4.3.1 Interjections

Interjections are conventionalized verbal forms that may constitute an utterance of their own, but may also represent an “entry” into a unit of talk and thus be followed by more talk. They do not enter into formal relationships with other lexemes or syntactic units, and do not tend to have lexical content (Clark & Fox Tree Reference Clark and Fox Tree2002: 76). The production of interjections is motivated by the general communicative context as they represent an instantaneous and often unmonitored reaction to a verbal or nonverbal stimulus from the environment. They are produced in a quasi-automatic way as formally autonomous units.

Grammatically, interjections are at the boundary between lexical and nonlexical, merely phonological expressions of states of minds or response cries in terms of Goffman (Reference Goffman1978). While some authors do not regard interjections as part of the language system, others do attribute lexeme- or word-status to them. Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985: 853), for instance, describe interjections as “purely emotive words which do not enter into syntactic relations.” Ameka (Reference Ameka1992), who argues that interjections have a semantic structure, distinguishes two classes of interjections: primary interjections, which have a lexical counterpart (e.g. hell, Jesus, shit), and secondary interjections, which have no lexical counterparts and cannot be used in any other sense than as an interjection (e.g. oh, oops, sh, wow). He further distinguishes two classes of interjections: those indicating a cognitive state or a change in knowledge state (e.g. oh, ah), and those expressing emotive stance such as delight, excitement, disgust or feeling of pain (e.g. wow, gosh, ouch). Wierzbicka (Reference Wierzbicka1992) adds a third type, namely interjections expressing desire or volition (e.g. shh, shoo, ahem) serving, for example, as requests for attention. Depending on how interjections are conceptualized, more routine functions can be distinguished, e.g. expressing greetings and farewell (e.g. hello, hi, bye), degrees of accepting a message (weak: hm, uh-uh, stronger: yo, yep), or confirming the validity of an assertion (bingo), and many interjections have parallel uses. While those authors that argue from a conceptualist point of view consider interjections as forms with semantic content (e.g. Ameka Reference Ameka1992; Wierzbicka Reference Wierzbicka1992; Wilkins Reference Wilkins1992: 119), others argue that the meaning of interjections cannot be defined in semantic terms; they can only be defined by the conventional practices they are used for (Clark & Fox Tree Reference Clark and Fox Tree2002: 77).

Some authors include uh and uhm (also er), which for others are merely “filled pauses,” in the category of interjections, e.g. Clark & Fox Tree (Reference Clark and Fox Tree2002). This is certainly plausible given that they are, like other interjections, produced as units of their own and provide insight into the current cognitive state of the speaker, documenting an ongoing mental planning activity. Basically, the function of uh and uhm is to project a delay in ongoing speech and to point forward to upcoming speech. There is some debate on whether or not the two interjections are freely interchangeable: based on evidence from larger corpora of spontaneous speech, Clark & Fox Tree (Reference Clark and Fox Tree2002) propose that uh and uhm signal different kinds of delays: while uh is used “to announce the initiation [ … ] of what is expected to be a minor delay,” uhm is used “to announce the initiation [ … ] of what is expected to be a major delay in speaking” (Clark & Fox Tree Reference Clark and Fox Tree2002: 79). These findings led the authors to the hypothesis that uh and uhm are planned like other words (Clark & Fox Tree Reference Clark and Fox Tree2002: 75). However, in her corpus-based study of hesitation phenomena within and at the boundaries of prefabricated units or “chunks,” Schneider (Reference Schneider2014: 60–62) provides counterevidence against this assumption: in her data, there are no significant differences in the mean length of pauses after uh and uhm, which in both cases are between 0.49 and 0.5 seconds on average. This would mean that their meanings are unlikely to differ.

In any case, both studies make clear that uh and uhm are expressions of speech planning activities, signaling shorter or longer hesitations, which shows that speakers are able to monitor the location of an imminent delay. To simply somewhat, the delays indicated by uh and uhm are usually caused by the need for stalling for time in order to reflect on the continuation of talk. The use is also motivated by marking discourse boundaries and signaling readiness to adopt or keep the conversational floor. Since speech planning can be assumed to be most difficult at the beginning of a unit of talk, where speakers have to plan the content of a message and select a structural pattern from a variety of possible syntactic formats, it would make sense to find a much larger proportion of uh and uhm in the initial field than at other points in time in the production of a unit of talk. This is indeed what Clark & Fox Tree (Reference Clark and Fox Tree2002: 94) observed, based on an empirical corpus-based study on the frequency of use of these interjections in different positions within and at the “boundaries” of an intonation unit (defined as a stretch of speech produced under a single intonation contour): the usage rate was highest at the initial boundary of an intonation unit, where also most of the prosodically prolonged realizations of these interjections were produced (roughly 50 percent of all prolonged uses, the rest occurring after the first word or later within a unit of talk), which indicates that delays tend to be longer at the beginning than at any other point in time in the production of a unit of talk. The use of uh and uhm attests that speakers keep track of how an upcoming unit is delivered (relating to performance, e.g. production with a delay), which is a task that is clearly within the confines of macrogrammar.

Clark & Fox Tree’s (Reference Clark and Fox Tree2002) observations on unit-of-talk-initial uses of uhm are congruent with my own observations on the use of uh and uhm in the ICE-GB: uhm is by far more frequent than uh in the initial field, and uhm it is more frequent in the initial field than in any other time slot in the production of a unit of talk. Based on a study of the occurrence of uh and uhm in the first ten conversations of the ICE-GB (S1A-001–S1A-010), I conducted a small-scale empirical study of the uses of these interjections in four contexts: in the initial field, in the final field, and at two different points in time at which uh/uhm occurred within the continuing phase, one being at a juncture from one syntactic segment to another, the other one at unspecific points within the production of a microgrammatical unit, e.g. after a subject or an auxiliary, in which uh/uhm clearly interrupt a segment-internal hierarchical relation. (26a) illustrates the use of uhm in the initial and final field within the same unit of talk. In (26b), uhm occurs in the transition zone from one (microgrammatical) syntactic segment to another: speaker B first produces a statement before he shifts to a major point deriving from it, which is introduced by so.

1. 1. a.

      129	A:	wh- what do you personally get ou’ of (.) ↓this
      		particular dance (.)
      130	B:	u:hm (3.1) it- (2.6) I find that I:  have to  (. ) be  more
      		honest about the way that I work. (.) uhm (.)
      131		[ … ]                                                          [ICE-GB S1A-002]

   2. b.

      136	A:	do you want to continue into the performance.
      137	B:	yeah I will=
      138		=I’m  graduating  in  JUNE  uhm  ( . )  so  (1.8  )   it’s  (  .  ) given   me some direction alREAdy.                     [ICE-GB S1A-002]

Note that in the initial field, uh/uhm may occur together with other macrogrammatical expressions, e.g. so uhm or well uhm. A context in which uhm is produced within an emergent microgrammatical unit is illustrated in (27).

1. (27)

   015	A:	NO: because I plan to be out of phonetics as quickly as
   		possible.   (.)    [°really°. ]
   016	B:	[yes,       ]
   		but it might well help if you’ve uhm actually done some.
   [ICE-GB S1A-008]

In (27), uhm occurs after the subject–auxiliary sequence you’ve, which makes a lexical verb and further arguments expectable and thus occurs during the production of forms that are involved in hierarchical relations. Such uses are least specifiable in terms of regularity.

The figures in Table 4.2 indicate the occurrence of uh and uhm in different time slots during the production of units of talk. Instances in which uh or uhm occurred two or more times in a row in the same time slot within a single unit of talk (e.g. uhm (.) uhm) were counted as one token. Based on this procedure, the ten conversations (=20,978 words) included into the empirical analysis yielded an overall number of 115 occurrences of uh and 315 occurrences of uhm. In the sample conversations uh was thus by far less frequent than uhm.

The data show that uh and uhm occur preferably in the initial field, and thus serve as means to prepare or announce the production of an upcoming unit of talk. The functions range from getting the addressee’s attention, indicating turn uptake and preparing the ground for an upcoming turn, and delaying the production of a message. Delays have different motivations, such as avoiding overlap with a prior or another self-selected speaker or stalling for time in order to carry out conceptual and structural planning of the upcoming unit. The tendency toward uses in the initial field is clearer with uhm than with uh. One explanation for this finding could be that in the initial phase speakers require, or automatically produce, phonologically more salient indicators of turn-uptake or continuation than at other places in an emergent utterance, given that in this phase speakers are subjected to turn-taking mechanisms and thus need to compete for turn-uptake or continuation. Since uhm is phonologically more salient than uh the first serves the task of indicating the speaker’s intention to adopt or keep the role of the speaker better than the second. Another explanation is that uhm allows for a somewhat longer planning time of upcoming talk, given that it provides more phonological material for prosodic lengthening (both the vocalic nucleus and the bilabial nasal lend themselves for lengthening). The beginning of a unit of talk represents a phase of comparatively high cognitive stress since speakers need to assemble linguistic forms into a proposition and choose a particular syntactic format. Thus, it can be expected that it is particularly in the initial phase that speakers, once they have self-selected for the upcoming turn but not yet completed conceptual and syntactic planning, are somewhat more often caught in a situation that requires a more or less extended time span for carrying out planning activities. This assumption would be congruent with Clark & Fox Tree’s (Reference Clark and Fox Tree2002) observation that uhm signals a more extended delay than uh even though, as discussed above, this finding is not undisputed. Note that uh/uhm are also frequent in unspecified time slots (“anywhere”), in which their use can be related to ongoing planning activities in emergent talk caused by, for example, the temporal unavailability of a lexical expression in the speaker’s mind, careful selection of an adequate expression, or a brief suspension of continuation due to an idea or activity that competes with the further development of a current thought or activity. The lowest frequency of occurrence of uh and uhm is in the final field, which is plausible if we assume that at the potential end of a unit of talk speakers need no longer select from a broad range of possible continuations as they have (almost) completed the expression of a focus of consciousness. Monitoring possible continuation would then become relevant only at the beginning of a new unit of talk, and thus in the initial field of a subsequent unit, unless speakers decide to expand an emergent unit of talk beyond a potential point of completion, or to foreshadow continuation, in which case uh/uhm may occur as planning markers in the final field.

To conclude, interjections in the initial field serve different functions. On the cognitive level, they may indicate an ongoing planning activity, as with uh/uhm. On the discourse-structural level, they continue prior talk since they are per se produced as a reaction to some prior stimulus, e.g. surprise (oh) or disgust (ugh). In terms of the organization of speaker–addressee relationship, they may express an emotional (e.g. wow) or knowledge-based reaction (e.g. ah) to what was said before, serve the expression of a directive speech act (sh, shoo), or indicate that the speaker is taking over the conversational floor or continuing a turn. Since the indication of cognitive or emotive stance makes an account expectable, interjections raise expectations on further talk in which the speaker explicates his/her stance. In this sense, interjections serve the organization of conversational interaction in that they coordinate the speaker’s roles between the participants by legitimizing turn uptake.

