Discourse Reflexivity-A Discourse Universal? The Case of ELF

This paper is concerned with metadiscourse, or discourse reflexivity, as a fundamental property of human communication. It assumes that lingua franca evidence is useful for discovering essential, possibly universal aspects of discourse, since participants must adapt to highly diverse and unpredictable circumstances while maintaining communicative efficiency. Metadiscourse is clearly not a feature speakers can dispense with, as is seen in the study reported here. Since ELF speakers cannot rely on much shared linguistic or cultural knowledge with their interlocutors, they need to secure mutual understanding by explicitness strategies, such as discourse reflexivity. This study shows that analysing interaction in dialogic speech events reveal important uses of metadiscourse that have not surfaced in earlier studies, which have used written monologues as their principal source of data. A much more prominent ‘otherorientedness‘, or orientation towards intercolutors, is evident in dialogue than in monologic language. The tendency of discourse reflexivity to collocate with hedges is also supported here.


Introduction
As evidence accumulates about other species, humans are undergoing the humiliating experience of noticing that they are not so unique after all. Other species construct tools, transmit culture, communicate about things beyond the here and now… Yet some things seem to remain ours only: our working memory capacity is way ahead of other species, we seem to be the only ones with a theory of mind, and our capacity for selfawareness and reflexivity is exceptionally high, perhaps unique. Not surprisingly, then, while other species turn out to have far more complex systems of communication than previously surmised, to the best of our knowledge only our languages can talk about themselves; in other words, the reflexive capacity of human thought is projected onto language. Human languages are reflexive systems.
Reflexivity as the capacity of language to refer to itself was recognised and discussed by many eminent linguists before it became a matter of widespread interest in applied linguistics. But it was only in the late nineteen eighties that it really took off, in the form of metadiscourse research, and has since enjoyed wide and steady interest among scholars. Pioneers in this line of research were William Vande Kopple (1985) and Avon Crismore (1989). From the 1980s onwards the term has often been used in a wider sense than language about language (as is very clear from the papers in the present volume). Metadiscourse research has embraced a broad range of phenomena, covering whatever is separate from the ‗primary' discourse (Crismore and Farnsworth 1990), the ‗topical' (Lautamatti 1978(Lautamatti /1987 text matter, the ‗propositional content' (Mauranen 1993), or in terms of Halliday's theory (e.g. 1985), the ideational metafunction of language.
Broadening the scope of metadiscourse beyond metalanguage, or reflexivity, the term suggested by Lyons (1977) for language about language, probably reflected a growing awareness in linguistics of the interactive aspects of language. Since the 1980s linguists have become increasingly aware that interpersonal features are ubiquitous and that even the most objective-looking written text is not without its interactional dimension. And it is indeed the objective kind of written text that has attracted the greatest proportion of metadiscourse studies: the most closely investigated domain is academic writing. The typical data may also have affected conceptual preferences. Thus, research has focused on expressions that indicate writer-reader interactionwhich can be seen as incorporating characteristics of speech into writing.
The all-encompassing notion of metadiscourse as the ‗nonideational' text matter, or what lies outside Sinclair's ‗autonomous plane '(1982/2004) has by now largely served its purpose of raising awareness about the nature of text as fundamentally interactional. Research into interactive aspects of language has expanded and diversified enormously since the early eighties. Studies of hedging, discourse particles, stance, evaluation, vagueness etc, have seen remarkable development and turned into burgeoning research fields in their own right. Perhaps the time has come to take a closer look at what the specific contribution of metadiscourse might be in this vast array of non-propositional language, and narrow down our own focus to what might in the end yield deeper insights into the workings of metadiscourse per se.
My intention in the present paper is to go back to the basics: to look at metadiscourse as language about language. Less traditionally, I draw my data from spoken interaction where English is used as a lingua franca. What connects the paper to the dominant research tradition is its source of data, which comes from university settings.
This study makes no comparisons to native speakers. We already know that native speakers of English use metadiscourse in academic speech (e.g. Mauranen 2001Mauranen , 2002Mauranen , 2003aMauranen , 2004Ädel this volume), and we know lingua franca speakers do. To understand academic speaking it is necessary to rid ourselves of the baggage of native English practices. Clearly, and more so than ever, English is the global language of academia. It is important to keep in mind that this globality means that most of the users are non-native speakers. Academic research is international by nature, not in itself associated with the preferences of a culturally or nationally defined language community. University institutions have been much more local, especially after the expansion of tertiary education in very many countries to include larger and larger proportions of each age cohort. But this has been changing at accelerating speed: universities are on a fast track to becoming globally intertwined, with increasing numbers of students and staff moving around from country to country. If the purpose, then, is to understand present-day academic speaking in English, we should look at the way English is used globally. To this end, English as a lingua franca is a better representative than native English.

Reflexive discourse
As far as we know, reflexivity is distinctive to human communication.
Since it is generally held that human languages are more versatile and effective than the communicative means of other species, we should expect this capacity for self-reference to confer advantages which go beyond the most basic requirement to get vital information across.
Reflexive language seems to be part of the more general ability to reflect upon our own experiences and actions; with some mental effort, we can distance ourselves from immediate experiences, identities, attitudes, and gut reactions, and subject them to conscious contemplation. Even though the processes we can bring to consciousness are only fragments of our entire mental activity, we are nevertheless able to think about our own thinking, to make it an object of thought itself. Such processes also allow us to talk about our talking: we can be aware of our verbalisations, and we can indicate this by means of verbalising itself.
