ² [df = 1, n = 24,196] = 323.7, p < 0.01). Importantly, this is not only true in absolute number of tokens but also normalized in relation to the overall amount of intonation phrases.

Figure 7 presents a mosaic plot that visualizes the overall use of DMs across different participation groups (physicians, caregivers, patients, and their relatives). The relations of DMs per group are normalized by the use of intonation phrases.

Figure 8 zooms in on the distribution of single DMs in relation to the interactive roles. Significant differences are highlighted by small plus symbols in the corresponding color or bar. The red plus symbols refer to significant differences in patients and relatives versus medical staff, whereas the blue plus symbols indicate differences between physicians and caregivers.

Figure 7 visualizes the finding that medical professionals use significantly more structural DMs than patients or their relatives (normalized per intonation phrase). This finding is not surprising given that the medical staff is responsible for conducting the conversation overall, and particularly for managing the delicate balance between leading the conversation with empathy while maintaining a clear focus on medical objectives (cf. Ford, 2017; Lindtner-Rudolph & Bardenheuer, 2015).

In addition, more ambiguous instances of structural DMs are observed in patients and their relatives in comparison to the clinical staff (χ² [df = 1, n = 1150] = 68.4, p < 0.01.). This can be attributed to the fact that they use DMs more frequently in a responding position, where other functional uses (and hence more overlap) such as positive evaluation, agreement, or display of understanding, are also expectable. Although all single DMs appear across the various participation groups, and the distribution is relatively balanced (Figure 8), it is remarkable that patients, especially, use DMs with lexical pendants (cf. Arens, 2023) that convey positive evaluation (super, schön, prima, etc.) and display of understanding (alles klar) significantly more often than it would be expectable if all DMs were distributed equally. Vice versa, the transition marker so (cf. Barske & Golato, 2010; Pfänder et al., 2024) is produced less than expected in patients and their relatives in comparison to their medical counterparts.

Within the healthcare personnel, however, differences can also be observed in terms of physicians producing overall significantly more DMs than caregivers (χ² [df = 1, n = 14,173] = 6.5, p < 0.05). An explanation for this could lie in the differing professional roles and requirements. Patients interact more frequently with caregivers. During certain care routines, there may be more space for informal conversation, which tends to lead to more “topic shading” (Schegloff & Sacks, 1973, p. 305). In these interactions with caregivers, conversation itself can be seen “as the ‘procedure’ of Palliative Care” (Gramling & Gramling, 2019, p. 14). In contrast, physicians generally might have a more structured agenda with a greater number of topics to cover, requiring a more focused and structured conversation. In other words, “‘topic’ may be a stronger and more important source of coherence” (Betz & Sorjonen, 2021, p. 20). Hence, more DMs might be needed in physicians’ interaction to organize and structure this.

Since okay emerged as the predominant DM in topic shifts (Section 3.1.3), it is not surprising that physicians use okay significantly more frequently than caregivers (χ² [df = 1, n = 926] = 31.2, p < 0.01). Reversely, the caregivers produce more tokens of so than the doctors (χ² [df = 1, n = 919] = 67.4, p < 0.01). As shown in Section 3.1.2, so is the main DM in transitions to non-verbal activities, which the professional caregivers perform more frequently (when carrying out actual caring tasks) compared to physicians due to the differing requirements of their professional roles.

3.4. Proportion of the Use as a Structural DM Within the Multifunctional Spectrum

The aim of the present study was to provide a comprehensive account of structural DMs as a group, in order to identify, for example, similarities and differences between markers such as gut and okay. After analyzing the frequencies of the individual DMs, their distinct distributions across various macro-level transitions within the conversation (Section 3.1), their co-occurrence patterns (Section 3.2), as well as their usage depending on institutional roles (Section 3.3), this section outlines an additional pathway that contributes to a more holistic analysis of DMs as a group. This is achieved by determining the proportion of the macro-level structuring function within the DMs’ individual multifunctional spectrum.

In contrast to the preceding results, the macro-level discourse-structuring functions of DMs in spoken language are still underestimated in traditional descriptions by German (and other) dictionaries: “Standard dictionaries however do not report on possible closing or structuring functions as they occur in spoken interaction” (Oloff, 2019, p. 201).

To this end, all instances of the two most frequent DMs were identified and annotated with regard to their function in conversation (see Section 2), namely, okay (‘okay’, cf. Beach, 1993; Metcalf, 2010; Nilsson, 2013; Oloff, 2019; Betz et al., 2021; Betz & Deppermann, 2021; Betz & Sorjonen, 2021; Mondada & Sorjonen, 2021) and gut (‘good’/’well’, cf. Willkop, 1988; Meier, 2002; Schwitalla, 2002; Harren & Raitaniemi, 2008; Auer, 2021; Arens, 2023).