4.3.2 Initial Adverbs

Some expressions in the initial field “look like” adverbs, from which they typically derive. However, strictly speaking they must be distinguished from “genuine” adverbs, that is, from forms that are integrated into hierarchical relationships with other forms. The group includes lexemes like of course, right, really, or honestly, which are classified in various ways in the literature, e.g. as disjuncts (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 613, 648), stance adverbials (Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 133), sentence adverbs (Brinton & Traugott Reference Brinton and Traugott2005: 139), or pragmatic markers (Fraser Reference Fraser1996; Traugott [Reference Traugott, Brems, Ghesquière and Van de Velde2012] for surely and no doubt). While many authors tend to regard these lexemes as adverbs and thus as clausal constituents, in the present framework they are classified as macrogrammatical units for the following reasons. First, they are syntactically unintegrated in the sense that in the initial field they lack integration into a hierarchical relation with another form. They do not restrict or modify the meaning of a form or unit within a microgrammatical unit, e.g. that of the predicate, and they are not assigned their morphosyntactic or semantic properties from any other constituent. Moreover, in the initial field they lack important features that characterize prototypical adverbs, e.g. positional flexibility or possible premodification or intensification by other adverbs. An analysis as clause adverbs is inappropriate as they need not be followed by a clausal unit, but can introduce a unit of any syntactic format and length. Secondly, they differ from adverbs in the sentence-grammatical sense in their semantico–pragmatic scope potential: while adverbs have scope over a constituent of a phrase or a clause, adverb-like expressions in the initial field typically have a wider scope, which goes beyond the meaning of the upcoming unit as it extends to the communicative situation, encompassing, for example, illocutionary force, speaker attitude, or speaker–addressee relation.

Consider, for instance, the problem of so-called illocutionary adverbials such as unit-of-talk-initial (also -final) frankly or seriously, which encode a lexical concept but are, strictly speaking, not part of the proposition expressed in a microgrammatical unit, non-truth conditional, and not involved in a “basic-level explicature” in terms of Sperber & Wilson (Reference Sperber and Wilson1995). Yet, they are associated with a particular conceptual meaning, providing a “higher-level explicature,” that is a conceptual representation in which the proposition is embedded under a propositional attitude and illocutionary force (Sperber & Wilson Reference Sperber and Wilson1995: 182; Carston Reference Carston, Davis and Gillon2004). On a gradient from microgrammar to macrogrammar, they are clearly closer to the macrogrammatical end.

The use of really in the initial field is perhaps the clearest example for a microgrammatical element that has developed parallel uses within the domain of macrogrammar. As a form involved in microgrammatical relationships, really typically intensifies the meaning of an adjective (e.g. really nice) or a verb (e.g. It really brings you closer to the truth), and is thus part of a hierarchical relationship with the form in its scope, which is structurally higher-ranked. As an expression in the initial field, however, it is not included in a binary morphosyntactic and semantic relationship with another form. Consider the use of really in (28), where A and B talk about possible ways to refurnish B’s bedroom. A is advising B on the color of the curtains over B’s bed; later (line 37) B expands the topic to ask for B’s opinion on a wallpaper color she has chosen.

1. (28)

   31	A:	no but it’ll be it’ll be if you decide to have blue it’ll be the
   		blue that you see.
   32	B:	yes.
   33	A:	see what I mean.
   34		u::h let me get you a photograph. (.)
   35		it will just be the outside little bit.
   36 →		really you wouldn’t see much at all. (.) uhm
   37	B:	so WHAT if I choose ↓this wallpaper, (.) uhm what shall we do
   		with this? (.)
   [ICE-GB S1A-086]

In the initial field, really forms a unit of its own that is not part of the network of hierarchical relations within the following microgrammatical (clausal) unit you wouldn’t see much at all. It serves as a text-organizing device in the sense that the speaker foreshadows a reinforcement of her main argument, which is that the blue curtains would not let through much light from the outside. The meaning is not anchored in the structure of the following unit, but refers to the situation of discourse: really is used in a sequential slot in which A has already expressed her opinion (see what I mean, line 33), and continues fleshing it out. Really connects two subsequent units of discourse (the units of talk in lines 31–35 and the unit produced in line 36) and thus serves as a means to signal a coherent continuation of talk in that it marks the upcoming unit as arising out of the immediately preceding context and that the preceding talk is the interpretive anchor for its comprehension. Moreover, it underlines the speaker’s strong commitment toward what she is about to say and is thus highly subjectified. The communicative effect is that it focuses the attention of the addressee at the upcoming unit in that it projects an idea that the speaker fully endorses, and thus serves as a persuasive strategy on the interpersonal level.

Adverb-like elements in the initial field express meanings that relate to text organization, speaker–addressee interaction, and the cognitive processing of a message, and thus to meanings outside those expressed in the unit they precede. Some of them express, for instance, subjective meanings like attitude towards the co-participant (e.g. honestly or frankly, which indicate, at least rhetorically, a certain degree of intimacy between the co-participants), projecting the illocutionary act and the kind of action they are about to perform (e.g. asserting something confidential), or expressing stance or degree of commitment towards what was said before, such as full endorsement (e.g. of course). Note that those adverb-like lexemes that may occur in the initial field have parallel uses as constituents within a microgrammatical unit, where they are integrated into a hierarchical relationship with another form.

4.3.3 Discourse Markers

Discourse markers (DMs) are a rather elusive category as there is no generally accepted definition of the elements subsumed under this category and as the set of formal and functional features characterizing them is rather diffuse. Due to the lack of a clearly definable boundary between DMs and expressions without DM status, there is also no generally accepted list of DMs in English. Part of the problem is that the term discourse marker has been conceptualized in different ways. Basically, one can distinguish two different “schools,” one of which represents a more “narrow,” the other one a “broader” view on what qualifies a lexeme to be classified as a DM (see Diewald Reference Diewald, Degand, Cornillie and Pietrandrea2013). Under a more narrow view, discourse markers are elements that are usually (and for some authors even necessarily) syntactically integrated into a sentential matrix and thus syntactically dependent on a host syntagma. Their function is to signal a propositional relationship between two neighboring textual units or “discourse segments” (S), i.e. they relate aspects of the explicit message of a second unit to aspects associated with a message of a first unit under a schema “S1–DM+S2” (Fraser Reference Fraser1999: 938, 950). According to this definition, DMs are e.g. conjunctions (e.g. but, because, or, so) (Cuenca Reference Cuenca, Degand, Pietrandrea and Cornillie2013), adverbials (e.g. of course) or prepositional phrases (e.g. as a result of), provided that they serve the introduction of a separate message with a propositional content of its own, i.e. one that is not embedded in the message of the prior unit. Two examples (but and so) are provided in (29).

1. 1. a. A: She didn’t even invite him. But that’s another story.

   2. b. C: I’m not in a hurry so how about having a talk over coffee?

The defining function of such elements is discourse connectivity: DMs have a text-connecting function, linking two adjacent units that range in size from clauses to larger textual units and indicating the relationship between them. In terms of Schiffrin (Reference Schiffrin1987: 31), they “bracket units of talk,” marking a boundary between a prior and an upcoming discourse unit. The scope of a DM is over two adjacent textual units, each expressing a separate message.

Under a broader conceptualization, DMs are elements that are syntactically non-integrated lexemes with no fixed position and without constituent status; they are syntactically as well as semantically and often prosodically independent. Functionally, they do not only link two textual units, but refer to dialogic interaction in general, encompassing nonpropositional, metadiscursive aspects, such as relationships between different conversational actions or between the speaker’s and the addressee’s views. In this sense, the basic functional distinction between DMs in the two theoretical views is that in contrast to the first view, in the second, broader view the type of connectivity they establish is not restricted to propositional relations between adjacent utterances or discourse units, but also pertaining to relations between nonlinguistic, nonpropositional aspects of discourse (Diewald Reference Diewald and Fischer2006: 408). DMs serve the management of conversation in a broader sense and have scope over utterances, illocutionary types, and actions accompanying an interaction in general. Text connectivity is thus only one among several aspects of their functional spectrum: they refer to general aspects of dialogic interaction, such as thematic structure, the turn-taking system, strategies of argumentation, and cognitive aspects. The DM actually, for instance, does not only link two units of discourse, but refers to the expectations and thus the to cognitive state of the addressee, for whom the upcoming information will be or might be unexpected (Taglicht Reference Taglicht2001), as in (30); it may also signal topic change and thus organize discourse into a coherent whole despite potential disruption caused by a change of the topical direction.

1. (30)

   133	B:	uhm sh- she’s got quite a small house a small garden a small
   134	A:	uhm whereabouts
   135		whereabouts
   136		sh- she’s got a small GARden <unclear word> °the fire the
   		bonfire and,
   137 →	B:	actually it’s not a small garden.
   138		no it’s not small it’s quite big.                   [ICE-GB S1A-025]

In (30), B marks her upcoming utterance as expressing a divergence from A’s belief or knowledge state by means of actually: while A assumes that she’s got a small garden, B seems to be certain that it’s quite big. As a DM, actually introduces a new unit of talk whose propositional content is seen as unexpected for the addressee by the speaker and dissonant to the addressee’s views or assumptions.

Considering examples such as (30), it appears more appropriate to adopt a broader view on DMs: DMs are syntactically unintegrated expressions that do not only establish links between neighboring discourse units on the propositional level, but serve a broader set of indexical functions on the interactional level, relating also to communicative, nonpropositional aspects of talk. In spite of these different conceptualizations of DMs, there seems to be agreement on some basic features that characterize them: DMs do not affect the truth conditions of the unit they accompany, they do not add or change propositional content, they tend to occur initially, though some DMs are positionally flexible, and they indicate how the following unit is to be understood or interpreted within a given communicative context, thus having procedural rather than referential, lexical meanings (Blakemore Reference Blakemore1987, Reference Blakemore2002).

Individual DMs typically exhibit a broad set of functions (see Fischer Reference Fischer2000). Jucker (Reference Jucker1993), for instance, identifies four different functions for the DM well: it serves as (i) a marker of insufficiency, indicating some problem on the content level relative to the preceding utterance, e.g. when speakers cannot give the information which the questioner has requested; (ii) a face-threat mitigator, mitigating some sort of confrontation, e.g. when an assessment is followed by disagreement or when an invitation is declined; (iii) a device used for indicating a topic change or introducing direct reported speech; or (iv) a delaying device. A similar diversity of functions has been observed for actually, which expresses functions like mild contradiction, surprise, change in discourse topic, or contrast to what was said before (Lenk Reference Lenk1998; Taglicht Reference Taglicht2001; Aijmer Reference Aijmer2002). As these examples show, DMs are multifunctional and adaptable to the sequential environment in which they are used. Descriptions of the multifunctionality of DMs differ as to whether the respective authors adopt a meaning-minimalist stance, according to which the diversity of functions can be reduced to one functional core, or a meaning-maximalist one, under which it is assumed that the different functions that can be identified for single DMs are unrelated and not plausibly derivable from a core meaning. I cannot do justice to the vast amount of studies dealing with DMs here; the reader is referred to studies that figure most prominently in this field, such as Schourup (Reference Schourup1985), Schiffrin (Reference Schiffrin1987), Jucker & Ziv (Reference Jucker and Ziv1998), Lenk (Reference Lenk1998), Fraser (Reference Fraser1999), Aijmer (Reference Aijmer2002), Fischer (Reference Fischer2006), and most recently Heine (Reference Heine2012) or Degand, Cornillie & Pietrandrea (Reference Degand, Cornillie and Pietrandrea2013).

From a macrogrammatical perspective, DMs are not syntactically integrated into the structural unit they project: they are not embedded in a binary hierarchical relationship with another form (and therefore do not classify as adverbs), but represent formally autonomous units in the initial field of a unit of talk. They establish structural relations outside single units of talk, thus serving conversation-organizing functions, and have an important cognitive function in that they project a particular kind of upcoming action, allowing the listener to process a message not as merely contributing propositional content, but as a message of a particular kind, e.g. one expressing mild contradiction (well) or potential violation of the addressee’s expectations (actually). They also serve a variety of functions on the interpersonal level, such as mitigating face-threats inherent in disagreements or topic shifts. As cues for discourse processing, they contribute to the development of a coherent mental model of ongoing discourse by integrating single units of discourse into a coherent whole.