Languages have a vast array of terms and labels that we can use for language related matters in everyday talk (he's pretty talkative; the way you write is sort of chatty; I have an uncle who speaks Malay…), as well as sophisticated terminologies of linguistics, but these are not necessarily reflexive. They are not always ways of speaking about the discourse at hand. Reflexive discourse is distinct from referring generally to language and its uses in that it relates to the ongoing discourse. Even though the borders may occasionally be fuzzy, the principle remains clear: reflexive discourse is discourse about the ongoing discourse.
By indicating our awareness of our talk, we share this with our interlocutors (One thing I wanted to ask you…) Reflexive language is therefore highly interactive, whether couched in monologic or dialogic form, and it sharpens conversationalists' mutual understanding of how to relate to the discourse that is being co-constructed. In this way, it contributes to the two fundamental uses that language has: sharing experience and negotiating interaction.
So far I have been talking about reflexive discourse rather than metadiscourse in this section. What is the relationship, then, between reflexivity and metadiscourse? I would like to see them as roughly synonymous, both referring to discourse about discourse. However, this conceptualisation excludes a number of things that have become commonly associated with metadiscourse, such as many non-reflexive interactive elements, as already pointed out. Strictly speaking it also excludes some elements that I have previously included in text reflexivity (Mauranen 1993), notably those I classified as reflexivity of low explicitness, such as connectors, because many of them do not make reference to discourse. In all, metadiscourse as a term has become to be used for a wider range of phenomena than reflexive discourse. Clearly, the difference between reflexive discourse and other discourse signalling items is a hard distinction to maintain, as can be seen for example in Carter and McCarthy (2006: 221). They talk about discourse markers that -enable speakers to monitor and manage the ongoing discourse by commenting explicitly on the process of talking itself‖, which is very close to reflexive discourse. As they go on to illustrate these discourse markers, some clearly refer to language, mostly with verbs like say, speak, and put, whereas others have no such reference (or rather, as it were, or well). The latter often signal reformulations or alternative expressions, even though they are not about language and not discourse reflexive.
The term metadiscourse has been contested altogether, notably by Sinclair (for instance 2005) on account of a misuse of the prefix meta-, which normally refers to something external to a concept or object, or an abstraction from it. In the philosophical sense, if we have a language, an object language, used for talking about the world, then a metalanguage would be a separate system for talking about that language. Basically a metalanguage is a formal language for analysing an object language. This is the way it is also used in mathematics, so that we speak of the ‗metalanguage of mathematics'. Analogously, we might talk about formal systems of linguistic analysis as the metalanguage of natural language, or the terminologies of theoretical linguists. This is not of course how the terms metadiscourse or metalanguage are used in linguistics; metadiscourse is not used for referring to a separate system set up to discuss discourse, but a label for certain integral parts of the discourse itself. The criticism for using ‗metadiscourse' to refer to ordinary, normal segments in ongoing discourse is justified to this extent. On the other hand, not all science follows the lead of mathematics, either, so for example in biology ‗metapopulation' refers to separated but interacting populations of a species taken together, see Hanski (1999).
The now established use of metadiscourse to refer to certain elements in discourse probably owes a good deal to the line of thinking where conveying information or sharing knowledge is the primary function of language, or the main level of language. We might call it the ‗primacy of the message'. In this way, anything else would be something extra, less essential, and less important. The early terminology reflects such thinking, with divisions into ‗primary discourse' and ‗metadiscourse' (see, e.g. Crismore and Farnsworth 1990).
Be that as it may, the term metadiscourse is well established in linguistics for broadly referring to discourse about discourse, or to an even wider range of discourse elements that incorporate interactive and text-organising functions. These two uses have been labelled as ‗broad' and ‗narrow', or ‗integrative' and ‗non-integrative' (see, e.g. Ädel 2006;Mauranen 1993). The labels that perhaps best describe the difference are those suggested by Ädel in this volume): ‗interactive' and ‗reflexive'. Although this use deviates from those of some other disciplines, linguistics has its own terminology. I shall be using the term discourse reflexivity whenever it is necessary to maintain the distinction between this sense of metadiscourse and its wider understanding, but when the difference is not relevant or not confusing, I employ metadiscourse in the same sense: discourse about the ongoing discourse.
Despite this less than felicitous term, the scholarship that it has inspired has captured important facets of natural language use; any conceptualisation of metadiscourse that I am aware of gives language about language a central position in the analysis. I would like to argue that this is indeed the core of metadiscourse. I also want to argue that it is a major element of natural language. Discourse reflexivity is a distinctive characteristic of language, ubiquitous in our speech, and it deserves close attention from linguists.
As noted above, discourse reflexivity contributes to the fundamental uses of language, sharing experience and negotiating interaction. More specifically, reflexivity helps discourse achieve two main purposes: to make discourse more explicit and precise, and to manage discourse strategically. Some examples from the MICASE corpus (Simpson et al. 1999) illustrate these uses in speech.
