Epistemic Stance Markers in German and English as a Lingua Franca Media Sports Interviews

This study investigates a particular type of media interview to explore the use of epistemic stance markers in professional media sports contexts. The study focuses on interviews with male professional football players usually taken straight or very shortly after the match (post-match interviews or PMIs). Two data sets were investigated using a simple quantitative and an ethnographic conversation analytic approach: 57 interviews conducted with German professional football players in German, and 27 interviews conducted with professional football players of various nationalities and first languages, including German, in English as a lingua franca (ELF). The aim of this study is to find out how the use of epistemic stance markers such as I think in English and the German equivalents ich denke/ ich glaube contributes to the foregrounding of the player’s perspective and thus the maintenance and negotiation of an epistemic gradient, which is essential for a smooth and unproblematic progression of the interview.


Introduction
This study investigates a particular type of media interview to explore the use of epistemic stance markers in professional media sports contexts. The study focuses on interviews with male professional football players usually taken straight or very shortly after the match (post-match interviews or PMIs). For the purposes of this article, two data sets were investigated using a simple quantitative and a qualitative ethnographic conversation analytic approach (ethnographische Gesprächsanalyse, Deppermann, 2008): one dataset contains 57 interviews conducted with German professional football players in German, the other dataset contains 27 interviews conducted with professional football players of various nationalities and first languagesin English as a lingua franca. The aim of this study is to investigate how the use of epistemic stance markers such as I think in English and the German equivalents ich denke/ ich glaube contributes to the foregrounding of the player's perspective and thus, the maintenance and negotiation of an epistemic gradient, which is essential for a smooth and unproblematic progression of the interview.
The PMI is an established element in the television broadcast of a football match and can be viewed as being part of a genre chain (Chovanec, 2018: 30) that creates a narrative extending beyond the actual match both temporally (by including reports on past and future events) as well as spatially (by reporting from different locations such as the stadium, the studio and the wider urban environment, Wilton, 2017), using a variety of documental and fictional as well as scripted and unscripted genres (Adelmann & Stauff, 2003: 110).
The PMI has a number of typical characteristics that mark it as a distinctive media genre: the interviews take place right after the match and are comparatively short -on German television, they usually consist of three question-answer pairs. While the questioning turns are usually kept very short, the player's replies are more extensive and elaborated. This self-presentation becomes ritualised in the interview through the rigid structure, the frequent use of repetitive and often formulaic language and the overall aim to conduct a harmonious and essentially cooperative exchange (Wilton, 2019; see also Caldwell, 2009 andFile, 2012). Thus, the PMI is not the place to discuss and evaluate the match in all its details or to hold the player accountable for his actions, but to turn the player's individual experience into a collective emotional experience which is shared by the audience (Montgomery, 2010;Wilton, 2019). This observation already points to the underlying epistemic relationship that the PMI has to establish and maintain in order to progress smoothly: the player's perspective is the dominant one, and it is the task of the interviewer to establish that dominance in the questioning turns. At the same time, however, he or she has to present him/herself as an expert in the field of football. The player's task is to mark his perspective explicitly and to counteract any implicit or explicit violations of his epistemic authority. The use of stance markers, therefore, plays a vital role in this interactional accomplishment.
These observations hold true also across linguistic and cultural boundariesfootball is a global sport being played and broadcast locally, regionally and nationally all over the world. Professional football is now highly international, with a volatile transfer market of players migrating temporally to another country, often several times during their careers. Internationality is also created by highly popular tournaments such as the World Cup or the Champions League, and national competitions of internationally renowned leagues such as the German Bundesliga or the English Premier League. Such events are broadcast around the world, often addressing an international audience in English as a lingua franca (ELF). Players participating in such tournaments are frequently required to be available for PMIs, for which ELF is then the language of choice. In such ELF situations, Seidlhofer (2011: 18) states, "it is usually taken for granted that speakers will have a command of English that varies along a continuum from minimal to expert, but that they regard themselves as capable of accomplishing the task at hand" and "that they can meet the requirements of participation in a particular speech event". Previous research on the datasets discussed in this article has shown that players (and interviewers) are able to fulfil these expectations even with a limited command of English. The ritualised structure and the repetitive use of locally produced and conventionalised lexical material typical of the PMI allows players to fill their slot and participate satisfactorily in the exchange (Wilton, under review).
Section 2 introduces the main characteristics of the PMI. Section 3, a quantitative and qualitative investigation of the data reveals distinct patterns of distribution and interactional function for epistemic stance markers in the German and ELF versions of such interviews. Section 4 summarises and evaluates the results before giving some suggestions for further research.