4.3.4 Address Terms/Vocatives

Address terms in the initial field are primarily used to summon the attention of the addressee, especially in a communicative situation in which the co-participants need to achieve joint orientation to an upcoming message. This involves activities such as attracting and focusing the addressee’s attention and the arrangement of a joint interactional space including bodily co-orientation of the participants, e.g. in situations where mutual visibility and fixing gaze need to be arranged to provide the necessary ground for the incipient unit of talk (Mondada Reference Mondada2009; Deppermann Reference Deppermann2013). This is the case when speakers move in space and/or deal with different activities, or when they are located in different rooms. In such contexts, speakers need to secure the availability of the addressee for the message to be produced, dealing with mutual availability, attention, readiness for interaction, recognition of speaker roles, and coordination of gaze. These are the prerequisites that need to be achieved in order to establish joint orientation to the upcoming message and for which vocatives or address terms play a central role. This is illustrated in (31), where the co-participants are moving to a table for coffee. They seem to be situated in the visual field of speaker D, given that all co-participants are in the same room, but physically not co-oriented in the way necessary to recognize who D is addressing with his utterance, at least not right from the beginning, as they are busied with choosing a seat. D addresses his guests with their first name.

1. (31)

   52	D:	anybody else like a piece of anything that they can see on here
   		that I haven’t given them?
   53	A:	<<unclear words>>
   54	B:	<<unclear words>>
   55 →	D:	Peter where are you sitting?
   56		you’re there. (.)
   57 →		TIM? (..) are you going were you going to sit here. [(.) ]
   59	C:	[here.]
   58	D:	would you like a piece of that?
   60	C:	yes please.                                                        [ICE-GB S1A-012]

From the design of D’s turn we may conclude that the addressees shift bodily positions in succession to select a seat at the table and have multiple, changing visual foci. Thus, D needs to coordinate the attention of the different potential addressees and to update the referential addressee as he changes interactional partners. By means of using the first name of the addressees as a beginning of the units of talk in lines 55 and 57, D establishes the prerequisites for the reception of his requests as he is not able to build on already established arrangements of the participants’ coordination, such as when conversing while sitting at a table, or in dyadic interactions. These prerequisites include securing the availability of the addressees for the request, signaling that one is willing to take the speaker role, picking out the intended addressees distinctively (in multiparty situations), and attracting the addressee’s attention for the upcoming request (which is the prerequisite for a response). In short, address terms serve as warrants for the addressee’s attention to the upcoming message.

Address terms are not part of microgrammatical relations as defined in Chapter 3, and not part of the propositional content expressed in a unit of talk as their omission would not alter the propositional content or truth-value of the unit they precede. However, they are not outside a unit of talk as they make some kind of continuation relevant: they mark the onset of a new unit of talk and are used as a device to achieve the necessary conditions for a successful uptake of what follows. Address terms have macrogrammatical functions in that they organize joint cognitive orientation to an upcoming unit of talk and situate this unit within the social coordinates. As explicit means to pick out the addressee distinctively, they not only serve an attention-getting function, but also convey subjective meanings such as positive affection since address terms make the interactional relation to the addressee and thus the speaker’s orientation toward him/her as a distinctive individual explicit. Address terms are also used to mark a unit of talk and the ensuing action as set off from prior talk, and thus have a discourse-organizing function. Clayman (Reference Clayman2012), for instance, notes that address terms used in responsive actions serve to mark the response as independent of the initiating action, e.g. a question, detaching it from its presuppositions and separating the upcoming utterance from the motivational entanglement with the preceding question. In this sense, they serve as cues on how an upcoming unit is to be understood in a given discourse context.

4.3.5 Parentheticals/Comment Clauses

The initial field may host a fixed set of configurations that, from a sentence-based grammatical perspective, look like main or matrix clauses consisting of a subject (usually a first person singular pronoun) and a mental verb that requires a complement clause as an object, such as I think, I believe, I guess, I mean, also you know or it seems. Such units, as well as forms like to-infinitive clauses, such as to be honest, or finite clauses introduced by as (e.g. as I see it), which usually function as adverbial clauses, are analyzed under different labels, such as parenthetical disjuncts by Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985: 1112), comment clauses (Brinton Reference Brinton2008; Kaltenböck Reference Kaltenböck2008, Reference Kaltenböck, Aarts, Close, Leech and Wallis2013), or parentheticals (Dehé & Kavalova Reference Dehé and Kavalova2006; Dehé Reference Dehé2009, Reference Dehé2014). The term parenthetical suggests nonclausal and morphosyntactically “unintegrated” status of the expressions to which it refers. For practical reasons, I will restrict the category of parenthetical clauses to subject-cum-mental verb combinations.

The syntactic status of parentheticals allows for two alternative interpretations, at least in English: utterance-initial I think or I believe lacking a complementizer (that) are ambiguous between a main clause taking the rest of the sentence as object, and a parenthetical or discourse marker, which is morphosyntactically and semantically separated from the structural unit following them. Note that in medial and final positions an interpretation as “main clause+complement clause without complementizer” cannot be obtained since they do not introduce a complement clause, but even this is not undisputed since it may be argued that such an analysis is possible under the assumption that such clauses are the result of a syntactic movement operation (most recently Newmeyer Reference Newmeyer, Schneider, Glikman and Mathieu Avanzi2015).Footnote ² In initial position, however, a main clause interpretation is, for many authors, not excluded. Dehé & Wichmann (Reference Dehé and Wichmann2010a, Reference Dehé and Wichmann2010b), for instance, have shown that the meaning of I think varies along different prosodic realizations, depending on whether the main stress falls on the subject, the verb, or whether I think is prosodically unmarked. In the last case, a discourse marker reading without propositional meanings is most likely, given that prosodic reduction correlates with reduced semantic content (see Pierrehumbert & Hirschberg Reference Pierrehumbert, Hirschberg, Cohen, Morgan and Pollack1990). Moreover, research by Dehé (Reference Dehé2009, Reference Dehé2014) has shown that a variety of intonational phrasing patterns is possible with parenthetical clauses such as I think, allowing both for prosodic separation such that the parenthetical clause is phrased in its own intonational phrase (IP), and prosodic integration, such that the clause is phrased in one IP with surrounding material. Parenthetical clauses whose primary function is one of mitigation and which have only very reduced semantic content are unstressed and integrated into the contour of the unit that follows. Additionally, they are prosodically characterized by increased speech and lowered volume. Prosody may thus disambiguate between various possible functions of parentheticals.

Since the present approach is based on the linear, emergent nature of speech, it will be argued – based on the findings referred to above – that initial parentheticals that lack prosodic prominence no longer function as matrix clauses with an embedded subsequent subordinate clause. Rather, they form the initial field of an upcoming unit and thus precede a structural unit without having a microgrammatical relation with it; that is, their use is not related to hierarchization and embedding. An important indicator for the lack of embeddedness is, next to the absence of a formal marker of a binary, hierarchically organized relationship, that parentheticals may be followed by units of any structural type, ranging from a single lexeme to a longer textual segment. Thus, from a macrogrammatical perspective the syntactic structure involving a parenthetical expression can be analyzed as consisting of an initial parenthetical followed by an independent (non-subordinate) assertion in any syntactic format. The parenthetical itself represents a chunk (pragmatic tag [Weinert Reference Weinert2012: 237] or epistemic parenthetical [Thompson & Mulac Reference Thompson and Mulac1991]) that is integrated into the illocutionary force of the unit it accompanies, but that does not establish or take part in an embedded structure in sentence-grammatical terms. The informational import does not lie on the parenthetical, but on the assertion, which also typically contains the topical subject, which is why “conversationalists are normally oriented to the material in the complements and independent clauses and not to the CTP-phrases” (CTP=complement-taking predicate; Thompson Reference Thompson2002: 154).

The case of parenthetical expressions is congruent with the observation that speakers favor the aggregation of segments that are not complements syntactically, but combinations of evidential or epistemic “phrases” and an assertion linked by loose discourse-pragmatic relations (Thompson Reference Thompson2002). In the initial field, the mental verbs involved in parenthetical expressions project an upcoming message whose syntactic format is not specifiable on a priori grounds (it may range from a single lexeme to a larger unit of discourse), and express an epistemic–evidential evaluation relating to the upcoming message. A semantic reading of these verbs cannot be obtained. Initial I think, for instance, does not indicate the speaker’s mental processing activity, but marks an upcoming message as an opinion or an assumption to which the speaker is principally committed, but whose commitment is mitigated for some reason, e.g. due to a lack of epistemic certainty, as illustrated in (32).

1. (32)

   45	A:	d’you d’you have d’you have that book °on him°.
   46		do-
   47	B:	I couldn’t find it.
   48		I don’t know what.
   49		I think I must have lent it to somebody.       [ICE-GB S1A-045]

In (32), prosodically unmarked I think marks the assertion following it as a vague idea that seems most plausible to the speaker, but whose validity cannot be scrutinized at the moment of speech. This is also reflected in the choice of modality (must have lent), where must has no deontic, but epistemic meaning indicating that the assertion is not a fact, but an inference based on some piece of evidence that is, however, not specified (see Traugott & Dasher Reference Traugott and Dasher2002: 113). I think is thus an interpretive cue for the addressee to take the validity of upcoming utterance as preliminary. I think also functions as a downtoner of illocutionary force with opinions, e.g. in contexts in which a speaker expresses a suggestion or an assessment (e.g. I think he’s very good actually, S1A-049). As a mitigator of illocutionary force, I think frequently co-occurs with other mitigators, e.g. probably (e.g. I think that’s probably what I’ll do, S1A-053).

A text-structuring – rather than modifying – function can be observed for I mean, which is used to signal that the upcoming utterance does not express new information, but a reformulation or specification of some prior piece of talk that is supposed to represent a clearer or more condensed way of expressing e.g. an idea or, as in (33), the main point of an information-seeking request.

1. (33)

   95	B:	it’s very good work but I think (.) that art and dance actually
   		include (.) the idea of therapy within them. (.)
   96	A:	when you say recovering the whole person it suggests that there
   		is something lost.
   97 →		I mean you know is there something incomplete?
   98 →		I mean [(.)    ] uh what’s w- what’s what’s incomplete? (.)
   99	B:	[yeah. ]                                      [ICE-GB S1A-004]

The parenthetical I mean is, in both cases, not analyzable as being embedded in the following unit: it could not be followed by a complementizer (that), and as a main clause it would be incompatible with the clause types of the following units (interrogative clauses). Rather, A marks the upcoming requests for information as being inferentially related to her initial assertion (line 96), which expresses an inference that she has drawn from B’s prior talk. I mean thus indicates that the requests following it will make an aspect of a prior utterance (in this case the question deriving from the inference that A has drawn) more explicit and thus more clearly identifiable for the addressee. A literal reading of I mean as referring to the expression of a personal opinion or meaning cannot be obtained; that is, I mean is not propositional, but rather providing a processing cue on how an upcoming piece of talk is related to prior talk.