To add precision to communication, speakers make explicit how they wish their interlocutors to understand their contributions, how they interpret other speakers' talk, and what they expect from others' contributions. Thus, they can indicate in which light they wish their speech to be taken: I mention that just for those who are interested it is totally irrelevant to what I'm talking about, and how they understand others' speech: are you_ were you saying that or am I just hearing you, differently Speakers can also indicate their intentions by prospecting ahead: now i just wanna clarify, and i don't think this'll take long at all and retrospectively label preceding stretches of discourse: your complaint is that this claim is not the whole truth about the universe.
The last example, labelling discourse, already shows how reflexivity can be brought to serve the speaker's more strategic interests. Clearly, such labels assign discourse roles to the stretches in a way that reflects the current speaker's interpretation, which thereby gets imposed on the discourse under way.
This leads us to the other major function of reflexivity, which is to help manage discourse strategically, and to manoeuvre the discourse for desired ends. Speakers use reflexivity to this end for example in order to take the floor: may I ask a quick question?
to yield or offer the floor: I was wondering if you could comment on the differences or to impose order on the discourse: In brief, then, discourse reflexivity plays crucial roles in negotiating the flow of discourse between participants. In written discourse the terms of negotiation are necessarily somewhat different, with readers mostly remote to writers. Nevertheless, reflexivity is one of those text features where writing most clearly incorporates features of speaking. It is not surprising therefore that metadiscourse is generally depicted in terms of writer-reader relations, thereby representing the interactive aspects of language in written text.

Earlier observations on discourse reflexivity in ELF and native English speech
In view of the hitherto scant empirical research on ELF, not many studies have investigated metadiscourse in ELF yet. Previous comparisons on reflexive discourse in academic speech between native speakers of English and lingua franca users have shown that it is common in both (Mauranen 2005(Mauranen , 2007a(Mauranen , 2007b. In a similar vein, Penz (2008) found that participants in an intercultural European project employed a good deal of metadiscourse. In contrast, some early ELF research with simulated conversational or telephone data (Firth 1996, House 1999 suggest that ELF communication is so content-oriented that when faced with communication problems, speakers do not negotiate meanings or use metalanguage to sort them out but resort to topic changes or a ‗let it pass' strategy instead, to keep the conversation going. The discrepancy might be accounted for by situational parameters, by scholars' different conceptualizations of metadiscourse, or by the small amount of data in the very early ELF studies. It is also important to note that the early studies did not involve any data from native speakers, so that it is not known whether natives use metadiscourse in comparable circumstances, and whether the lingua franca users were exceptional in this respect. After all, metadiscourse studies have primarily associated it with academic discourse, which characteristically favours a high degree of explicitness.
Apart from the overall finding that metadiscourse was common in my own earlier studies, the main ELF uses were also similar to L1 English.
Most of the observed differences were concerned with form, which tended to be approximate rather than entirely accurate (I just like to make a brief comment of this). Some form-function pairings were also ignored in that certain fairly subtle distinctions were overlooked (e.g. the difference between REFER TO and BE REFERRING TO), which probably testify to some systemic simplification in lingua franca speech. Some senses and uses were also absent from expressions that otherwise were used in the same ways as in comparable native speech.
Even if discourse reflexivity may not be equally salient in all circumstances, it is nevertheless present in English L1 and lingua franca communication alike, and we might assume this is more widely true of other L1s and lingua francas; for example such a typologically distant language from English as Finnish (see Luukka 1992, Mauranen 1993 employs metadiscourse. This suggests that reflexivity may be a ‗discourse universal', (for discussion, see Mauranen 2003b) that is, such a major element of communication that languages generally possess means for expressing it and that these means are available to speakers as resources which they can draw on as necessary. Discourse reflexive expressions are thus very likely to be what Hunston and Francis call -necessary features of language‖ (1999: 270).

Data
This paper focuses on dialogic speech events. The aim is to open up research into dialogic use of discourse reflexivity, because what investigation there has been on spoken academic metadiscourse has focused either on lectures or undifferentiated monologic and dialogic data. Comparing speech to writing in itself brings new insights to analysing the functions of metadiscourse, as shown by Ädel (this volume), but taking dialogue under special scrutiny may open other new avenues. I start by briefly looking at a couple of phraseological units around typical discourse reflexive verbs (PUT and SAY). I then change the viewpoint and focus on the particular, by picking three dialogic events from my data and looking at them as social practices which emerge through layers of social formations.
The database I draw on here is the first part of the ELFA corpus (Spoken English as a Lingua Franca in Academic Settings; www.eng.helsinki.fi/elfa), ELFA(i), as I call it. From this corpus, I have included all dialogic files, altogether 400,000 words. ELFA was compiled with a deliberate bias towards dialogic events, and transcripts of events comprising a monologic and a dialogic part, such as presentations followed by discussion, common in conferences, seminars and thesis defences, were split into two along these lines.
For the analyses in Section 5, I use the ELFA(i) corpus as a whole. For the analyses in Section 6, I have selected three seminars for looking at metadiscourse in the context of whole events. The selection followed three simple principles: the events should be identified and labelled as -seminars‖ by the host academic community, they should come from different faculties, and the range of speakers' first languages should be as broad as possible. The purpose was to take a close look at a few events that were constant along some parameters (mode, genre, event type, language), but independent of each other and spanning a broad range in other respects (disciplinary domain). The events that ended up in the sample in this way were from medical science, political science, and women's studies. In line with the present dialogic orientation, these were all multi-party discussions, with the number of speakers ranging from 6 to 15. In one event, the discussion was preceded by an oral presentation, which is not included in the sample. The duration of the recordings was altogether a little over 2.5 hours, the total number of speakers 28, and the number of first languages 9, with Finnish (10 speakers) and German (7 speakers) as the largest groups.