The post-match interview as a media ritual
In characterising the PMI as a media ritual, its typical -and often criticised or ridiculed -features can be regarded as constituting a recurrent event that has a media social function rather than a purely informative or democratic function, i.e. to hold public decision makers accountable for their actions. Previous investigations into its structural and interactional characteristics have revealed the following features: (1) The interviews show a distinctive pattern of repetitions and uptakes (Wilton 2019) that connect the player's replies to the previous journalist's turn. By explicitly repeating lexical material from the journalist's turn in their own contribution, players connect to the questioning turn on the surface of the interaction, relating their contribution to the previous turn and thus designing it as a relevant contribution. The pattern reveals that repetitions overwhelmingly connect player's answers either with the previous journalist's turn or their own previous turn(s). Journalist's turns hardly ever display repetitions or uptakes of lexical material used in a preceding contribution by the player, i.e. they do not refer back to the player's replies to advance thematic progression. This results in the interview consisting of (usually) three thematically rather independent Q&A pairs that are only loosely connected to each other by the overall thematic orientation to the match. Similarities between interviews within and across the German and ELF datasets reveal that the first question tends to address the match in general (e.g. its result), the second question zooms in on a notable aspect of the match (individual achievement of player, a goal, a foul, a special tactical move etc.) and the third and (usually) final question focuses on the relevance of the match for the future (progression of the tournament, subsequent changes in strategy etc.). This pattern reinforces the impression of a routinised procedure in which the details of the player's contributions remain largely without consequences for the successful completion of the exchange.
(2) A further characteristic contributing to the ritualistic impression is the frequent employment of formulaic language, often in combination with repetitions and uptakes. Typically, the beginning and the end of an interview are accomplished through conversational routines (Coulmas, 1981) such as congratulatory and leave-taking phrases, good luck wishes or mutual expressions of thanks. A restricted set of phrases and mostly general sports vocabulary are used to comment on a recurrent and restricted set of actions (Wilton, 2019). Players use formulaic language to expand and structure their turns, which are expected to be longer than a simple affirmative or negating reply. Players might take up phrases that were used in the previous questioning turn, conventionalised phrases that are used in everyday or sports contexts and phrases that are produced locally (individuelle Formulierungsroutinen, Dausenschön-Gay et al., 2007: 182) and often used more than once, all of which help the players to produce speech in a physically and cognitively demanding situation. It is much more common in the interviews to use words and phrases from everyday language or the general sports register than highly specialised terminology. For example, in the German interviews, it is much more frequent to comment on the scoring of a goal with ein Tor machen (to make a goal) than with ein Tor schießen (to shoot a goal) or even ein Tor erzielen (to score a goal) (Wilton, 2019).
(3) The supporting function of formulaic language becomes even more explicit in interviews that are conducted in English as a lingua franca (Wilton, under review). Despite varying degrees of competence, interviewers as well as players manage to accomplish the task of providing a commentary on the match which foregrounds the player's perspective. The predictable structure, predefined interactional roles, reduced importance of coherence, content and thematic progression and the frequent use of set phrases and formulaic language enable the participants to deliver acceptable contributions in the appropriate slot in the interaction. The data show that in contrast to non-institutionalised, less predetermined ELF conversation, PMIs exhibit a very low level of negotiation for meaning, repair or mutual support in the production of meaningful speech among participants.
(4) The basic idea of an interview is to elicit information. However, in the case of a PMI, this information is unlikely to consist of general and basic information about the match, as both journalists and the audience in-and outside of the stadium have just witnessed the event in question. What is required from the participants is to provide room for the delivery of an evaluation of the match by one of the active participants. However, as the players are interviewed right after the match and therefore had little to no time to reflect on it, this evaluation can only be expected to be preliminary and/or rather general. Consequently, the PMI is not the place to analyse the match in all its details nor to hold the player accountable for his actions and critically evaluate his performance. Instead, both participants strive to reach a consensual evaluation of the match as an alignment of the perspectives of the audience, the journalist and the active protagonists (Wilton, 2017) without risking conflict (Caldwell, 2009;File, 2012). Interviewers frequently either remain neutral or empathise with the interviewee either in their defeat or in their celebrations (Montgomery, 2010).
The above observations suggest that the PMI is a genre with ritualistic features that serves as a means to unite players, journalists and audience by turning the player's individual perspective on and experience of the match into a collective emotional experience for a community brought together by the media and not primarily by physical co-presence (Wilton, 2019). As such, they contribute to the establishment of para-social relationships (Gleich, 2009;Horton & Wohl, 1956) that provide the illusion of a close, personal, one-to-one relationship between (members of) the audience and the persona of a public figure. Rituals are part of our social organisation, setting apart events that are important within a community from the everyday flow of life (Becker, 1995: 635). They become invested with a symbolic sense that is shared by the community (Mikos, 2008: 35), and create collective emotions that intensify the individual's perception of and participation in a temporary collective reality (Bergesen, 1998: 49).
The media play an important part in the constitution of sport events as media rituals (Bartsch et al., 2008: 11-12). If the media social function of PMIs is to make the player's perspective collectively accessible, then that perspective/experience has to be systematically, even ritually, foregrounded in the interview.
Means to establish this foregrounding can be found on various levels: first of all, media technology and infrastructure serve to foreground the player by zooming in on him after a first full shot of both participants. Furthermore, the interviewer has the power over the microphone and therefore over the organisation of turn-taking -the microphone visible in front of the player's face is a concrete manifestation of the fact that the player is expected to talk at a certain point in time. Microphone and camera are also reminders of the presence of the media audience for which the player's contribution is designed and made available.