4.3.6 Clause-Like Fragments

Clause-like fragments are chunks that derive from microgrammatical units that underwent a progressive loss of internal syntax, thus acquiring a more formulaic character by the loss of originally compositional semantics and of the ability to govern other structural elements or units, and the adoption of functional values that are not derivable from the component parts. A frequently occurring unit in the initial field is the “the+N+is” chunk, e.g. the thing/point/problem/trouble … is, which was originally composed of a definite noun phrase headed by an abstract noun and the third person singular form of the copula verb to be, which syntactically requires a complement. The format occurs also in other languages, e.g. in German (das Ding/die Sache ist (dass), see Günthner [Reference Günthner2008, Reference Günthner, Auer and Pfänder2011]) or Spanish (la cosa es (que)). Based on how such fragments are used in spontaneous speech, an analysis in terms of microgrammatical units in the format of a main clause followed by a complement clause can, however, not be supported as they often occur without a complementizer, are usually not followed by a clause, but by a larger discourse segment, and they do not express propositional content to which the content of the subsequent unit is subordinated. On the contrary, it is the information following the fragment that expresses the main idea. Moreover, the projected unit could occur without the initial fragment that frames it without being “ungrammatical,” i.e. the unit introduced by the “the+N+is” chunk is structurally not subordinated, but autonomous. In this sense, the fragment has a formal and functional affinity to other fragmented or formulaic expressions that project an upcoming message, such as pseudoclefts with lexical specification (e.g. what I mean (is)), or initial complement-taking predicates/parentheticals (e.g. I mean/think/believe), which exhibit the same features. Due to their function of projecting a piece of focal information (Günthner Reference Günthner2008, Reference Günthner, Auer and Pfänder2011; Miller & Weinert Reference Miller and Weinert2009 [1998]: 243), clause-like fragments cannot represent a unit of talk of their own but require an upcoming structural unit that expresses the main information. The main task they serve is that of focusing the addressee’s attention to what the speaker considers an important piece of information; the second part or segment is thus framed as focal information or a core argument or idea that remains relevant for ensuing talk. The second part can take various forms ranging from a clausal structure to a longer stretch of discourse that expands over several units of talk, and it may express a single proposition or, as in (34), a series of propositions. In (34), B explains how he (a person called Uts) has long been playing with a girl by never making explicit whether or not he had any genuine interest in her.

1. (34)

   80	A:	and she is she s- she spent the entire evening convincing her
   		that Us UTS was desparately passionately in love with her (.)
   		and it was: uhm,
   81	B:	fatal <<laugh>>
   82	A:	↑well I mean it’s=
   83 →		=the thing is he was supposed to be coming over for her birthday
   		°coming back for her birthday (.) uh because he’s over there
   		now°.
   84		a:nd uh she hadn’t seen him for nine months or something.
   85		her birthday’s in May. (.)
   86		and he phoned and yes he was going to get a f- a a standby
   		flight (.) and he never turned ↑up.
   [ICE-GB S1A-093]

The chunk the thing is serves two functions: the speaker projects an upcoming piece of information that may have any syntactic format and length and that includes information that is focal, expressing a main argument or crucial information that is required to understand a particular state of affairs. Additionally, this construction allows speakers to justify the adoption of the turn or, as in (34), to keep it over a longer stretch of time. The speaker thus segments a unit of talk into two parts, an initial field and a core message. The expression in the initial field allows speakers to deal with (i) projections by opening up a conversational space that makes further talk expectable and necessary, and framing upcoming talk as expressing focal information, (ii) the organization of interaction (acquiring the right to adopt/continue a turn, focusing the addressee’s attention), and (iii) the cognitive planning of upcoming talk, which need not have been completed upon the production of the fragment. (ii) and (iii) are particularly important when the speaker is about to produce a longer discourse unit, as in (34), for which s/he requires the temporary suspension of transition-relevant places in order to complete the message. It is therefore not surprising that the fragment is typically used to introduce a more complex argument or description that spans over several syntactic segments or units of talk. Prosodically, clause-like fragments in the initial field and the following stretch of talk are usually not separated by a pause (though this is not excluded); the ensuing talk is typically “latched” to the initial fragment.

Clause-like fragments are characterized by minimal semantic content and belong to the domain of macrogrammar: the fragments are not part of a morphosyntactic relationship to another unit and thus not embedded, but serve the cognitive organization of a unit of talk in that they grant the speaker some planning time, focusing the listener’s attention and creating expectations in the listener on upcoming talk by delaying the production of the main information.

4.3.7 Expressions in the Initial Field as Macrogrammatical Elements

The discussion of the different kinds of expressions available for use in the initial field of a unit of talk has shown that all of them evade categorization in terms of established “parts of speech” or “word classes” and do not qualify as “constituents” in the traditional sense of sentence grammar: they are not morphosyntactically integrated into binary relations with other lexical or phrasal constituents and do not establish such relations. The expressions include lexical units and formulaic fragments or chunks and are thus conventionalized units that tend to form a relatively closed class. Expressions in the initial field exhibit the following features, which characterize them as macrogrammatical:

1. (i) they are used to construct and organize spoken text rather than to create microgrammatical relationships,

2. (ii) they tend to be morphosyntactically autonomous,

3. (iii) their internal structure is not or only weakly compositional,

4. (iv) they do not contribute to, change or restrict the propositional content of the unit they project; their semantic content is minimal,

5. (v) they serve as addressee-oriented processing or attentional cues, framing an upcoming message and indicating how it is related to prior talk and/or how it is supposed to be interpreted.

4.3.8 Functions of the Initial Field

Expressions in the initial field allow speakers to deal with a set of recurrent communicative tasks that are relevant in the initial phase of the production of a unit of talk. First, speakers need to deal with turn-taking issues, which include activities such as foreshadowing a longer contribution and legitimizing the adoption or continuation of the speaker role. The initial field is thus an important locus for the organization of conversational interaction, where speaker rights and the amount of conversational space that a next speaker is granted need to be negotiated. Second, spontaneous speech requires speakers to bridge the time gap between turn adoption and syntactic planning, the latter of which might not have been progressed when the speaker assumes the right to speak. The initial field thus correlates with intense speech planning activities.

Elements in the initial field are also understood by the co-participants as representing a particular kind of reaction to just-prior talk. This default assumption derives from the relationship of adjacency inherent in talk. As Schegloff (Reference Schegloff2007: 14–15) states, “[n]ext turns are understood by co-participants to display their speaker’s understanding of the just-prior turn and to embody an action responsive to the just-prior turn,” unless the speaker marks the turn as not addressing some aspect of the just-prior turn, e.g. through displacement markers such as by the way. Thus, listeners understand each new turn as responsive to the preceding one since prior turns typically form the starting point of subsequent turns. Speakers are naturally oriented to prior talk while designing a next contribution in order to provide a coherent (and intelligible) continuation of talk. The topic and the action established in a prior turn constrain listeners within “presuppositional matrices” (Heritage Reference Heritage2013: 333) in that next utterances need to be type-conform to immediately preceding ones. Typical reactions to prior turns encompass the expression of stance or the divergence from the normative constraints that a prior turn bears on the next turn. The principle of continuity also holds for multi-unit turns: next units are interpreted against prior units given that, in the default case, speakers continue their own talk in a coherent way.

Next to this backwards-oriented, responsive function, speakers use the initial field to provide an early interpretive cue as to the type of action that they are about to perform, and thus to project into the discursive future. Early projection facilitates cognitive processing in that addressees can start the interpretation of upcoming talk right from the point at which a unit of talk is initiated, which reduces the time gap between production and processing and allows for maximal synchronization between what the speaker is doing and the listener’s understanding of it (Auer Reference Auer2009: 3).

All of these tasks, which involve macrogrammatical expressions, have been predicted by the temporal logics of communicative tasks that speakers need to deal with in the initial phase of unit of talk production discussed in Chapter 3. The tasks are summarized in Table 4.3.

To conclude, the initial field is a point in time at which speakers instruct their addressees to recognize the connection between a prior and an upcoming unit of talk in terms of continuity of topic and action. It is an important resource for setting up expectations in the listener on what there is to follow, focusing and guiding the listener’s interpretation from the earliest point of utterance interpretation possible. It is also used to reset expectations that derive from prior talk and to deal with various aspects of speaker–addressee interaction, such as turn-taking and attention management.

4.4.1 The Post-Initial Phase

The temporal phase immediately following the initial field, which can be regarded as forming the post-initial phase in which the linguistic forms produced still project more to come, is characterized by a high frequency of self-repairs, cut-offs and restarts. These tend to follow expressions in the initial field, as shown in (35a), which exemplifies two sudden changes of a syntactic trajectory that had already been underway (syntactic repairs), and (35b), which illustrates the restart of a syntactic fragment.

1. 1. a.

      72	A:	have you still got that damn machine on.
      73	B:	uhm yes. (.)
      74 →		but you can’t- you see you’re not meant to notice it.
      75 →		it’s got to be: the tape’s got to be °on for°
      76		↑they disregard the first fifteen minutes anyway(..)
      		because they reckon that it won’t be natural.
      [ICE-GB S1A-047]

   2. b.

      138 →	C:	well you know I sent my- I my sent my son off to boarding school,
      139		but I thought that as you were here I might as well send
      		my daughter here. (.) <<unclear words>>
      [ICE-GB S1A-012]

The observation that the (post-)initial phase is highly susceptible to dysfluencies such as hesitations and repair phenomena has already been made by other authors (e.g. Beattie Reference Beattie1979; Barr Reference Barr, Cavé, Guaïtella and Santi2001). The reasons can be found in the greater cognitive demands on planning processes at this point in utterance production, that is, the kinds of cognitive activities that become relevant after backlinking and responsive activities: the speaker undergoes a phase in which a number of syntactic alternatives are “in play,” that is, available for use, which need to be simultaneously tracked, and in which several syntactic interpretations of the first word or words that has/have already been produced are possible. In this phase, the speaker needs to “build up syntactic projections while assembling syntactic patterns in a phase of high cognitive stress” (Auer Reference Auer2009: 4). In other words, the cognitive task of the speaker is to select a suitable structural format from a broad range of possibilities for the kind of message s/he wants to convey and the kind of action s/he intends to perform, to distribute thematic roles, and to make preliminary decisions on the information status and the relative size of an emergent pattern. Thus, in the initial and post-initial phase speakers deal with the assembling of linguistic signs in order to open a syntactic projection that creates expectations in the listener’s mind on the further development of a structural unit, and to eliminate possible alternative interpretations of the kind of structure and action they are about to produce. Early predictability of a structural unit-in-progress is the precondition for synchronization, which allows for fluent dialogic coordination between the actions performed by the speaker and the co-participant.

Due to this relatively high cognitive stress at utterance beginnings, it is not surprising to find macrogrammatical elements quite early after the initial field: if speakers are still in the planning phase, they are likely to encounter planning difficulties at the conceptual or syntactic level in the early phase of structure building, in which they are required to deal with the problem of committing themselves to a particular structure that builds around one or more words already produced. They are thus most likely to pause and signal various kinds of planning activities after the production of what could be heard as the first constituent of a microgrammatical unit, and thus before or immediately after an initial commitment is made, but less so when the unit is already in full progress (see Clark & Wasow Reference Clark and Wasow1998: 205) and thus when fewer syntactic alternatives are still “in play.”

This view implies that cognitive load predicts the occurrence of dysfluencies and the use of macrogrammatical expressions that document planning activities and delays. Some evidence for this assumption has been provided e.g. by Oviatt (Reference Oviatt1995) and Shriberg (Reference Shriberg1996), who have shown that dysfluencies occur more often before longer than before shorter utterances, and thus when speakers carry out more extensive planning that spans over a larger structural unit. Clark & Wasow (Reference Clark and Wasow1998) provide empirical evidence that speakers tend to disrupt an utterance-in-progress and produce more repetitions and dysfluencies in proportion to the relative complexity of an upcoming constituent: for instance, determiners and pronouns are repeated much more often at the beginning of longer units with a larger number of syntactic nodes in the constituent (a phrase or a clause) than before shorter constituents with lower syntactic complexity. The complexity hypothesis, which predicts that the more complex a constituent the more likely speakers are to suspend speaking after an initial commitment to it, is thus a claim about processing limitations.

These findings refer to dysfluencies in the post-initial phase, which refers to a point in time at which the first constituent of an upcoming structural unit has been produced and at which speakers make a first commitment to a particular structural pattern including that constituent. The commitment is, however, cancelable and thus preliminary, given that many commitments are premature: as argued by Maclay & Osgood (Reference Maclay and Osgood1959: 42), “since structural choices typically involve fewer alternatives than lexical choices the speaker will often initiate a constituent before he has completed his lexical decisions” and, one may add, his/her syntactic decisions. Premature commitments are perhaps driven by time pressure in ongoing talk, where time does not only belong to the speaker, but also to the addressee (Goffman Reference Goffman and Goffman1981: 198), and where speakers that pause too long, especially at the beginning of a unit of talk, might be heard as opting out, being distracted, or as being unsure of what to say. By making a preliminary commitment at the start of a new unit of talk speakers may forestall such misattributions, with “speakers already expecting, at some level of processing, to suspend speaking immediately afterward” (Clark & Wasow Reference Clark and Wasow1998: 208).