Some corpus findings: PUT and SAY
As an illustration of the patterning in discourse reflexivity I chose two verbs that can be used reflexively (PUT and SAY) to focus in on. The choice of the former was based on seeing it listed in reference books and finding a number of occurrences of it (to put it bluntly), while the latter was selected because I had noted earlier (e.g. Mauranen 2001Mauranen , 2004) that it plays an important role in academic argumentation (what I'm saying).
To begin with PUT, it was used in dialogic speech largely with reference to verbalisation, specifically a certain way of expressing something (I thought that was nicely put), which is the way in which it is also used in Standard English as represented in major dictionaries and other reference works. A clear majority of the instances referred to verbalising (32/50), and well under a half (18/50) to other senses of the verb (should I just put it on). Its principal use was reflexive (let me put this clearly): about two thirds of the verbalising sense (24/32) were reflexive, and half of the total use (24/50). One third of verbalising uses of PUT were instances where the speaker referred to was not a participant in the speech situation (propositional surface of moral discourse as Blackburn would put it).
In the discourse reflexive sense of PUT, two patterns emerged, both meaning ‗formulate', or ‗verbalise in a certain way'. One pattern (Example 1), to PUT it + adv., was the relatively idiomatic prefacing formula to PUT it bluntly/simply/ briefly/mildly: (1) er erm so er or to put it positively the non-naturalist has to claim sort of, dwells, in women's suffering, to put it blunt if you put it would clear my thinking but er put it simple erm our natural language you plan to do @@ to put it to ask in a nice way well sometimes On the whole, the use is in line with Standard English, but the form is occasionally something of an approximation, for example adjective forms (blunt, simple) being used instead of adverb forms.
The other pattern of PUT also fitted into a fairly common phraseological pattern let me/let us + V comm , as can be seen in Example 2.
(2) are i after the forum] put it let me put it this way if you buy me about their activities and let me put it this clearly NGO's some analysing a text yeah let's put it very simply what ever the text is by organisational knowledge creation let's put it that way and use i think that well let's put it as a question can there be a quality A third observation from these searches was that discourse reflexive use of PUT co-occurred with mitigating expressions, such as modals (could, might) and other hedges (a bit) and if -clauses: (3) yeah i think we cou-you could put it that way but that's more like necessary connection erm, or you might put it er use a a metaphor remember but and I'm also going to put it a bit more sharply than and society or i-if i, put it more clearly where can we see this blunt if you put it a bit bluntly so that's what i find problematic that he, he cannot, well if you put it blunt to have sex with bess This tendency of metadiscourse to collocate with hedges has been noted before (e.g. Mauranen 2001Mauranen , 2004. I have called it ‗discourse collocation', because this is a co-occurrence tendency between two discourse phenomena, not between specifiable lexical items (or grammatical items, as in colligation). As the earlier research was carried out on data from native speakers of American English (the MICASE corpus, http://lw.lsa.umich.edu/eli/micase/index.htm), the present observation suggests that the connection is more general. Although research on written text usually assumes metadiscourse to be an act of consideration towards the reader, its role is more complex in dialogue. Metadiscourse implies an imposition of the speaker's perspective on the discourse, and in so doing reduces the negotiability of the dialogic perspective. Hedging mitigates this, as one of its important uses is to open up for negotiation the meanings made in the discourse, or, in slightly different terms, to indicate epistemic openness (Mauranen 1997). The combination thus serves to restore the balance between expressing speaker perspective and keeping it negotiable. That this combination appears in ELF suggests it is not confined to Anglo-American discourse conventions but is a more fundamental feature of discourse.
It has been customary to associate metadiscourse primarily with first person pronouns. However, in this data put it co-occurred half of the time with impersonal reference (generic you) or inclusive we, which are typical other-involving expressions (let's put it /you could put it). The rest divided equally between speaker (I) and hearer (you) references. This suggests again that a dialogic situation highlights different aspects of metadiscourse as compared to monologic communication, especially writing.
The other verb form under scrutiny, SAY, shows a similar tendency in person reference: about half of the instances (32/67) refer to the interlocutor, half (35/67) to the speaker. It appears that discourse reflexivity is not so entirely speaker-oriented as we are used to thinking; dialogic interaction brings this out as the role of the interlocutor gains prominence relative to monologue.
Since other-oriented reflexivity has been much less investigated than speaker-oriented reflexivity, I look at SAY in the other-oriented function. The basic metadiscourse pattern is you + BE + saying, with three main functions: ‗clarification', ‗interpretation', and ‗springboard'.

CLARIFICATION
The speaker wants the previous speaker to repeat, elucidate or confirm what he or she has said (Example 4). Elucidation requests are typically questions, to which the interlocutor is under obligation to respond. Responses can confirm or dispute the interpretation, or offer an elaboration of the original intended meaning.