Secondly, the timing and the setting of the interview are important. The players are interviewed right after the match, either still on the pitch or in the mixed zone, often still wearing their jersey and being out of breath, sweating and generally restless, i.e. still visibly marked by the physical experience of the match. The impression created is one of temporal and spatial immediacy. Thus, the media setting reinforces the superiority of direct experience through active involvement to that of indirect experience through observation.
Thirdly, the very nature of an interview implies an asymmetry of knowledge. The person asking the questions is assumed to have less information about an issue than the person being interviewed. In the case of media interviews, this is most obvious in expert or news interviews. The role of the interviewer is ideally that of a neutral agent acting on behalf of an anonymous and diverse audience, assuming a less knowledgeable position than that of the interviewee. In the PMI, this asymmetry is in danger of being jeopardised by two factors: on the one hand, journalists have been criticised frequently to be too uncritical, even ingratiating in PMIs. Depending on the outcome of the match for the player being interviewed, journalists frequently align their perspective of the match with the (supposed) perspective of the player: i.e., pride of a victory is reinforced by congratulatory and celebratory phrases and positive evaluations, while a defeat is commiserated upon and evaluated negatively: Overall, however, in these news interviews we find not neutrality and detachment but, on the contrary, strong affiliation by the interviewer with the interviewee. And the function of these latter moves by the interviewer (e.g. 'you timed it to perfection') is in part to provide the warrant for the particular character of the interview itself. In effect they establish for the overhearing audience some special quality of the contestant's performance at the same time as congratulating them in person. In this way they encourage a different kind of alignment between the audience and the interviewee. Whereas in the adversarial accountability interview the audience is invited to scrutinize the interviewee for signs of evasion, in these interviews the audience is offered the opportunity to co-celebrate with the contestant. (Montgomery, 2010: 196) On the other hand, both participants in this type of institutional interaction are professionals in their respective, but overlapping fields of expertise (sports journalism and football), they experience the same event, albeit from different perspectives (observer and actor) and through different types of involvement (reporting and playing). Consequently, they have different kinds of epistemic access to the event, resulting in different, but overlapping territories of knowledge. Even though both participants know the essentials of the match, the interview provides them with an opportunity to portray the player as an expert of his own experience for the benefit of the media audience. The data show that as long as the player's experience of the match is given epistemic primacy by the participants, the interview develops in a cooperative and consensual way. However, if the player's epistemic authority is challenged, the interview can develop into a competitive or even adversarial exchange in which fields of expertise and epistemic access are contested (Wilton, 2017).

Epistemic stance markers in post-match interviews
One means to express epistemic stance is the employment of conventionalised epistemic stance markers such as I think in English and ich denke/ich glaube in German. Generally, I + predicate combinations focus the ongoing talk on the current speaker: "I + predicate combinations in discourse are self-revelations. They are the prime sites of the speakers' self-stylization" (Baumgarten & House, 2010: 1185. Part of this selfstylisation is taking stance towards what is relevant in the ongoing discourse. The first person pronoun "is the most basic and prototypical source of subjectivity in language because it always explicitly refers to the speaker and thereby automatically introduces an explicit argumentative perspective to the discourse" (Baumgarten & House, 2010: 1185. In combination with verbs of cognition, the first person pronoun typically reveals subjective evaluations, attitudes and knowledge claims. This process of subjectification leads to a loss of grammatical dependency, therefore making I think/ich denke/ich glaube more versatile with respect to the syntactic positions in which they can be used. Their semantic meaning becomes more vague and context-dependent, while their pragmatic functions become more prominent (Imo, 2011;Baumgarten & House, 2010).
In her study of I think in political interviews, Simon-Vandenbergen found that the high frequency of the stance marker is linked to the "type of discourse in which the formulation of viewpoints is central" (2000: 59). This is certainly also to be expected of the PMI, in which the players are invited to present their point of view of the match. Furthermore, in such contexts, the assertive use is more frequent than the tentative use, which expresses uncertainty (Aijmer, 1997;Simon-Vandenbergen, 2000). A distinction between the two types of use is claimed to be possible according to syntactic and prosodic features (Aijmer, 1997;Simon-Vandenbergen, 2000;Imo, 2011). Referring to Fairclough (1992: 204), Simon-Vandenbergen (2000: 60) links the tentative use of stance markers not only to a hesitant and uncertain stance towards an issue, but suggests that the frequent use of I think might reflect an increasing conversationalisation of public discourse, reducing the formality of some media genres such as interviews and blurring the distinction between the private and the public.

Data and methodology
In order to explore these pragmatic interactional functions in the PMI, the following analysis will take a quantitative as well as a qualitative approach. The analysis is based on 57 German and 28 ELF interviews with male professional football players which appeared on television, club websites and social media sites such as YouTube. The interviews were transcribed according to the GAT2 transcription conventions (Selting et al., 2011). To reveal the frequencies and distributions of the stance markers in both datasets, a search tool designed to read GAT2 transcripts was used to provide a preparatory and supportive quantitative analysis to the following more detailed qualitative analysis, which uses an ethnographic conversation analytic approach to identify and analyse the functions of stance markers for the ritual realisation of the interview. A systematic (quantitative) comparison between the datasets in terms of the interactional functions of the stance markers is not intended. Thus, the approach taken in this study can be described as corpus-assisted rather than corpus-based (Partington, 2011). With this approach, I follow the study on epistemic stance markers by Baumgarten & House (2010).