Thus, there is evidence that dysfluencies typically occur in contexts where speakers have a high number of syntactic options to choose from. Given that the choices are more numerous at the beginning of a structural unit than at later points in the production process, the cognitive load deriving from dealing with various syntactic (as well as lexical) formats at unit of talk beginnings is an explanatory factor for the phenomenon that structures are often abandoned and replaced by a different one (which means that a new commitment is made) after the production of the first element(s), or are taken up again and continued after a brief moment of hesitation when the commitment to a particular syntactic pattern is strong enough.

What is relevant for the discussion of the microgrammar–macrogrammar distinction is that macrogrammatical expressions occur frequently at such caesuras in the post-initial phase. These expressions serve different functions, such as smoothing out dysfluencies and indicating different kinds of ongoing speech planning activity, such as reformulation (I mean), searches for expressive means (uh, uhm), or references to shared knowledge or assumptions (e.g. you see or you know, as in (36a–b) below). The latter typically signal that the speaker is about to express a state of affairs that is accessible to or highly plausible for the addressee, which might release the speaker from spending too much effort in explicit formulation and preciseness of expression.

1. 1. a.

      125	B:	uh I was (.) I mean I wasn’t (.) shocked that the that
      		the the students had not met people with disabilities or
      		people in wheelchairs. (.) uhm (..)
      [ICE-GB S1A-001]

   2. b.

      127	A:	and the physical contact [(.)    ] out there is the is the
      128	B:	[uhm]
      127	A:	only <<unclear words>> problem of being isolated in your
      		(.) °wheelchair°.
      129	B:	uh uh in a in a chair?
      130	A:	mhm.
      131	B:	I think so.
      132		yes.
      133 →		I mean I think uhm (.) space (1.5) is you know I mean (.)
      		you know just the obstacles that you have in a room (.)
      [ICE-GB S1A-003]

In (36a) speaker B replaces a syntactic unit introduced by I was by a negative construction (I wasn’t) after a micropause and the reformulation marker I mean; in (36b) the speaker appears to have more serious formulation difficulties, first starting a microgrammatical unit consisting of a “subject+copular verb” combination (space is), which projects a subject complement, but then abandons the structure switching to macrogrammar, shifting between addressee-involvement (you know) and an attempt at a reformulation (I mean), before initiating a new grammatical format that is, however, relatively fragmented as it represents an isolated noun phrase with internal postmodification and thus a referent that is not integrated into a state or an event schema involving a predicate. However, the unit expresses the speaker’s current focus of consciousness (obstacles that you have in a room), which can be plausibly linked to the meanings associated with the initial, abandoned referent (space) and the topic of the prior talk (being isolated in a wheelchair): the topic of feeling isolated in a wheelchair and the reference to space allow for an integration of the noun phrase into a topical frame centering around the problem that obstacles in a room that provides only little space for someone in a wheelchair to move around contribute to the feeling of isolation. The interpretation of the unit, however, requires the ability to construct a coherent model of discourse by the listener by integrating different segments of talk into a coherent whole. In both cases, macrogrammatical expressions occur right after the beginning of a microgrammatical unit.

A further regular phenomenon at unit of talk-beginnings that is relevant for explaining the deployment of macrogrammatical expressions shortly after the initial field relates to Schegloff’s (Reference Schegloff, Ochs, Schegloff and Thompson1996: 78) observation that units of talk often start with a candidate beginning that is cut off and followed either by a new element (e.g. a referent) or by the reemployment of the original beginning, the latter of which is illustrated in (37a–b).

1. 1. a.

      250	C:	does an artist have to live with an artist?
      251	B:	yes,
      252		what happens?
      253	C:	I don’t see why.
      254	A:	uh do do they have to or can they?
      255 →	B:	Brow- well uh BROWning and uhm and Emily you see lived
      		together.
      [ICE-GB S1A-020]

   2. b.

      30	D:	have you decided what to do with exercise number six.
      31	C:	when did you do that?
      32		how many times did you do that?
      33	D:	a couple of times.
      34 →	A:	u:hm the only time=I think the only time was at the
      		bottom of what’s called a hundred and thirty.
      35	C:	yes.
      [ICE-GB S1A-026]

In (37a–b), the reemployment of the beginning is linked to the original beginning by the insertion of a macrogrammatical element, which allows the restart to be recognized as leading into a unit of talk of a particular kind. In (37a), the production of the proper name Browning is self-interrupted after the first syllable and followed by well, which serves as a delayed signal indicating that the response given by B will diverge from the premises set up by A’s information-seeking request: B will not respond to A’s question following the options offered by A (do they have to or can they?), which refer to whether an artist can only have a partnership with someone who is also an artist or whether this is just an option and more or less a coincidence. Rather, B provides an example in which both partners happened to be artists, from which A needs to infer the kind of relation to his request. B thus marks his utterance as not directly contingent to the premises of A’s request after he had already started producing the first constituent of an upcoming (microgrammatical) unit. Well seems to document B’s orientation to the apparent divergence from the kind of response made relevant by A, marking the unit of talk as disjunctive at the earliest point possible. In (37b), speaker A interrupts the beginning of a microgrammatical unit in order to insert the macrogrammatical expression I think, which serves as a delayed device to mark the content of the upcoming unit as epistemically weak, lacking facticity, and thus as potentially challengeable. In both cases, the speakers interrupt a candidate beginning at an early point in order to indicate information relevant for the processing of the upcoming unit of talk and to improve its integration into the sequential context or to indicate the way it is to be interpreted, before the abandoned beginning is redone. The insertions have a clear macrogrammatical character since they organize a unit of talk by linking smaller fragments into an integrated unit of talk. Additionally, they mark the relationship of the unit in which they are produced to prior talk or to the speaker’s mental world in a maximally economic way without making it the central concern of the speaker, who continues the structure already initiated after the reemployment of the beginning.

To sum up, in terms of dysfluencies we find two different ways in which the development of a unit of talk may be carried out in the initial phase of its production (“beginning” refers to the production of the initial constituent(s) of a microgrammatical unit):

1. (i) Beginning + cut-off (+ macrogrammatical insertion) + restart with a new element,

2. (ii) Beginning + cut-off (+ macrogrammatical insertion) + reemployment of beginning.

Macrogrammatical expressions are not always used in such contexts, as is indicated by round brackets above, but a frequent phenomenon. They document what the speaker is about to do or which tasks s/he is currently dealing with, e.g. indicating formulation difficulties, representing causalities of a high cognitive load in the planning phase, or marking epistemic uncertainty.

Theoretically, there should be a third possibility, which is that speakers begin a unit of talk and continue the structural unit without reemployment of the beginning (= “beginning [+ macrogrammatical insertion] + continuation”). However, this pattern occurs very rarely in the data and is by far outnumbered by either a repair of the initial element (usually a form that is part of an emergent microgrammatical structure), or its repetition. One reason for the preference of restarts may be the speaker’s intention to redo a microgrammatical form in its ideal delivery and thus to facilitate parsing for the listener (Clark Reference Clark2002): speakers might be guided by the intuition that forms that are part of hierarchical relations are easier to parse and understand when they are integrated into a coherent sequence with the forms to which they are hierarchically related than when they are disrupt. Thus, their reproduction restores continuity after the disruption caused by the suspension. A further possible reason refers to processing limitations on the side of the speaker: it might be easier for speakers to formulate a structural unit from the beginning rather than from the middle, assuming that producing a structural unit from the beginning helps speakers keeping track of where they are. Both reasons have been discussed in the context of word repetitions at utterance-beginnings by Clark & Wasow (Reference Clark and Wasow1998: 207), who point out that, in principle, both of them apply as they represent two sides of the same coin.

A brief note on pattern (ii) is in place here. The repetition of an initial element or segment of an emergent structure is a highly frequent pattern among dysfluency phenomena in the initial or post-initial phase of the production of a unit of talk. In many cases it is realized without macrogrammatical elements and delays, as shown in (38).

1. (38)

   306	C:	you’d better be nice to her.
   307	A:	yeah.
   308	C:	yeah.
   309 →	B:	she could come she could come down on Tuesday and say I want
   		fifty pounds (.) <<laughter>>
   310		she could say or else I’m going to snitch or I’ve already taken
   		down out down the West End and go and see <<unclear syllables>>
   		something fantastic.
   [ICE-GB S1A-030]

Such repetitions at unit of talk-beginnings have various reasons. First, speakers need to set up the conditions for the successful uptake of their talk and focus the attention of the addressee(s) to their talk. As the initial phase is overlap-vulnerable, especially after turn-transition, e.g. through laughter caused by prior talk or overlap of two competing speakers that self-select for the next turn, speakers may reemploy the beginning of a structural unit in order to avoid it from being unattended by the addressee. In this sense, repetitions are everything else than functionless performance “errors.” A second reason is related to cognitive pressure deriving from real-time speech production: given that speakers act under time-pressure, they try to avoid temporal gaps (as these may be associated with unwanted implications, see above) and rather make preliminary commitments to a particular constituent. This would imply that speakers are more or less aware of the preliminariness of first constituents and count on the possibility of suspending speaking right after their production. The idea of a “premature” production of a constituent before decisions on the design of the structure and lexical choices have been completed is at the heart of Clark & Wasow’s (Reference Clark and Wasow1998) commitment hypothesis, which gave rise to several testable predictions on when and how speakers produce disruptions and restarts.

Dysfluencies become less frequent in the final phase of the production of a unit of talk. The reason is that during the course of the further production of a unit of talk, once a microgrammatical unit has been initiated, the variety of alternative developments becomes smaller until toward the closure of projections the final components are increasingly predictable so that speakers are more and more unlikely to revise an emergent structure that has become identifiable as a pattern of a particular kind. For the listener, the various alternatives that had been “in play” at the beginning of a structural trajectory are reduced over time until the ultimate structural format can be identified with reasonable certainty, which allows for the anticipation of a possible point of completion at which a transition of speaker roles is legitimate. The predictability allows not only for smooth turn transition to the next speaker with minimal time gaps, but also for early starts (Jefferson Reference Jefferson1973, Reference Jefferson, D’Urso and Leonardi1984), that is, overlapping talk at the potential end of a structure-in-progress. In such cases, a first speaker brings a structural unit to completion by producing highly predictable elements, while a second speaker already sets in at the point at which the production of further elements is almost entirely predictable and initiates a new unit of talk.

4.4.2 Medial Fields at Later Points in Time

It has become clear at several places in this study that understanding the internal structure of a unit of talk in grammatical terms begins with an analysis of what is in it, that is, by identifying and locating subunits, which may belong to the domain of microgrammar or macrogrammar. In other words, one needs to lay open the internal organization of units of talk and associate the different segments one may encounter with the different kinds of things the speaker is doing with the production of these segments.

The assumption that units of talk are internally organized in smaller segments allows the analyst to deal with even seemingly chaotic units, such as the ones in lines 20–22 in (39), where B is talking about how she told her grandmother that she had just been left by her boyfriend.