Figure 9 illustrates the proportions using stacked bar plots with the dark color tier representing instances of uses as a structural DM, while the lighter colors in the middle indicate ambiguous cases, where overlap with other interactional functions occurs (see legend). The light gray bars represent the use of the respective type in a different function. This applies to Figure 10 and Figure 11 as well, which display the number of occurrences for each function of okay and gut separately, presented as ranked horizontal bar plots.

Figure 9, Figure 10 and Figure 11 highlight the relevance of the structural function of DMs, at least in institutional interactions such as at the Palliative Care unit.

For okay (Figure 10) and gut (Figure 11), discourse structuring at a macro-level, i.e., “in closings and transitions” (Mondada & Sorjonen, 2021, p. 93), emerges as the second most frequent use of both polyfunctional expressions, showing a similar picture in both the absolute and relative numbers of occurrences (okay: n = 133, 17%; gut: n = 134, 15%). All the ambiguous cases that include discourse-structuring, but also other functions, are not even included in this, but visualized in the form of the light-colored bars (see legend in Figure 9).

This means that transitional functions even surpass positive evaluation and concessive use in the case of gut (cf. Arens, 2023) and agreement marking as well as the use as a question tag and continuer for okay (cf. Betz et al., 2021).

However, the composition of the specific functions differs to some extent: whereas gut (‘good’) is still predominantly used as an adjective (n = 478, compared to okay as an adjective: n = 71), okay most frequently functions as a display of understanding (cf. Betz & Deppermann, 2021), which is only the fifth most frequent function for gut (n = 43) in the Palliative Care corpus.

Together with the ambiguous cases that overlap with other functions (particularly agreement and auto-dialogical uses, for the latter see Auer (2021)), okay is involved in discourse structuring at the macro-level in 38% of all its occurrences in the corpus, while gut fulfills this function at least in 26% of the cases (Figure 9).

In sum, the Figure 9, Figure 10 and Figure 11 underline the prominence of the discourse-structuring function of okay and gut, which still continues to be neglected in standard dictionaries (cf. Oloff, 2019). In both DMs, there is only one prototypical use (gut: adjective, okay: display of understanding) that surpasses the amount of instances of the use as a structural DM at a conversational macro-level.

4. Discussion and Conclusions

This study’s aim was to provide a systematic account of structural DMs operating at a conversational macro-level in a comprehensive corpus of German Palliative Care interactions (cf. Buck, 2022; Section 2). Various analytic perspectives were applied in order to deepen the understanding of DMs as a group of functionally similar, competing, and complementing expressions.

First, the analysis of the competing DMs in similar contextual environments revealed functional similarities but also differences across single DMs (Section 3.1).

While so is produced most frequently in openings and transitions to non-verbal activities, okay and gut are predominant in topic shifts and pre-closings, where generally more different types (such as also, genau, super, wunderbar, schön, alles klar, etc.) as well as a higher number of tokens in general occur. Although previous observations with regard to their functional similarity as transitional markers hold true (cf. Meier, 2002; Auer, 2021; Pfänder et al., 2024), the results showed once more that empirical analyses of natural language data are inevitable in order to sharpen the nuances and more precise distributions of DMs’ usage.

The results of so support the finding that, unlike for okay, gut, and other DMs, it “does not close prior sequences, but rather marks a readiness to transition from one action/topic to the next course of action” (Barske & Golato, 2010, p. 247).

In openings, the “contact signal” na (Schwitalla, 2002, p. 262; see also Rensch, 2018) can also occur, which is not the case for any of the other macro-level conversational transitions.

Moreover, the analysis of possible and recurrent co-occurrences (Section 3.2, see also Cuenca & Marin, 2009) shed light on the DMs’ usage and constraints and should be seen as an important piece of the puzzle in the DM research area.

Okay + gut emerged as the most frequent combination in the Palliative Care data and is much more likely than the opposite gut + okay (in line with Arens, 2023, p. 262). Also is barely combined with other DMs in first position; however, it is more likely if also is following DMs like okay, gut, or genau.

Co-occurrence with temporal (jetzt, dann, cf. Imo, 2010; Staffeldt, 2018) and conclusive elements (also, dann, cf. Deppermann & Helmer, 2013) could also be found frequently, confirming previous observations (cf. Meier, 2002; Imo, 2010; Pfänder et al., 2024).

However, they are not equally distributed, as so is more frequently combined with jetzt. In contrast, all other structural DMs show a preference for co-occurrence with dann. The co-occurrences and their restrictions as well as the position within a DM cluster (Excerpt 13, see also Cuenca, 2024) could provide insights into an implicit hierarchy among different DMs, which still needs more exploration in the future.