Epistemic stance markers in PMIs -a quantitative survey
As a first approach to the data, a simple quantitative survey done on both datasets reveals the high overall frequency of first personal pronouns singular and plural (ich/I and wir/we). As Tables 1 and 2 show, in terms of overall frequency of ich/I the datasets are very similar. This similarity also holds for the normalised frequency, which is 22,38 for the German and 25,78 for the ELF dataset. Table 3 illustrates that the most frequent combination of the first person singular pronoun with a verb is I/ich + verbs of cognition (voc) such as I think/I believe/I mean and ich denk/ich glaub/ich mein. 1 The combination I/ich + voc in the datasets is restricted to the three forms that can be conventionalised as stance markers. Other combinations of I + voc such as ich vermute/nehme an/weiss/bezweifle or I assume/surmise/know/doubt/suppose occur very rarely (once or twice) or not at all. The comparison in Table 3 shows first of all that combinations of I/ich + voc take up almost half of all instances of I/ich in both datasets (42,12% and 47,76%, respectively). Furthermore, it is quite clear that the distribution of I+voc variants are different between the datasets. While ich mein/I mean is used only very rarely in both German and ELF interviews, ich glaub is used more frequently than ich denk in the German dataset, but the English equivalent I believe is not used at all, giving the combination of I + think a comparatively high frequency (92% of all I + voc). Other studies have remarked on the ubiquity and high frequency of the I + think combination as stance markers in spoken American and British English: I think and I don't know belong to the high-frequency I + verb collocations in spoken American and British English. I think is the single most frequent I + verb combination in the spoken components of the Corpus of Contemporary American English (COCA) and the British National Corpus (BNC). I don't know is the single most frequent negative collocation in both corpora. In other words, I think and I don't know are almost ubiquitous as stance-markers in spoken L1 English. (Baumgarten & House 2010: 1186. Similarly, in his study of German ich glaube, Imo (2011: 169) argues that the pragmatic similarity of ich glaube and I think as stance markers justifies their treatment as translation equivalents.
If we now look at the distribution of occurrences of ich/I + voc across speaker types we see that the overwhelming majority of ich denk/I think and ich glaub are produced by players (Table 4). Furthermore, it becomes clear that instances of I believe, ich mein and I mean can be disregarded as they either do not occur at all (I believe), very infrequently (ich mein) or are due to idiolectal preferences of a single speaker (6 out of the 8 instances of I mean were produced by the same speaker). In their analysis of I think in ELF conversation, Baumgarten & House (2010: 1187ff.) identify the expression of stance with I think through three formal structures: 1) as a main clause in a simple clause construction, 2) a main clause complement clause construction with I think as the main clause, 3) utterance-medial or final finite adverbial comment clauses as verbal routines.
This is similar to German, where the stance markers can either occur as a main clause ich + denk(e)/glaub(e)/mein(e) followed by a dependent complement clause or they occur with subject-verb inversion as denk(e)/glaub(e)/mein(e) ich in clause or utterance-medial or -final position, where they are less syntactically and prosodically integrated (Imo, 2011). Furthermore, the verbs are frequently phonetically reduced by the omission of the final schwa sound, and can be reduced to the verb only, as in in der zwEIten halbzeit hat das glaube jeder gesehn (in the second half, (I) think everybody saw that).
It is useful for the functional analysis of epistemic stance markers to refine the distinction between instances of I think/ich denk/ich glaub followed by a complementiser (that/dass) overtly marking subordination and instances of I think/ich denk/ich glaub followed by a dependent main clause. Imo (2011) shows that the distinction between German ich glaube as a main clause for an overtly subordinated clause, marked by complementiser and a finite verb at the end of the clause, and ich glaube as a main clause followed by a dependent main clause expresses a reduction in saliency or profile determinacy of ich glaube, decreasing its power to project. Furthermore, the use of post-positioned glaub(e) ich, in particular when it is prosodically and/or syntactically integrated, takes on the function of an adverb or modal particle, suggesting a broader application in spoken discourse than the fixed verbal form would suggest. The main overall function is to mark the speakers perspective in a process of subjectification (Imo, 2011: 186, Baumgarten & House, 2010: 1190, which might indicate different degrees of assertiveness or tentativeness. In the German dataset, 9 of the instances of ich glaube and 12 of the instances of ich denke are followed by a complementiser. In the ELF dataset, overt subordination with the complementiser that does not occur at all, which is in line with the findings of Baumgarten & House (2010: 1190: In the majority of cases the complementizer that is omitted. The L2 speakers show a much higher ratio of that-omission (7.6% of full structures) than the L1 speakers (25% full structures), which suggests that the L2 speakers are less aware of the structural variability of the collocation, and possibly also of the associated meaning differences with respect to the 'tentativeness' and 'deliberativeness' of the speaker's stance.