1. (39)

   16	B:	and she was just saying=
   17		=oh yes well you know (.) you must get over it as best as you
   		can=
   18		=it’s not the end of the world °and stuff°.
   19		↑and I said I know=
   20 →		=but uh you know when I was talking to my cousin Emma uhm it was
   		it was <<unclear word>>,
   21 →		so I was you know I said it was well Monday when I went to work
   		I said it was a week two weeks ago yeah °two weeks ago last
   		Monday°.
   22 →		but (.) ↑I kept on feeling wobbly all day you know and I kept on
   		feeling really miserable and thinking but ↓you know.
   23		I thought=
   24		=God this is- if I’m feeling like this because this is the first
   		person that’s really,
   25		you know I meant something to him I was getting quite fond of
   		him.
   [ICE-GB S1A-049]

The macrogrammatical elements (highlighted in gray) reflect the different kinds of actions speaker B is doing while incrementally building up a structural unit in the linear flow of time, separating one action from another. The unit of talk in line 20 begins after a series of macrogrammatical elements in the initial field: the first marks a shift in focus, namely from how she was talking to her grandmother toward the talk she had with her cousin, which is indicated by but, the second one is a planning marker indicating a cognitive resting point required to plan the new sequence (uh), the third one an addressee-oriented marker (you know) that expresses that the addressee will be able to grasp the implications of what was said in the preceding talk and/or of what will follow. What follows is the projection of a biclausal microgrammatical unit with a first part (when … ) that strongly projects a second part (then … ), the first one expressing an event that became the starting point for a projected second event. However, at the transition from the completed first part to the anticipated second part the speaker hesitates, producing a macrogrammatical expression (uhm) in order to deal with planning activities. The second part begins with a repetition (it was it was) and ends with an inaudible part. The medial field that hosts the cognitive planning marker uhm thus represents a major juncture in the production of the unit of talk, linking an already produced segment to an anticipated second segment that is required to resolve the initial projection of a biclausal structure.

In the next unit of talk (line 21), the speaker signals continuation with a conclusion by producing the macrogrammatical expression so in the initial field, which indicates that what follows is directly contingent to what was said before. This is followed by the beginning of a microgrammatical unit (I was), which makes a subject complement expectable and is likely to express the speaker’s emotional state. However, B changes her plans midway, dropping the further development of the unit and rather alluding to shared understanding (you know) before continuing with a different kind of information. A further, major juncture in the production of a microgrammatical unit occurs at the point at which the speaker is about to express a state (I said it was), but leaves the expectable subject complement unexpressed and shifts to background information (Monday when I went to work), the shift being marked by well. The next juncture in what is to become a microgrammatical unit is caused by a self-correction (it was a week two weeks ago), which goes almost unmarked (no cut-offs, hesitation or a filled pause) and reflects the real-time planning process in which B has to change the deictic perspective from last Monday to the moment of speaking. The repair is followed by a macrogrammatical expression of self-confirmation and deferred realization (yeah), which epistemically upgrades the new piece of information two weeks and strengthens the repair.

The unit of talk in line 22 in (39) is marked as focusing on a new aspect by means of initial but, which signals that something disjunct is being done: B is now focusing on her emotional and physical state, which is described as feeling wobbly. The expression of this state is followed by you know, which forms a final field as it occurs at a potential completion point, but is followed by a continuation that is interrupted at a point at which B is about to refer to her thoughts ( … and thinking). At this point, a shift of focus is indicated (but), followed by a marker that refers to shared understanding of the implications (you know), both of which are ambiguous between forming a medial field or a final field indicating potential completion from a process-based perspective, as can be concluded from the terminal pitch.

While, from an analytic perspective, the interpretation of macrogrammatical units is usually retro-constructed from the final speech product, a speaker- and listener-based perspective understands these units as documenting the kind of cognitive tasks a speaker is dealing with at particular points in time in the incremental construction of talk, and the kinds of projections s/he is creating when upcoming talk is yet unknown. The few, rather coincidental observations based on (39) above illustrate the relevance of macrogrammar for the organization of ongoing communicative tasks in terms of what has preceded and what will follow. Most of these tasks are recurrent generic jobs (Schegloff Reference Schegloff, Ochs, Schegloff and Thompson1996: 81), which are made relevant by prior talk or derive from how talk-in-progress proceeds. Some of them are highly relevant in the course of the production of an utterance and thus in medial fields, such as referring to shared understanding, highlighting the implications of what the speaker is about to say, or indicating a shift to a new piece of information midway. What follows from an analysis of this kind is that understanding the internal structure of a complex unit of talk is impossible without understanding the sequential organization of a speaker’s minor and major actions. The macrogrammatical expressions occurring at several places within a unit of talk were shown to mark major junctures in the continuation of a structure-in-progress, and are thus everything else than randomly occurring “fillers.”

Speakers need not only establish internal coherence within a single unit of talk, but also need to link their utterance in various ways to the prior speaker’s talk and thus to the sequential context in which it is produced. Backlinking may occur within (rather than at the beginning of) an emergent unit of talk when speakers need to deal with a concurring task beforehand. An example is (40), where A and B talk about one of A’s favorite books.

1. (40)

   39	A:	[I      ] think it’s a really good book,
   41	B:	[uhm]
   42		and do you read the book over and over again?
   43 →	A:	WELL (.) after a certain length of time [yeah ] I would imagine.
   44	B:	[yes   ]
   43	A:	because it was really (.) one that always sticks in my mind.
   45	B:	°right°.                                             [ICE-GB S1A-016]

In line 43, speaker A first deals with a restriction of the propositional content of B’s information-seeking request before directly answering it, stating that he reads the book again after a certain time has passed. This prepositional phrase is followed by a marker of affirmation (yeah), which serves as a direct response to B’s request at a point at which the modification of the content of the question has been completed. A then continues with a clausal unit that expresses a somewhat mitigated confirmation of the content of B’s request (I would imagine) and with an explanation of the state of affairs. As a macrogrammatical device, yeah in the medial field serves the task of providing a straightforward response to the co-participant’s request for information and, as a side effect, also serves the integration of the unit of talk produced thus far into the sequential context, marking it as directly contingent to B’s utterance. Moreover, it marks a boundary in the utterance-in-progress as it represents the transition from specifying information that is relevant for responding to B’s request to a follow-up that partly mitigates the confirmation and weakens its epistemic value (I would imagine).

The ability to integrate different segments into one coherent unit of talk is particularly important in cases where syntactic segments that are morphosyntactically independent are aligned consecutively, as illustrated in (39) above, and in (41). A advises B, who is planning to teach English at an institute, to consider important aspects such as the age group B would be comfortable with. In line 189, speaker A starts producing a microgrammatical unit that projects a biclausal structure expressing two alternatives in which the first part involving a subject–operator inversion (would I like to work) strongly projects completion by a second part (or not). However, the production of the second part is delayed as the speaker engages in a specification of the first part.

1. (41)

   188	A:	but I susPECT primary might be °more likely°.
   189 →		but then you’d have to think would I like to work with that age
   		range=
   190 →		=you might well you might say yes that’s such a you know the the
   		the five six till (.) ten eleven sort of age a::nd or no:th (.)
   191		that’s another thought.
   [ICE-GB S1A-033]

The divergence from the syntactically bipartite frame opened up with the production of a subject–operator inverted syntagma would I like to work with that age range is marked by well, which occurs somewhat delayed after the beginning of an intervening segment (you might well you might say, line 190). The next juncture occurs after A initiates what can be interpreted as the beginning of an evaluative construction (that’s such a … ), which projects an evaluative expression in the format of a lexical noun phrase that is, however, not delivered. The suspension of the continuation is marked by you know, which includes the addressee into the discourse as it alludes to shared understanding. The segment that follows (the the the five six … ) does not syntactically fit to the prior segment in the sense that a definite noun phrase cannot be morphosyntactically embedded in a such a construction. Rather, it forms a structurally autonomous unit that fits only loosely to the prior part and, in terms of its internal structure, represents an ad hoc formation analyzable as a phrasal compound (the five six till ten eleven sort of age). After the completion of the noun phrase – a microgrammatical unit with internal hierarchical relationships – the speaker projects a continuation with and that is followed by the production of the second part of the bipartite structure that was initially opened up (or not), which closes the unit of talk as a whole.

The example in (41) illustrates how syntactic junctures arise from ad hoc shifts in ongoing speech activity, which do not only occur after the completion of microgrammatical units, but may also temporally suspend the production of such a unit in order to attend the production of other segments. Such temporal suspensions of a segment-in-progress typically arise from competing communicative tasks that the speaker considers necessary to perform, e.g. the expression of two competing thoughts. This is also illustrated in (42), where speaker A talks about a flat he intends to buy. While producing his utterance, A deals with two concurring tasks, the expression of a “problem” related to the flat (=the lack of availability of public transport) vs. the indication of continued interest in it despite the problem, provided that the price will be the one that he offered.

1. (42)

   207 →	A:	the only problem with that Estelle Road place. (.) >I mean I’d
   		still be happy to have it at the price that I offered.< (.) uhm
   		(.) is the other thing that we’ll need (..) about all those
   		places=
   208		=and that’s the: uhm, (..) actual public transport convenience.
   		(..)
   [ICE-GB S1A-023]

In line 207 the speaker shifts from one communicative task (introducing the problem) to another (expressing his continued interest: I’d be happy to have it). Since both could be interpreted as going against each other (the problem might be interpreted as a possible counterargument to buying the flat) the speaker needs to attend the task of making explicit that he is nevertheless interested in it rather early before misunderstandings arise. This dual-task situation is iconically reflected in the structure of the unit talk in line 207, which begins with the introduction of the “problem” in the format of a projecting phrase that anticipates focal information, but whose completion with the projected predicate is temporarily suspended in favor of the expression of A’s general attitude toward the flat (I’d still be happy to have it … ). This digression, which is delivered with increased tempo, creates a juncture at which the speaker makes use of the macrogrammatical expression I mean, which marks the upcoming information as providing specifying information. The micropause and the planning marker uhm following the complex noun phrase the price that I offered indicate the end of the speaker’s current focus on the intervening task, upon which he resumes the initial syntactic thread by providing the copular verb is that is projected by the initial noun phrase the only problem … At this point, the speaker rephrases the original idea, which is now referred to as a the other thing that we’ll need and interpretable as a structural resonance in that is reflects the syntactic format of the initial projecting noun phrase. The structure is completed with the projected predicate only in the next unit of talk (line 208), which includes a resumptive pronoun (that in that’s the … ) that reactivates the prior syntactic subjects (the only problem/the other thing that we’ll need) and is followed by the statement of the problem (that’s the actual public transport convenience).

As shown above, major junctures occur when speakers carry out ad hoc shifts of a syntactic or communicative plan. They may also result from repairs, as in (43a). Moreover, they generally arise from the incremental character of spontaneous speech, where speakers move from a syntactic segment that may represent a completed unit of talk of its own to a next one. Junctures thus occur e.g. when increments are added to a potentially completed unit of talk, as in (43b), or generally when speakers move from one idea to the next, as is illustrated by the repeated use of you know in (43c).

1. 1. a.

      89	A:	and I think that’s- you know I think people need (.) uhm
      		skills for doing that and uh (.) I think Adam certainly
      		has those skills for for our group (.) uhm
      [ICE-GB S1A-003]

   2. b.

      94	B:	because they were doing it on their own you see.
      95	A:	mm.
      96 →		could they get the right sort of materials (.) you know
      		to make it tone in?
      [ICE-GB S1A-009]

   3. c.