(4) two societies and you say that you am i right that you saying that you kind of want to way of expressing it, (differs) i-is you are you are you saying that er, the imagery @yeah or@ joo [@@] [@@]. er are you are you saying erm that this particle functions could you just show that last example once more were you saying that in this the you that's confusing This is a means of ascertaining that the shared experience being coconstructed in the interaction is indeed shared (5). At the same time it keeps the interpretation negotiable; it is easy even if not necessary for the interlocutor to confirm or dispute the offered interpretation. The speaker can also express the interpretation more tentatively (at least I took you to be saying), in which case the invitation to respond is stronger.
(5) supervenience using quantification over possible worlds so you're saying things like er yeah for any objects X and Y for any worlds W and B if X is the same answer as for the first case so you're saying that there you cannot really make the difference between the two because and in a way this also then covers probabilistic systems so you're saying that probabilistic system is fair at least under some er [a reasonable another thing that that i at least took you to be saying was that that er literary works are individually different and to the best of my (xx) you were saying to the best of my i think er there isn't any any any difference SPRINGBOARD The speaker paraphrases the previous speaker's meaning as a point of departure for a new direction in the discussion. This was the largest group of you + BE + saying. In these cases, see (6) below, the speaker does not indicate that the interpretation is open to negotiation but continues to develop his or her own agenda. No response, confirmatory or otherwise, is thus expected of the interlocutor concerning the correctness of the interpretation, and in fact it is not easy to do this without making an intervention.
part of what you are saying that indigenous knowledge and, and scientific er knowledge, but i have if what you are saying about crystallisation is true or not you could use the D-S-C for that when you were saying you know maybe er i would have wanted to talk about western what you're saying proves that tanzania is pretty much a stable country and then time i mean it's true what you're saying that er the the home is is very important talking about er place ballets well you're saying one possibility is definitely cornell realism another possibility is erm oil hydraulics that's just like you are saying water hydraulics, but er is it just possible to change the oil with water forums er as diaspora outside fiji er you were s-er saying about digital divide between the disadvantaged and the so i was just wondering The examples show that You + BE + saying plays an important role in ongoing argumentative dialogue. It contributes to creating a coherent, interrelated discourse by relating speakers' arguments to each other. Other-reference is integrated into the argumentation as a piece of shared understanding, to which the next stage of the argument is anchored. The speaker may be more tentative about interpreting the interlocutor, in which case the interpretation is open to negotiation (the clarification and interpretation types), or simply present it as given (the springboard type), when it serves as a more straightforward point of departure for the next stage. What we see here, then, is how reflexivity helps to co-construct discourse as a joint product between speakers.

Focus on the particular: three seminars
The corpus examples have already suggested that as we move our attention to dialogue, discourse reflexivity reveals features that have not been observed in the field earlier. To push the boundaries back a little further still, I would now like to shift the angle to individual discussions and the ways in which discourse reflexivity works in particular contexts.
The events under scrutiny are ‗seminars'. These are comparatively transient communities of practice, set up for one or two terms. As multiparty discussions they are interesting from the perspective of discourse reflexivity, involving as they do frequent turn shifts and interaction management, as well as managing the discourse as text. They are relatively spontaneous and open-ended events, even though within limits: they operate within layered structural frames from the institution and from practices that the groups have established for themselves. Within those limits, the discussion is co-constructed by the participants fairly freely, so that it can take different directions, and there is a strong element of unpredictability, as in any discussion or conversation.
There are thus commonalities in seminars. A seminar is a ‗chain event' in these study programmes, a serial event that gets performed by the participants in repeated encounters. Some of its norms and practices are imposed upon it by the institution that contains and validates it, others the group can decide for itself, and yet others are tacitly agreed upon, and evolve in the course of events. The language of instruction and communication is institutionally determined, while things like appointing a chairperson, the possibility or desirability of interruptions during presentations, or first-name use may be collectively decided. More subtle norms of language use tend to be tacit, a matter of linguistic selfregulation within the group.
I chose three seminars, as already explained in Section 4. The discussions in medical and women's studies constitute the entire event for which certain written texts had been read as preparation, while in political science the discussion section was preceded by an oral presentation not included in the analysis.
Despite being called seminars, the three events are fairly different, which also shows in the ways in which they employ discourse reflexivity. Some variability is of course to be expected within a genre, despite its uniform labelling. Disciplinary differences alone set up expectations of distinctions, which however cannot be pursued here, as we only had one seminar from each discipline. Cultural differences are often relevant but do not transpire as particularly evident or relevant here, because these events are international even though they take place in the same national matrix culture.
The seminars discussed below are all from a late stage in the series. Although we thus have no access to the initial stages of negotiating linguistic norms, it is nevertheless unlikely that reflexive discourse gets explicitly discussed, because it is the means, not the object of negotiation in itself, and rarely something speakers are aware of. In brief, the sessions here are all late-stage single sessions in a seminar chain. The speakers are familiar with each other and group practices are well established.

Medical seminar: 'understanding facts'
The medical seminar consists of a discussion around a topic for which there had been a set text for students to read. One of the student participants acts as a discussion chair, while the seminar leader, a senior staff member, remains in the wings most of the time. She makes an appearance every once in a while to direct the discussion towards points of her choice, and her main role seems to be to answer questions students have not found an answer to. The event seems to unfold according to a routine; this being the final session in the seminar series, the participants appear to be familiar with the procedure and go through the motions smoothly. The participants are actively involved, take turns, overlap, and engage in backchannelling and laughter.