Interactional functions of stance markers and their role in the negotiation of epistemic authority
To explore the interactional functions of epistemic stance markers in their discourse contexts, we will now look at examples from both datasets in more detail. In Excerpt 1, the player uses ich glaub three times in his reply.

Excerpt 1
Interviewer: Int Philipp Lahm: PhL This sequence shows how the functions of the first person pronoun singular and plural are typically distinguished in PMIs: ich occurs only as part of an epistemic stance marker, while we is used to refer to the team's actions in the evaluative description of the match. Even when asked explicitly about their own actions such as scoring a goal, players tend to avoid reference to themselves and prefer to refer to the whole team as an actor. The use of this so-called "modesty plural" (Du Bois, 2012: 324) serves to minimise self-praise and threats to the player's negative face. Thus, the stance taking function of ich in combination with glaube takes a prominent role in the player's turn.
As Baumgarten & House (2010: 1192 note, a typical context for the occurrence of I think is (at the beginning of) an answer to a question. Although the formal characteristics of the interview presuppose a Q&A structure, the questioning turn does not necessarily contain what would formally be classified as a question. Heritage (2013: 385) and Clayman (2010: 257) argue that questioning can be done by utterances that in other contexts would not qualify as a question, such as declaratives with falling final intonation. Similarly, the player's reply does not necessarily fulfil the criteria of a typical answer in the sense of providing previously unknown information, but is seen more generally as a reply or response which might or might not contain an answer to a previous question (Lee, 2013: 416). In an institutional context such as the media setting of the PMI, the pre-allocation of turn types provides a structural frame within which turns can deviate from a strictly questioning or answering format and still be treated as acceptable contributions (Ehrlich & Freed, 2010: 5). Therefore, the actions performed within the Q&A structure by the participants in a PMI can be more broadly characterised as an initial request or invitation to comment and/or evaluate by the interviewer and a subsequent delivery of a comment/evaluation by the player.
In line with this characterisation, ich glaube is used three times in sentenceinitial position with a starting point function (Kärkkäinen, 2003: 121ff.): in lines 13 and 18, the player starts his evaluation of aspects of the match with ja ich glaube as a matrix clause with a dependent main clause, marking and asserting the following evaluation overtly as his subjective assessment. Ja in this context serves as a marker of an initiated planning process (Imo, 2013: 176) after a specifying wh-question (Fox & Thompson, 2010), introducing an explanation that was requested by the previous turn. The player responds to that request by accepting the presupposition that something has changed and elaborates on the team's improvements that led to the victory. The use of ich glaube in line 18 can be seen as marking a boundary towards a slight topic shift from positive evaluation to negative evaluation (Kärkkäinen, 2003: 143). In line 22, he employs glaube ich, the inverted form, in combination with aber (but) and a summarising insgesamt (all in all) in order to set off and introduce his final evaluation of the whole match. The employment of but as a marker of contrast (Schiffrin, 1987: 152ff.) serves three functions in this position: it marks a semantic contrast to the immediately previous evaluation of the team's negative performance, it marks a return to the positive assessment at the beginning of the player's turn and it serves to introduce a final return to the interviewer's invitation to evaluate the match. This foregrounding of an evaluation as contrastive and/or independent from previous evaluations occurs relatively frequently in both datasets: 14% of all occurrences of I think are in combination with but, 15% of ich denke occur in combination with aber, as do 16% of ich glaube.
All three occurrences mark the player's turn as the expected and required subjective assessment of the match, foreground the player's perspective against the actions of the team and structure the turn into an initial part containing a positive evaluation, a medial part containing a negative evaluation and a final part containing an overall summarising evaluation. Thus, the starting point function, in addition to routinely marking the beginning of the current speaker's perspective, indicates the treatment of boundaries between parts of talk (Kärkkäinen, 2003: 143ff.).
The use of the inverted form glaub ich in sentence-or utterance-medial or final position is thought to indicate a less assertive, more tentative stance towards the proposition in question (Imo, 2011;Aijmer, 1997). Prefaced by aber and insgesamt, the expression loses its prominent sentence-initial status, and as an inverted form, its power of projection (Imo, 2011: 182). However, glaube ich in this case still appears to convey an assertive stance rather than a tentative one, because it is used in a summarising statement that is meant to foreground the player's overall assessment. The summarising function of glaube ich is marked and enhanced by the contrast to previous talk with but and by the summarising adverb insgesamt, making the interpretation as a marker of uncertainty less likely.