The provided qualitative transcript excerpts were useful in order to gain more detailed insights in the ‘ecological habitat’ where the DMs are embedded as well as the way they are used in natural interaction. However, further detailed analysis based on Interactional Linguistics (cf. Couper-Kuhlen & Selting, 2017) will be necessary to reveal how exactly certain DMs fulfill discourse-structuring functions (e.g., Auer, 2021; Betz et al., 2021; Arens, 2023; Pfänder et al., 2024), but also how they do that in comparison to other DMs within the same language.

Another interesting finding of this study is the variety of DMs that occur in the data, especially for topic shifts and closings, including the different origins and grammaticalization paths of single DMs (cf. Degand & Evers-Vermeul, 2015). Although this is no new insight for German (cf. Willkop, 1988; Schwitalla, 2002; Trotzke et al., 2020), these might be relevant observations for future research. Given the relative homogeneity of the Palliative Care corpus, comparisons with other communicative settings would be fruitful.

Another dimension of further comparison could be (micro-)historical, as DMs are prone to language change (cf. Auer & Günthner, 2005; Degand & Evers-Vermeul, 2015). For instance, the international DM okay is on the rise in many languages around the world (cf. Betz et al., 2021; Betz & Sorjonen, 2021) but was not yet mentioned in Willkop’s (1988) exhaustive study on structural DMs in German. Other native (e.g., fein, Schwitalla, 2002) and non-native DMs in German (for instance, the French bon, cf. Meier, 2002) are mentioned in the literature, but absent in the Palliative Care data, which could be due to regional or situational variation.

Interestingly, the particle naja (and, similarly, tja) has also been described to occur in topic shifts and conversation closings, noting that it often conveys a nuance of resignation, though (cf. Willkop, 1988), but is, however, not found in the Palliative Care interactions. Given the omnipresent threat of death at the Palliative Care unit, this might just be the wrong setting to end a conversation with a notion of resignation, which would underline the finding of optimistic projections in the closing sequence (cf. Jefferson, 1984). Another explanation for the lack of naja in the Palliative Care corpus could be that naja is rather informal and thus not appropriate for institutional communication. Naja has been described to be used when “the speaker is interested in continuing but leaves it up to the co-participant to do so” (Barske & Golato, 2010, p. 262) and would therefore not fit a setting of deontic asymmetries, where physicians and caregivers lead the conversation.

In addition, differences but also similarities could be found across different participation groups (Section 3.3). The healthcare personnel produced significantly more and less ambiguous structural DMs than patients and their relatives, which again reflects the institutional asymmetries found in medical interactions (cf. Mishler, 1978; Heritage, 2011). Physicians, however, use more DMs than caregivers (particularly okay), which can be explained by their different professional roles and requirements.

While caregivers have to fulfill lots of caring activities (e.g., wound care, assistance with dressing, washing, etc.) and use more so for those transitions to non-verbal activities (Section 3.1.2; see also Pfänder et al., 2024), physicians have a more dense agenda to discuss with the patients, and hence, a well-structured discourse management is required, evoking more DMs, especially in topic shifts (Section 3.1.3). Caregivers, in contrast, may have more time to actually use conversation “as the ‘procedure’ of Palliative Care” (Gramling & Gramling, 2019, p. 14) and give space for informal talk, thus the chance for “topic shading” (Schegloff & Sacks, 1973, p. 305) and less structural DMs increases.

Finally, the quantitative analysis of the different functions of the two most frequently occurring DMs (okay and gut) highlighted the significance of the discourse-structuring function, where it emerged as the second most frequent function in both expressions (Section 3.4). This underscores the necessity for standard dictionaries to include solid empirical findings on spoken language in general, and on discourse-structuring functions in particular (see also Oloff, 2019).

Moreover, such form-based analyses as in Section 3.4 are useful and necessary for gaining a better understanding of the DMs’ multifunctional spectra, as well as for identifying and elaborating on functional differences among similar DMs. For gut (see also Arens, 2023), the use as an adjective is dominant, followed by the use as a structural DM, for positive evaluation, as concessive use, as a display of understanding, and as an agreement marker (Figure 11). Although all of these functions are also found for okay, their relative frequencies differ notably (ranked in this order: display of understanding, structural DM, agreement, question tag, etc.; see Figure 10).

Taken together, this study hopefully demonstrated the benefits of combining multiple approaches to DMs within a single language and a relatively homogeneous communicative setting, which of course should be seen as a fundament for further contrastive work, across languages as well as across different interactional environments.

Both qualitative methods from Interactional Linguistics and quantitative methods from corpus linguistics are essential to empirically deepen our understanding of the very fascinating functional category of DMs, including different, highly multifunctional, and no less fascinating linguistic expressions that compete with but also complement each other.