Discourse reflexivity is confined to a small role. The discussion moves on from stage to stage by simple prompts such as and then and what about, and direct questions from the chair, as in Example (7) The discussion revolves essentially around the facts at hand, with brief, even minimal insertions of metadiscourse elements (I mean): (9) S7: where does the ascites come from . when you're big like this S2: i think it's the same thing when you're having hypertension so you have er your legs are swollen because the fluid goes to extra , cellular space S6: but is it the is the same mechanism because i've been thinking about this S2: @@ i don't know [i just thought it] @@ S6: [because i mean] when when the the legs are swollen isn't it because of the protein [lack of protein protein lack (xx)] S5: [yeah well probably (xx)] in this case because albu-albumin (S6: yes) is going is increasing so probably a lot of liquids (S6: mhm) are going out S6: but is it is it same with the ascites Reflexivity thus appears mainly in a clarificatory role, to indicate that something was unclear or hard to understand, and to elicit elucidation (do you mean): (10) S6: it's the protein complex take in in the first phase takes take the cholesterol, and then it goes to the, S3: do you mean the kilo (S6: yeah) micrones SS: @@ S6: yeah [(xx)] In addition, very occasional negotiations about language took place, and these concerned essentially terms or subtechnical vocabulary.
(11) S5: yeah it is specific and erm, there are ways to control the alcohol (mhm) addictation do you say S6: yes yes, it will go down in two or three weeks, when [you] S2: [so what] was the thing you said, C-S7: C-D-T SS: C-D-T S6: it's, carbohydrate deficient transaminase or something, C-D-T Overall, the main weight in the discussion revolved around understanding the content. The orientation of the participants and the whole verbal activity is very much to the discourse-external, physical world of medical phenomena, together with problems and professional practices related to them. These constitute the central referents talked about, and the topics of questions and clarification requests. The purpose of the discussion appeared to be getting the facts right, thereby developing professional skills. In brief, the main point of this seminar could be described as ‗understanding facts'.

Women's studies: 'Talking and sharing'.
For the second seminar, the chair, who is a staff member teaching this course, has asked the students to write something for her and those texts laid the foundation for the discussion topic of the day. The chair has selected the discussion topics from these and moves on along her pre-set agenda. She makes frequent reference to the submitted texts, earlier discussions in the group, and the general idea of the course. In this way, she draws on other discourses, which she presents as relevant to the topics at hand and engages in a lot of preparatory work for each topic or subtopic she introduces. As in the medical seminar, the participants seem to be familiar with the procedure, and actively involved in turn-taking, backchannelling, laughter, and cooperation in cases of difficulty. In contrast to the first seminar, speaker contributions are longer, they report more personal experiences, more evaluation of speakers' own and others' contributions. Discourse reflexivity is markedly more common than in the medical seminar.
In this event, discourse reflexivity is not merely more frequent, but also used in more ways than in the first seminar. This is illustrated in the following example (12), where S1 starts out by imposing order on the discourse (then I wanted to talk about), followed by an offer of the floor to S12 (would you like to explain). S12 starts with a distancing preface (well), upon which S1 starts to negotiate, modifying her offer (or would you think about it…). S12 now gives grounds for declining the offer to take the floor (well, I don't know how to explain it), whereupon S1 backtracks and starts producing herself the response that complies with her initial request.
(12) S1: … then i wanted to talk about some sort of expressions that we use . that are defined or thought of as positive, for example there's this <FOREIGN> hyvä jätkä </FOREIGN> would you like to explain it for us S12: [well] S1: [or would] you think about it and then i fill in S12: mhm well, i don't know how to explain it so S1: yeah [well] (S12: [mhm]) perhaps er . if other people that you're really er, you're a good good mate, or a good good friend or you do things you play well or As in the medical seminar, participants negotiated their accumulating shared understanding of the topic, but this was achieved with much more verbalising in women's studies (13): (13) S2: what er i-i (S1: mhm ) i can't understand what what what does it mean what is er real differences between women and men are is it that they were asking for er er is it the gender difference er er cultural difference or a biological difference was is this the question In contrast to the medical seminar, transitions were largely made explicitly and elaborately. The tendency of metadiscourse to collocate with mitigation is seen in Example (14).
(14) S1: [well i] i'll start with, some of my thoughts (xx) last session, try to summarise a little bit, what we discussed and, what you wrote about . this er somehow it became evident that er this categoration /…/ and actually er the most service occupations are such occupations . we are going to talk about emotional work… Speakers used discourse reflexivity to move back and forth in the discussion, creating cohesive links with their own earlier positions (as in 15), or those of others (as in 16). In addition to coherence in the discourse, such links may well contribute to group cohesion among the participants.
(15) S1: but do you think there's any er presseer, pressure from the society S8: yeah, definitely, at least in my experience it is yeah but then again i like i said i come from this huge family so many small children that it's just facing me every day almost so but maybe it's different for other people (16) S1: … so i'm agagain referring to your story when really talked about the division of labour actually in the household that (S3: mhm-hm) the male went into the coal mine and then the women did all the chores at home er there's this strict division there, and also, you talked about this sort of that for women this traditional way… Terms and concepts were discussed and shared understanding was sought, but the discussion was not oriented solely to discovering or establishing a predefined correctness. The terms commonly used in the field were seen as negotiable and open to redefinition as a result of the group discussion. The distinction between specialist terms and general language was thus blurred, and individuals were seen as being entitled to moulding terminological usage.