In Excerpt 2, again from the German dataset, the case for an interpretation of glaube ich as an uncertainty marker is clearer:

Excerpt 2
Interviewer: Int Jörg Butt: JöB The interviewer puts forward an evaluation which is clearly marked as his own and invites the player to agree with him. The invitation is realised as a closed question to which a positive answer, i.e. agreement is preferred (Lee, 2013: 423;Pomerantz, 1984: 63). Thus, the pressure to comply with this expectation is relatively high for the player, and his initial hesitation at the beginning of his turn in line 3 already expresses the need to deliberate. The following token of agreement JO is conventionally used to indicate that agreement is only partial or provisional and might even be followed by a counter statement (Imo, 2013: 169ff.). It is then quite plausible to interpret the occurrence of glaub ich in line 3 as a marker of tentativeness. In the continuation of his turn, the player recounts the match in its progression, but fails to align this with what the interviewer's evaluation suggests: a difficult start resulting in a deserved victory. Instead, from the description of the player, it becomes clear that the difficulties, but also the achievements occurred throughout the match and that apparently, there was no significant improvement during the course of the event. This discrepancy is reflected in the hesitant agreement the player showed as an initial response. In the remaining part of his turn from line 13 onwards, he makes an effort to explicitly align his own experience of the match with that of the interviewer as an observer, using the indefinite pronoun man (one) to potentially include anyone, including himself, who was in a position to reflect on the match while it was still running. Finally, he takes up the second part of the interviewer's evaluation and presents the deserved victory as his conclusion. Introduced with andererseits (on the other hand, line 18), it is clearly set off from the preceding assessment of a difficult match as a contrast, but the following assertion is less convincing than in the previous example. Particularly noticeable in this case is the use of the particles schon and auch (line 19) which can both be used to assert, but also to concede, especially so in the combination schon auch. The use of glaub ich and the almost random placement of schon and auch within the utterance indicate a hesitant and tentative concession to the interviewer's evaluation, making the summarising function less prominent here than in the previous case.
A very similar start to an interview in the ELF dataset results in a much more assertive response by the player:

Excerpt 3
Interviewer: Int Per Mertesacker: PeM The interviewer's closed question suggests that an affirmation is the preferred response. After some hesitation, in line 3 the player delivers an overtly affirmative response with an agreement token (yeah) and I think so followed immediately by a prepositional phrase giving the reason for the player's conviction (because of the second half). The affirmative formulaic expression I think so is used three times in the ELF dataset, and also occurs as ich denke schon twice in the German dataset -the particle schon here enhancing the agreement. Continuing his turn, the player elaborates on the development of the match, using I think in line 7 to express uncertainty about the number of times the team hit the post. When referring to events and actions rather than giving an evaluation, such as in this example, I think and its German equivalents are used as markers of uncertainty, regardless of their syntactic position.
The interviewer's second question, a wh-question, leaves potentially more room for the type of answer that can be regarded as appropriate. However, questions addressing the team's character are usually treated as opportunities for the player to assert the team's motivation, will to win and perseverance. Thus, the player's overtly assertive statement in line 14, preceded by I think and followed by a rejection of recent rumours, serves exactly that purpose. Furthermore, the player links his reply to the previous questioning turn by taking up two salient words -character and proved in lines 14 and 17 -and qualifying them positively -with great and a lot. The uptake in the final sentence of the player's turn directly refers back to the interviewer's question and functions as a summarising and closing statement of a turn that is designed as an authoritative evaluation of the match result.
In Excerpt 4, the same player uses I think twice in the same turn. Despite the fact that in both instances I think is used as a main clause followed by a dependent main clause, the functions of I think are different. In the first instance in line 63, I think is used as a starting point for an evaluation, while in the second instance in line 66 it is used as an uncertainty marker, indicating that the player cannot remember exactly when the match against Barcelona will take place.

Excerpt 4
Interviewer: Int Per Mertesacker: PeM The distinction between marking conviction towards an evaluation and marking uncertainty towards a proposition is also evident in Excerpt 5.
In line 23 the player evaluates the performance of the team in the first half with I think as a main clause followed by a dependent main clause. In line 25, he uses postpositioned I think to express his uncertainty about the number of chances in that first half. Additionally, the use of I think in connection with but in line 23 functions as a marker of change or contrast, separating the previous talk from what is to follow. In this instance, the player starts his turn by showing agreement with the interviewer's suggestion that an early goal would have been necessary for the team's success. Furthermore, the interviewer already suggests in his question that the player might have lost hope of a victory during the course of the match. With but I think, the players sets off his own evaluation from his previous agreement to the interviewer's suggestion and evaluates the first half of the match as not good enough before elaborating on the further progress and outcome of the match. He thus marks his own evaluation as different and independent from a) his own previous agreement and b) the interviewer's rather tentative suggestion.
In Excerpt 6 from the ELF dataset, the player heavily relies on I think to structure his turn and assert his perspective.

Excerpt 6
Interviewer: Int Shkodran Mustafi: ShM 01 Int: shkodran; have you been involved in a (.) !BE!tter; 02 (.) arsenal performance (.) than THAT. 03 ShM: i THINK so yes; 04 i THINK we had eh-05 (--) since i joined here i think we had a few (.) 06 GOOD performances i think eh-= 07 =the only (.) important thing is to to keep it UP; 08 °h i think we did wEll toDAY, 09 it was ehm (.) a DERby it's you know; 10 so it's a special GAME;= 11 =but i think eh-12 °h we showed CHAracter, 13 e:hm (.) we were there in the ehm important MOments, 14 and i THINK ehm (-) eh-15 when you do all THAT i think ehm-16 you win GAMES.