      258	B:	and uh whenever you went and stayed in hotels abroad and
      		things you know he’d be worrying about whether the tissue
      		box was on the right side of the bed and whether you know
      		why the flowers weren’t the right color and all that
      		nonsense you know and sort of (.) busy dispatching people.
      [ICE-GB S1A-010]

It is precisely at such junctures that medial fields and thus macrogrammatical expressions available for use in such fields occur. These expressions establish textual cohesion within an emergent unit of talk in the continuing phase of utterance production. In order to find empirical support for this observation, a corpus-based study of two expressions that appear to occur regularly at the juncture between segments within a unit of talk or between two subsequent units of talk produced by the same speaker has been conducted, namely I mean and you know. Each of the expressions indicates a different kind of link between the neighboring segments: I mean typically indicates continuation with a reformulation, a specification, a conclusion, an assessment, an explanation or additional details (Swan Reference Swan1997: 151), that is, the expansion of an idea (Schiffrin Reference Schiffrin1987: 296), sometimes also in terms of a concession, e.g. after “extreme-case formulations” (Couper-Kuhlen & Thompson Reference Couper-Kuhlen, Thompson, Hakulinen and Selting2005); you know usually indicates that the addressee is able to grasp the implications derivable from what has just been said or that what follows is assumed to be plausible for the addressee or shared knowledge (Schiffrin Reference Schiffrin1987: 268–270). The empirical study distinguished five different contexts of use, which are defined and exemplified in Table 4.4 below. The contexts under (ii)–(iv) belong to the concept of the medial field.

The figures in Table 4.5 indicate the total frequency of the two macrogrammatical expressions in the five different contexts and the frequency relative to their overall uses in percentages for the first 300 tokens of each expression in the spoken sections S1A of the ICE-GB.Footnote ³

The figures allow for the following conclusions. First, I mean and you know occur frequently in the initial field where they indicate continuation and a transition from a prior to an upcoming action or bit of information within a turn produced by a single speaker or when a speaker continues a prior action after a brief contribution by the co-participant, that is, at a juncture created by a co-participant’s intervening speech activity, as illustrated in (44).

1. (44)

   44	B:	>are you going to go to all< uhm the day on- of the phonology lectures.
   45	A:	°I think I ought to [do that.]°
   46	B:	[yes,     ]
   47		I think you had.
   48		ye:ah.
   49 →	A:	I mean [I don’t know how much I’ll take in.]
   50	B:	[I think I’ll go to most of them.       ]
   [ICE-GB S1A-005]

By means of I mean in the initial field in line 49 speaker A indicates that she is continuing to talk about the question of whether she will attend the phonology lectures, thus bridging the upcoming utterance to the piece of talk (line 45) she had produced before B intervened, and marking it as an elaboration of what she has said before.

The second and perhaps most important conclusion deriving from the data in Table 4.5 is that the occurrence of I mean and you know in a unit of talk is not random. There seems to be a division of labor between both expressions, I mean predominantly bridging two subsequent units of talk in the initial field, you know smoothing out junctures within units of talk and gaps occurring through planning pauses. This division can be explained with the function of these expressions: while I mean projects a continuation with a specification or reformulation, which is a task that speakers typically deal with in the initial phase of the production of a unit of talk, you know indicates that the addressee is assumed to be able to grasp further implicatures or the relevance of what is said, which is a task that becomes relevant either at the beginning of a new unit of talk that continues a speaker’s turn (referring to prior talk), or while a unit of talk is in progress as a means to monitor the addressee’s interpretation of emergent talk. Given that you know indicates the speaker’s assumption that there is no significant discrepancy between the mental world of the speaker and that of the addressee (Schourup Reference Schourup1985: 102), the conversational behavior of the addressee may influence the further course of talk, e.g. making further explanations unnecessary.

To conclude, both expressions are predominantly used to bridge transitions in ongoing talk from one segment to another and creating cohesion by smoothing out the potential disintegration of syntactic segments that results from the linearization of various pieces of talk into an emergent structure, or from turn-taking and thus between prior talk and an upcoming continuation of a turn. Repair activities are among the least systematic time slots in which these expressions are used, but considering their frequency of occurrence in all other contexts they represent a minority.

Generally, the points at which syntactic junctures occur are, of course, not predictable. From the speaker’s perspective, such gaps occur in ad hoc ways, which explains the instantaneous character of medial fields. While starting points lend themselves to a relatively context-free description, given that these are absolute points in the production of a structural unit, what happens when a unit of talk is in progress is relatively difficult to conceptualize in an abstract way as the degree of segmentation depends on a larger set of idiosyncratic factors. Some of these are e.g. the schematicity of a format chosen by the speaker (certain formats facilitate the mental segmentation of a message, and thus real-time planning, such as temporal sequentiality, e.g. while x–y, when x–y, or cause–effect relations), or the frequency of talking about a certain topic. It has been observed in experiments that speakers produce more dysfluencies when describing an abstract scenario they had not described before than when describing one they had previously described (Arnold, Fagnano & Tanenhaus Reference Arnold, Fagnano and Tanenhaus2003; Barr Reference Barr, Cavé, Guaïtella and Santi2001) and that speakers have more difficulties introducing new items into the discourse than talking about old ones (Almor Reference Almor1999). Mental schemas are thus a powerful resource affecting the structural design of talk.

What we can conclude from the examples above is that medial macrogrammatical fields are likely to occur at junctures or gaps within an ongoing unit of talk. Major junctures and gaps originate from:

1. (i) syntactic planning that did not go far enough, or a loss of the syntactic “thread,”

2. (ii) word-finding problems and formulation difficulties,

3. (iii) transitions from one microgrammatical segment to another,

4. (iv) dealing with concurring tasks, such as forwarding a message while providing background information required for the listener to be able to understand,

5. (v) repair activities, which need to be carried out in situ, that is, immediately adjacent to the repairable item (Schegloff Reference Schegloff1997; Sidnell Reference Sidnell2010: 111).

(i)–(v) are phenomena that speakers need to deal with instantaneously and that arise in ad hoc ways. Note that many cut-offs of an emergent microgrammatical unit (e.g. in the case of reformulations) are smoothed out by macrogrammatical expressions, thus providing important orientation for real-time processing by the listener, as in (45). With such cut-offs, the co-participants need to mentally overwrite the pieces of microgrammatical configurations that are abandoned underway and not continued at a later point in time (highlighted in gray), i.e. they need to stop the further processing of a syntactic segment at some point in order to restart processing at the beginning of a new segment. In (45), speaker A and B talk about how people usually describe other people, particularly those who they cannot stand, and how people can manipulate other people’s impression by how they talk about an individual.

1. (45)

   268	B:	you think these are these are people that other people are
   		describing as warm loving (.) friendly,
   269 →		you know=I mean it didn’t seem,
   270 →		uh here I am talking to all the-
   271 →		because I’m asking people their own point of vie:w (3.6) I’m
   		obviously (1.9) I mean in a sense it’s a nice position to be in
   		because I’m asking everybody to express what they think (.) so
   		they like me because they think oh here’s kind of subconsciously
   		they think here’s someone who’s prepared to listen to what I
   		think. (1.3) so:, (4.3)
   272		anyway. (1.6)
   273		there’s too much of me °talking°.
   [ICE-GB S1A-037]

Such examples provide evidence for relatively small-scale planning distances that speakers carry out in real-time speech production and for the observation that macrogrammatical expressions occur at caesuras in emergent talk (note e.g. the uses of you know, I mean, uh, because in lines 269–271). An interesting question is in which ways the abandoned segments contribute to the speaker’s message and thus implicitly guide the listener’s understanding, e.g. in the sense that they foreshadow topical elements that turn up at a later point in time in a new microgrammatical format. Thus, even though they are no more than grammatical “junk” from a retrospective (analytic) point of view, they may provide important processing cues supporting the comprehension process and document the speaker’s cognitive activity of “trying out” different formats before committing to a particular structure. For instance, the abandoned segment I’m obviously in line 271 anticipates the expression of a self-evaluation or reflection of the speaker, which is provided later in a different format (they think here’s someone who’s … ) and thus already foreshadows a particular kind of information. This idea surfaces in Ferreira & Bailey’s (Reference Ferreira and Bailey2004) concept of Overlay, which refers to a cognitive operation that allows one structure to be put on top of another, creating a layered structural representation. The top layer is the one that supposedly determines the interpretation, but since the reparandum portion is “underneath,” it can continue to exert some influence (see also Ferreira Reference Ferreira2005: 375).

To sum up, one of the major tasks in the continuing phase of the production of a unit of talk is to integrate smaller segments into a coherent whole in the sense that all segments can be understood as contributing to the illocutionary force and thus to the communicative action the speaker is about to perform. This task is highly relevant given that cognitive constraints (limited capacity of the WM) work against far-reaching syntactic plans and thus often force the speaker into an incremental, segment-chaining style. Since syntactic planning has usually not been completed when the first units of an emergent structure are expressed, speakers need not only be able to create units with internal hierarchical relationships and embedding of all forms (=microgrammar), but also need to deploy strategies to achieve cohesion between smaller syntactic segments.

The different kinds of tasks in the continuing phase discussed in this section are summarized in Table 4.6. The tasks are all associated with macrogrammar.

4.5.1 Conceptualization of the Final Field

The term final field refers to a point in time at which the speaker has brought a unit of talk-in-progress to potential completion and expands it by the production of further expressions. What speakers and listeners consider “complete” is a context-sensitive issue and cannot be predicted out of a concrete speech context or the co-text (Schegloff Reference Schegloff, Ochs, Schegloff and Thompson1996: 87–90; Auer Reference Auer, Couper- Kuhlen and Selting1996b: 61–62) as it does not merely depend on syntactic “well-formedness.” A relatively context-free definition of the potential end-point of a unit of talk can only be given on the basis of one major cognitive operation that guides language production and processing, namely the speaker’s and listener’s ability to draw mental projections, that is, to precalculate possible continuations on various dimensions of the communicative system (Auer Reference Auer2005, Reference Auer2009; Levy Reference Levy2011). Projection is a cognitive mechanism by which one (verbal or nonverbal) action or a part of it prefigures the next, thus creating expectations on the side of the participants on the further trajectory of a (syntactic, prosodic) configuration or a communicative action underway.Footnote ⁴ Projections are based on gestalts, which are schematic patterns created by speakers and identified by listeners that are perceived as an organized whole with the production of more or less well predictable elements in a gestalt-conforming way. Gestalt types in language can be found on the syntactic, prosodic, semantic, and pragmatic level.