(17) S1: [well] what is radical because we don't have any radical feminists in finland so it's no what's the, what's the definition of a [radical] S4: [no] i i'm not sure if if radical is the is the is the right word maybe i don't know er b-women who are ah arguing in every situation er with with all the gender Openly evaluative remarks were often made in connection with metadiscourse (18), apparently in recognition of the complexity or sensitiveness of certain topics. Language was thus crucially implicated in conceptual difficulty, as questions, answers, and just talking. This was quite different from the medical seminar, where difficulty was linked to understanding and conceptualising external-world phenomena, and terms were treated as referential items with an uncomplicated relationship to difficulties in the language-external world.
(18) S1: …and when er talking about this gender difference thing it's a bit difficult, when in Finnish we…well, and then when we talk about gender difference it's usually talked in terms of cultural differences between this femininity and …it's more useful to concentrate on the latter one and we shall we talk about gender difference that's the main focus, and i was a bit er not well not amazed but a bit er, baffled when some of you when you wrote … and that's a really difficult one, cause i i don't think at least i don't have a such a fon-mhm answer to tell what's this ... In all, the women's studies seminar event was characterised by frequent and versatile use of metadiscourse. As distinct from the medical seminar, there was a marked orientation to the situation at hand, the language used, and the participants themselves. A central ingredient was making reference to personal experiences, and the discussion also oriented to probing participants' views, interpretations, conceptualisations and feelings. These were related to language as well as the topics at hand, and participants presented their own contributions in relation to those of others, which obviously created demand for discourse reflexivity. Talking appeared to be valued in itself. The seminar chair did a good deal of the talking, with long prefaces to topic transitions, and many invitations for others to speak. She also brought other relevant discourses to play in the discussion, which seemed to enhance the sense of sharing perspectives. Hedging of metadiscursive imposition served to level out or soften asymmetries of power. These discourse strategies also appeared to be intended to offset some of the inherent power imbalance in a seminar discussion, where ultimately the teacher assesses the contributions of the others. In brief, this seminar could be characterised as ‗talking and sharing'.

Political science: 'Constructing arguments'
The third session consisted of discussions around two presentations by students, which were each followed by a discussion. The two discussion sections are considered here without the presentations. This seminar resembles the women's studies seminar in that the chairperson plays a strong role in selecting the topics as well as organising the flow of the discussion, and also in that metadiscourse played a central role in negotiating meanings and in moving the discourse on.
Transitions from one stage of the discussion to another are largely in the hands of the seminar leader who acts as the chair, and although they are often brief and formulaic (okay; and then) as in the medical seminar, the chair also uses more complex transitions to organise the discussion (19).
(19) S1: er okay before we go to the next topic, i i think that. in a way the question <NAME> made what made you study or be-become interested on this issue it is a relevant question cause this your topic leads us a bit further to more general (S2:yeah) discussion about human rights or in general whether we can… The seminar chair took care of the overall structuring of the discussion, as in the women's studies seminar, but in this case student participants were also keen to initiate topics. Sometimes they took the floor by just announcing the topic (about Stalin), but most of the time they used advance labelling of their speech acts, prospecting ahead and making it clear in which light their turn was to be taken (a couple of questions; can I ask you a difficult question; just a comment; I have couple of comments reactions to this; just wondering). Discourse reflexivity was employed to achieve complex topic transitions, for instance indicating awareness of turn-taking norms while introducing a new subtopic: (20) S2: … and like, yeah it's it's another thing in sorry @@ i i just go on but it's another interesting thing it was like that in estonian press… In referring to and resuming topics from earlier stages of the discussion, discourse reflexivity was also made use of, involving both speakers' own contributions (my question based on minority; as I told) and those of others (what you said; what you also said; the problems you were discussing). This created an impression of a coherent argumentative discussion, where speakers were jointly engaged. In this way, the discussion bore resemblance to exchanges following conference presentations, where the presenter's line of argument gives rise to questions, alternative views, and new points (see 21 below). Unlike conference discussion sections, though, this discussion was much longer (over half an hour) and could thus probe the issues in more depth.
(21) S3: of course a couple of questions erm this citizenship how much does it influence the people are they the russians allowed officially to work and everything S2: no yeah i can explain it's erm yeah sorry i i didn't probably didn't mention @@ it's a if you don't have a citizenship you can't vote Discussion contributions were evaluated in a number of ways, some with metadiscursive expressions (it's a good question; it is a relevant question). The chair's evaluative comments on presentations or methods often seemed pedagogically motivated and intended for the whole group, as e.g. in (22), where we can also see the mitigating effect apparently sought by the hedged (just some minor things) forewarning that something negative may follow (nothing to criticize you but). This hedge and metadiscourse combination acts as an advance notification of a possible face threat (cf. Brown and Levinson 1987).