The player replies to the interviewer's question with an overtly assertive I think so in line 3, with think prosodically prominent and complemented by a token of agreement (yes). In line 4, he begins his evaluation with I think, which again receives prominent stress, then breaks off and restarts in line 5 with a restriction that indicates that his evaluation only refers to his personal experience with the club (since I joined here). In the continuation of his turn, it is difficult to accurately assign the occurrences of I think to preceding and/or following speech. Prosodically, as indicated in the transcript, it appears that the player's evaluation in lines 5 and 6 is framed by a pre-and a post-positioned I think, which can be taken as an overt assertion of his evaluation. However, the occurrence of I think followed by a short hesitation marker in line 6 is latched immediately onto the following statement. Thus, I think in line 6 might function as a pivot, simultaneously asserting a previous and a following evaluation. Addressing the match in question in line 8, the player once more introduces his positive evaluation of the match with I think before commenting on the fact that a derby is a potentially difficult (special) match. The use of but I think in line 11, then, indicates a return to the positive evaluation of line 8 (Schiffrin, 1987: 117), emphasising the achievements of the team and ending in a summarising statement introduced with and I think. The instances of I think in lines 14 and 15 show a similar function to those in lines 5 and 6: they provide a frame to the first part of the sentence while the second I think also projects forward to include the following statement, giving it extra salience as a final and general evaluation.
What becomes clear from the examples so far is that variants of I think/ich denke/ich glaube are used extensively and repetitively in the players' turns where they fulfil various functions depending on the context and the position in which they occur. While in German, the functions can -but need not be -differentiated by a greater variety of syntactic constructions and phonological forms, the ELF data shows that I think is used generically as a multifunctional marker of epistemic stance. Thus, the extensive and multifunctional use of stance markers in the players' turns shows the importance of marking subjectivity and epistemic stance, regardless of whether it expresses conviction or uncertainty. This emphasises the foregrounding of the player's perspective as a typical characteristic of the PMI.
In contrast, interviewers very rarely use epistemic stance markers in their turns and when they do, players signal a problem in their following reply, as in the next excerpt.

Excerpt 7
Interviewer: Int Thomas Müller: ToM In Excerpt 7, the interviewer presents an observation as generally valid through the use of the indefinite personal pronoun man (one). He then puts forward his own personal observation, introduced by ich glaube as a main clause followed by a dependent main clause (line 26). This main clause is a closed question ending with a tag (oder?). This design a) presents the presupposition as the interviewer's personal assessment, b) asserts the interviewer's presupposition that the player must have fun and c) expresses a preference for agreement (Hayano, 2013: 405). In line 27, the player replies with ja klar, which is a combination of two response tokens (ja and klar, Imo, 2013: 167) and a German equivalent of of course. It has been shown that in response to a question, of course does offer confirmation, but that it "also treats the alternative (…) as inconceivable" and therefore speakers "contest the presupposition of the question that both confirmation and disconfirmation are possible and thus treat the question as unaskable." (Stivers, 2011: 87). In Excerpt 7, despite the fact that the interviewer's question is designed to invite agreement, the alternative of denying the presupposition remains viable and is addressed by the player in the design of his reply. In the continuation of his turn, the player explicates why he deemed the question unaskable: it is obvious, i.e. firmly grounded in general knowledge, that playing together with excellent players is fun. He marks this by asking a rhetorical question in lines 28 and 29, suggesting that anyone in his position would have fun. The uptake of touch in line 31, albeit with its different, i.e. German loanword, meaning, overtly aligns the player's reply with the interviewer's questioning turn. However, already in the following line 32, the player presents a contrasting perspective, introduced by nichtsdestotrotz (nevertheless). In the remainder of his turn, he goes on to explain why he thinks that the individual achievement of his teammates is less important than their contribution to the team as a whole. As a final move, he refers back to the issue of having fun by explicitly assigning fun to his teammates (line 41) and to the whole team (lines 40 and 43), including himself (uns allen). It becomes clear that he treats the interviewer's subjective reference to the player's personal experience as inappropriate on two levels: first, he rejects the possibility of not having fun as inconceivable based on epistemic access, thereby taking the moral high ground (Stivers, 2011: 88). Second, the rejects the invitation to comment on his emotional state by diverting attention away from the personal evaluation of the interviewer to an evaluation of his teammates' achievements for the teams' success.
In Excerpt 8, the interviewer's personal evaluation is rejected both on an epistemic as well as on a propositional level: In this interview, the interviewer puts forward an evaluation of a conceded goal as a personal statement marked by glaub ich and intensifies his commitment by stating that there can be only one opinion about the reason for the goal -a mistake by the player. Despite the potential interpretation of glaub ich as a downtoner (Imo, 2011: 180) and the final rise in intonation as a marker of a question rather than an assertive statement, the interviewer's turn conveys a strong commitment to the expressed proposition. This interpretation can be seen in the reaction of the player, who expresses his incredulity first with a short ironic laugh, then with an almost formulaic counter question doubting the interviewer's serious intent. The interviewer, however, starts his reply by asserting his serious intent before changing his strategy. This change is indicated by the discourse marker also in line 6. He then tries to describe the shot that led to the goal, but gives up any further elaborations on his reasoning in favour of a much more open and less provocative question about the player's experience of the shot. The player uses this invitation as an opportunity to mock the interviewer's assumed expert status in lines 9 and 10 before launching into a more detailed description of his perspective as a participant in the match. He finishes his turn by another mocking remark, ironically praising the journalist for his keen observation skills.
In this excerpt the player shows that the personal evaluation of the interviewer is inappropriate on the level of epistemic balance because observation is seen as inferior to participation in terms of epistemic access and therefore, concerning the two competing types of expertise, the journalist's knowledge must be subordinated to the player's knowledge. The interviewer violated this expectation and subsequently, the player reacts with uncooperative behaviour. Additionally, the player addresses the central issue of how the shot resulted in a goal to refute the interviewer's claim that it was the player's fault.

Conclusions
From the above analyses we can draw the following conclusions: (1) The frequency and distribution of first person pronouns reflect their functions in the PMI: we/wir is used by players to comment on the team actions, and by interviewers to refer to the entity of observers, including themselves and the audience, whereas I/ich is used primarily and frequently by players to communicate personal involvement, perspective and evaluative stance.
(2) The functions of I+voc as epistemic stance markers are consistent with the player's task of providing a personal evaluation of the match. They have been found to aid the player in structuring his turn by marking boundaries within the turn (starting point function, closing/summarising function, contrasting function), by marking tentativeness and uncertainty and by asserting and marking the independence of stance.
(3) In particular, for the collocations of I think/ich denke/ich glaube with but/aber three functions have been identified: a) They are used to introduce a part of the turn that usually follows a first assessment of the proposition or evaluation proffered in the prior turn to introduce the player's perspective and to demonstrate its independence. The presentation of one's own perspective as contrasting or different, following a previous recognition of or concession to the interviewer's proposition or question, can be seen as a strategy to claim epistemic authority. b) They mark a return to a prior point made by the current speaker. Again, this can be seen as strengthening one's personal perspective. As Schiffrin (1987: 177) states, "the use of but in point-making has an expressive relevance, in that a repeated point displays a committed orientation toward a proposition, and an interactional corollary, in that stating one's point can take precedence over interactional goals." c) They often occur towards the end of the turn, introducing a summarising statement as a closing strategy, in some cases in the German dataset intensified by summarising adverbs such as insgesamt and ansonsten. Again, this foregrounds the player's perspective as the final word on the matter, marking the boundary of a largely self-contained Q&A sequence.
(4) A delineation between an assertive and a tentative use is not always possible, and syntactic position alone is not a reliable indicator. The examples showed that tentativeness can be reliably established when the player does not evaluate or when the stance markers are accompanied by other markers of tentativeness or deliberation. In other cases, both interpretations are possible, and stance markers occurring between clauses might even assume a pivotal function, qualifying both the preceding as well as the following talk as subjective and/or assertive.
(5) There is a general similarity in the use of epistemic stance markers across both datasets. The extensive use of ich denke/ich glaube and its variants in German and of I think in ELF shows similar frequencies, distribution and functional range in both datasets. The media institutional setting in which the PMI takes place imposes a number of constraints on the turn types and their allocations in order for them to count as acceptable contributions and to which participants orient in the design of their turns.
It is evident from both the German and the ELF data that the player's perspective, his subjective experience and evaluation of the match is routinely, even ritually foregrounded through the extensive and multifunctional use of epistemic stance markers by the players. In contrast, the interviewers hardly ever use epistemic stance markers, and when they do, the players' replies show that there is a problem with the questioning turn. This asymmetry in the use of stance markers is characteristic for the PMI and reflects a genre-specific epistemic gradient that needs to be maintained by both participants if the interview is to be successfully accomplished.
What remains to be investigated in more detail in the future is the way in which interviewers design their turns in order to "do questioning" or "do an invitation to comment" while keeping the epistemic gradient intact. Journalists' questions (…) participate in a distinctive environment that embodies a mix of professional and public accountability. Both of these dimensions, in turn, leave their imprint on the questions that reporters ask of public figures. What such questions are meant to accomplish, and the specific manner in which they are designed, are conditioned by specialized journalistic tasks and norms, as well as general public attitudes and preferences. (Clayman, 2010: 256) As a first speaker, the interviewer is responsible for each first pair part of a Q&A sequence -in fact, the interviewer has an obligation to "do questioning" (Clayman, 2010: 257). The elicitation of an evaluation from the player can be done in a variety of ways; direct questioning being only one of them. Another one is offering an evaluation for the player to agree to or to refute. As argued above, journalist and player are both professionals, and journalists -also as representatives of an informed audience -seek to display their own professional expertise by asking knowledgeable questions or by putting forward assessments based on knowledge and experience in the field. Thus, the ritual foregrounding of the player's perspective as epistemically more authoritative requires some skill and careful handling of the epistemic balance necessary for a successful interview. 1. For the German language dataset, reference to ich denk/ich glaub/ich mein includes the variants ich denke/ich glaube/ich meine as well as variants with reversed order such as denk(e) ich/glaub(e) ich/mein(e) ich.