(22) S1: … er, then just some minor things nothing to criticize you but just you give a good excuse to to mention these things in finland finns are very proud about… Participants also negotiated the functions or purposes of each other's arguments and points: (23) S2: no, you mean why did i refer to that i was more like when i was defining why… The principal role of discourse reflexivity in this seminar seemed to be to elucidate the co-construction of arguments. What seemed to be at stake were interpretations of past or present political situations and stances, which the participants were negotiating among themselves (24).
(24) S1: but from this point of view what you said this is a minor thing about paasikivi being against finnish membership in UN i wouldn't say that he was against it was just a question a matter of the time [and then] S1 here evaluates S7's interpretation of a past political stance. He uses discourse reflexivity to make an other-reference to S7's interpretation (what you said… about Paasikivi being…), gives an indication of how his contribution is to be taken (this is a minor thing) and then from this springboard he launches his own interpretation with a somewhat tentative metadiscursive preface.
The last extract (25) illustrates several typical functions of discourse reflexivity in a small space. Here it prefaces discussion points (just a comment) and questions (wouldn't you say), signals clarifying (I mean), backtracking and self-rephrasing (let me correct myself) and negotiating terms (call it): (25) S5: just a comment wouldn't you say that is a situation that is quite prevalent in in a number of countries in europe i mean that happens if you if you only talking about er or let me correct myself not only but you are talking about sort of bureaucratic repression er that kind of repression call it repression it goes on for example in france… Reflexivity was used for preparatory work, relating arguments from the speaker, other participants, and third parties to each other. In this way, a network of arguments was being constructed, and the orientation was primarily towards arguments for and against certain interpretations of the world. Discourse reflexivity was also employed in offsetting power inequality in the manner of the women's studies seminar. Unlike the women's studies event, this seminar was more oriented to the arguments being developed than the personal experiences of participants. The main orientation in political science also contrasted with the fact-orientation of the medical seminar, as the focal points of discussion revolved around interpretations of facts rather than the facts themselves. These two seminars nevertheless showed certain affinities in putting the emphasis on non-personal events in the external world and downplaying personal experience and emotional responses that surfaced more in women's studies.
Three ostensibly parallel events chosen by external criteria revealed that there is much to be discovered about variability in the use of metadiscourse in spoken interaction. Patterns of metadiscourse use seem to be involved in constructing different conceptualisations of the world and associated epistemic beliefs. As variation in written texts has been known and much studied for a long time, dialogic speech deserves similar attention.

Conclusion
This paper has investigated discourse reflexivity in spoken dialogue. Three main points have emerged from the analyses. One is that discourse reflexivity appears to be crucial to successful spoken interaction: it enables fluent management of interaction in even complex multi-party discussions, and promotes communicative clarity and precision. In other words, it seems to confer the kinds of advantages to communication that were outlined at the beginning of this paper. Even though its amount varies, it seems ubiquitous. That this is so when English is used as a lingua franca among non-native speakers lends further support to the pivotal role of discourse reflexivity; it is not a feature of Anglo-American culture, or directly a function of language proficiency. We can postulate that it is a strong candidate for being a discourse universal (Mauranen 2003b), or in Hunston and Francis's (1999) terms, a ‗necessary feature of language'. Speakers will find means of expressing reflexive functions in discourse because these constitute critical elements of fluent conversation even if the particular expressions used do not match some prescribed standards.
The second major finding is that dialogue brings out new facets and different emphases on the functions of self-referential language in communication as compared to written text analysis. A notable phenomenon was the amount of other-oriented metadiscourse. This deviates from earlier perceptions of metadiscourse, and also of speakers' orientation in speaking. For instance Schiffrin (1987: 124) suggests that -speakers' monitoring of their own talk is more finely attuned than their monitoring of others' talk‖, and while it may of course be possible, the present results certainly cast doubt on such a generalisation. At least in argumentative discussion other-oriented reflexivity seems particularly salient.
The third point I wish to emphasise here is that the present data again supported the tendency of discourse reflexivity and hedging to collocate in ‗discourse collocation'. This has a bearing on our conceptualisation of metadiscourse. If we opt for a very broad, embracing notion of metadiscourse, we risk losing sight of its collocability and interaction with other discourse phenomena. The consequence of this is that our analytical tools lose some of their sharpness. I take this therefore as support for my initial argument that discourse reflexivity is a crucial aspect of human communication, which deserves to be studied in its own right.
All these findings have been scratches on the surface: there is a wealth of expressions in self-referential spoken language, which could be studied for patterning and functional diversity. Moreover, the settings in the present database were very similar in generic terms, as they were all academic dialogues. We could assume that many of the factors that have been found relevant to genre differences in written text reflexivity may be relevant to speech as well. The need for precision and explicitness in academic discourse along with its necessarily fairly complex organisation motivate a high level of metadiscourse. However, in any dialogic genre the need to manage spoken interaction in real time may assign discourse reflexivity a much more important role than tends to be found in non-academic written prose.
The contextual diversity of metadiscourse was remarkable in the light of the three seminar sessions. It is clear even from this small-scale study that the roles of professional or academic and national culture need to be kept separate, and the same goes for native languages and national cultures. To account for variability of the kind found here, new parameters ought to be taken on board, not assuming that first language, native culture, or disciplinary domain are omnirelevant categories even in academic settings. There is space for more work on contextual parameters which have a bearing on metadiscourse. Most importantly, new models of metadiscourse must take the dialogic perspective of interaction seriously on board.