Speakers are able to create projections and listeners are able to process projections based on their experience and intuitive knowledge of how gestalts are typically structured or sequentially organized. Knowledge of gestalts allows listeners to identify the points at which these may have reached potential completion on various dimensions. Syntactic projections require knowledge on the serialization of linguistic forms, which includes adjacent serialization (when the form following the one processed so far can be predicted with more or less certainty, e.g. adjacency of determiner and noun in English) and nonadjacent serialization (when the occurrence of a next form can be predicted, but not its exact position, e.g. a segment expressing the apodosis of an initial conditional if-segment) (Auer Reference Auer, Deppermann and Günthner2015). The ability to build up prosodic projections is based on segmental properties, involving knowledge about the possible prosodic patterns of intonation units, which are bounded by a terminal contour and contain at least one nuclear accent and one prominent pitch movement (Couper-Kuhlen & Selting Reference Couper-Kuhlen, Selting, Couper-Kuhlen and Selting1996: 14; Auer Reference Auer, Couper- Kuhlen and Selting1996b). It requires intuitive knowledge of global patterns of pitch movement to and from a syllable with nuclear pitch accent (a segment receiving highest pitch) and on the shape of terminal boundaries in a given language.Footnote ⁵ Determining when the highest or lowest pitch accent has occurred in an emergent utterance requires the co-participants to have some feeling for the pitch range in a concrete speaker’s speech in unmarked contexts (excluding e.g. intonation displaying surprise, anger or contrasts) since, in principle, listeners can make predictions only on the basis of their intuition regarding the limits of a speaker’s pitch contour (Auer Reference Auer, Couper- Kuhlen and Selting1996b: 68–69). Predicting possible completion points would then coincide with the point at which a speaker has reached a pitch limit, i.e. the upper or lower end of his/her intonational range, the first of which foreshadows an upcoming terminal intonation pattern, the latter usually signaling potential completion. Changes in rhythm, volume, and tempo are further resources for prosodic projections. However, the predictability of the number of units to follow a point of maximal or minimal pitch and thus of potential completeness is relatively small. Moreover, global intonation contours are only an ideal case since more often than not speakers assemble smaller accent units in a linear sequence, and it is impossible to predict how many and what kind of accent units or non-accented syllables will follow a final pitch accent. In other words, speakers do not tend to create far-reaching, but rather small-scale prosodic projections, which often do not exceed the limits of a single accent unit. Prosody with its local flexibility to adjust additional units into an emergent intonation contour is thus less suitable for building up far-reaching projections than syntax (Selting Reference Selting1996; Couper-Kuhlen & Selting Reference Couper-Kuhlen, Selting, Couper-Kuhlen and Selting1996: 14–24). Semantic projections relate to knowledge on what it takes for a unit to represent a complete proposition or an “idea,” definable as the point at which a piece of talk can be subjected to a test of their truth value and at which other speakers may express confirmation or negation. Pragmatic projections are based on the speakers’ knowledge on how particular illocutionary acts are usually carried out, and how conversational actions are grouped together and related to one another so as to form sequences. An individual action may be part of or foreshadow another one, as with adjacency pairs (Schegloff & Sacks Reference Schegloff and Sacks1973; Sacks Reference Sacks, Jefferson and Schegloff1995: 521–541), that is, with paired actions in which one action makes another one relevant, e.g. question–answer or offer–acceptance/decline pairs, or preparatory activities (“pres” and “pre-pres”), e.g. story prefaces (Did you hear that about Pete’s car?), pre-invitations (Are you busy tonight?), or pre-requests (Can I ask you a favor?), which project a specific action and are supposed to check on the appropriateness of its production and to prevent the occurrence of dispreferred second actions (e.g. rejections) or the blocking of an action underway (Schegloff Reference Schegloff1988, Reference Schegloff2007). Judgments on what it takes for a conversational action to be interpretable as complete is heavily determined by the sequential context in which it is performed: as Schegloff (Reference Schegloff, Ochs, Schegloff and Thompson1996: 88–89) notes, a simple no can in one context be taken as a completed action (e.g. when it serves as an agreement token with a preceding negative assertion), whereas in other contexts it cannot (e.g. when the speaker is not in the position to agree or disagree with the prior speaker). In such cases, no serves only the initiation of a turn that is expected to express an account.

It is important to note that projection does not determine what follows, but makes expectable a particular kind of continuation. Syntactic, prosodic, semantic, and pragmatic parameters thus serve as orientation cues in the real-time production of an utterance and the synchronization of the listener’s processing with the speaker’s activity. The different kinds of projections interact in complex ways, which cannot be discussed here in full detail. A detailed discussion of this issue has been provided by Auer (Reference Auer, Couper- Kuhlen and Selting1996b, Reference Auer2005) and Ford & Thompson (Reference Ford, Thompson, Ochs, Schegloff and Thompson1996).

Due to the participant’s ability to project continuations, the chances to predict the further trajectory of a structural unit underway continually increase during the production of a unit of talk, until projections on relevant dimensions are complete. This means that in the course of the production of a structural unit the processing effort continually decreases for both the speaker and the listener: while during the initial part of a trajectory participants go through a “phase of maximal planning,” where speakers need to select among various syntactic alternatives and listeners need to simultaneously track several possible trajectories of an emergent structure on various dimensions (e.g. syntax, pragmatics), the number of alternative trajectories decreases over time until in the later phase only a small number of alternatives is available for bringing a unit of talk to completion, many of which are less probable than others so that the remaining parts can be predicted (Auer Reference Auer2005: 9). As I have shown in Haselow (Reference Haselow2016b), a final field arises when the speaker has reached a point in time in the production of a structural unit at which projections relevant for the interpretation of a unit of talk are completed and at which the unit-thus-far is potentially independent of any following unit (t₁). Expressions forming the final field are thus outside the projections holding within the unit to which they are attached. The final field expands a unit of talk and creates a new point of (potential) completion (t₂ in Figure 4.1), without adding new propositional content or opening up new structural projections.

Figure 4.1 Schematic representation of the final field

The sequence produced up to the point in time t₁ in Figure 4.1 is unit-of-talk-constitutive; that is, it can potentially represent a unit of talk of its own. The point of potential completion at t₁ is not to be understood in categorical terms, like the end of a sentence. It is a decision made by the participants, based on whether or not listeners are able to identify the illocutionary point of a unit of talk underway, and thus the kind of action a speaker is performing, and whether or not speakers need to deal with the resolution of open syntactic, prosodic, semantic, and pragmatic projections.

However, it is not necessary that projections on all levels are completed in order to understand a structural unit produced until t₁ as complete: participants tolerate the incompletion of projections on different dimensions to a certain extent if the context and the co-text as well as the unit produced thus far provide sufficient interpretive cues. Syntactic completion, for instance, is often secondary; that is, speakers often do not complete projections on the microgrammatical level when the unit-in-progress is advanced to such an extent that it allows for the identification of the speaker’s communicative goal. The reason is that a syntactically “incomplete” structure can usually be interpreted against the context, co-text, or knowledge on how a particular action is usually performed verbally, and thus allows for the identification of the kind of action a speaker intends to perform without syntactic completion, as illustrated in (47).

1. (47)

   65	B:	I understood that it’s not out on VIdeo.
   66	A:	no.
   67	B:	I just wondered if you’d seen it.
   68	A:	yes=
   69		=I have.
   70 →	B:	and you- >OH you were the one who-,<
   71	A:	I was the one who’s seen it.                 [ICE-GB S1A-006]

B’s sudden realization of the fact that of all people it was A who had already seen the film is syntactically “incomplete” in the sense that it lacks the continuation of the relative clause that is projected by the relative pronoun who. Yet, the unit-thus-far is interpretable for A as the expression of an instantaneous change of B’s knowledge state. Since there are not very many syntactic alternatives in play after the production of the relative pronoun who (a finite verb complex and an object, e.g. has seen it, are the most likely candidate items to follow) the continuation is highly predictable and thus not necessarily required in order to identify the action carried out by B. Speaker A is thus able to confirm B’s deferred cognitive realization by reproducing and expanding the syntactic format initiated by B. Note that the unit in line 70 also leaves projections on other levels open: it is, for instance, prosodically incomplete as it lacks a terminal contour, ending with a cut-off and mid-level pitch, the latter of which projects more to come.

There are units which, if considered out of their co-text and context, appear “incomplete” in the sense that they project more to come, but which are not heard as leaving projections on any level unresolved within their individual context. An example is provided in (48), where actually, which forms the final field of the unit of talk in line 265, is attached to a bare noun phrase.

1. (48)

   262	D:	uhm <<piano in the background>> (7.5) could you just play it
   		through so we know what it’s (.) exactly cos >I’m not confident
   		that I know what it’s supposed to sound like.<
   263	A:	it’s pretty.
   264		it’s nice.
   265 →		a good chord actually. (.)                                [ICE-GB S1A-026]

If produced and heard in isolation, the noun phrase a good chord could be heard as projecting more to come: units beginning with a noun phrase are most likely to be followed by a verb in English since initial noun phrases are typically subjects, or they represent a “left-dislocated”/“fronted” noun phrase that is typically followed by a co-referential pronoun (e.g. a good chord – that’s what we need). However, given the co-text the unit is syntactically not leaving a projection unresolved as it follows a sequence with two copular constructions (it’s pretty, it’s nice), which can be considered as cognitively activated when the speaker produces the unit of talk in line 265. The noun phrase a good chord fits to the copular pattern produced in the preceding units of talk as it may also serve as a subject complement of it’s, parallel to the preceding adjectival complements. Semantically, the NP is non-projecting, as it fits to the list of positive assessments given by A before, providing a nominal reference to the assessment. Pragmatically, the unit of talk is also clearly interpretable as contributing to the sequence of units that express the speaker’s attitude, that is, to an action that can be accounted for, paraphrased, or questioned, and thus not leaving any projections unresolved. The final field represented by the FP actually is thus a point in time at which the unit-thus-far has reached a potential point of completion in the given sequential context. Such examples show that units of talk are usually not uttered as autonomous acts, but interact with prior aspects of talk and are deeply intertwined with the co-textual and contextual environment (see Linell Reference Linell2009: 296).

Identifying potential completion points is thus a complex issue, requiring the simultaneous tracking of several kinds of projections and the sequential context, and not tied to the rules of syntax alone. The role of syntax is important in so far as it is one important dimension on which projections are build up and completed, thus guiding the speaker’s planning activities and the listener’s anticipation of the further trajectory of a unit of talk-in-progress. It seems intuitively plausible that pragmatic projection is the crucial factor in deciding about the potential completeness of a unit of talk in the sense that unresolved projections on other levels can be tolerated as long as speakers have provided sufficient cues as to make their communicative goal identifiable. The relative weight of different kinds of projections in utterance interpretation requires further investigation.

4.5.2 Macrogrammatical Expressions in the Final Field

The final field does not only host single-word units such as final linking adverbs/final particles (e.g. then, though), but also more complex ones, e.g. independent if-clauses (e.g. if you don’t mind me saying), general extenders (e.g. and stuff, and things like that), or parentheticals (e.g. I think, I believe). Such units do not obey preconceived ideas of syntactic units based on lexical compositionality and exhibit a reduced semantics where only in some cases residual components of the original semantic content of the individual parts are identifiable. They form conventionalized form–meaning pairings or chunks with little or no internal compositionality and functional values that are not derivable from the meanings of the component parts.

Note that, as mentioned above for expressions in the initial field, using formal criteria based on sentence-grammar for classification is problematic since almost all of the expressions in the final field violate many of the defining features that characterize the lexical or syntactic category they seem to correspond to according to their form. Units like if you want or if you don’t mind me saying, for instance, are formally affiliated with and thus “look like” if-clauses as they seem to exhibit a subject–predicate structure introduced by the subordinator if. However, the classification as a clause is misleading since such forms are not interpretable as genuine clauses introduced by a subordinator: they do not represent a free combination of individual constituents, but relatively specific constructions with their own pragmatic, rhetoric, and discursive properties (see Brinton Reference Brinton2008: 163–166); semantically, they do not express a condition under which a state of affairs expressed in another clause is valid, and they are neither grammatically, nor semantically representing a subordinate clause (no embedding into a matrix clause, no semantic dependence on the proposition of the prior clause).Footnote ⁶ Another example are units known under the label “general extenders,” such as and stuff or and things like that, which appear to be interpretable as coordinated noun phrases from a sentence-based perspective but, again, represent formulaic units whose meaning cannot be generated by rules of compositionality. Their structural isolation is reflected in the frequent mismatches between the morphosyntactic and semantic properties of the antecedent unit and those of the generic noun that is part of the extender, e.g. you have to speak clearly and stuff (Pichler & Levey Reference Pichler and Levey2011: 444–445).

As with expressions in the initial field, the terminology used here is largely based on formal categories, which are important for identifying the different unit types among analysts, in combination with the specifying adjective final (e.g. final independent if-clause, final comment clause), which is supposed to indicate the syntactic distribution and the functional values that the respective expressions have in common. An overview of the different expressions in the final field, based on the ICE-GB, is provided in Table 4.7.Footnote ⁷ The order of the expressions reflects increasing internal complexity, ranging from low complexity (a single lexeme) at the top of the table to higher degrees of complexity (a more complex, “clause-like” construction) at the bottom. The functional values given in Table 4.7 are discussed below.

Expressions in the final field are only interpretable with reference to the unit to which they are added and for which they typically provide an interpretive cue: they have no illocutionary force of their own and they do not represent a new conversational action, but are part of the action performed with the prior unit. They often improve the perceptibility of transition-relevant places (TRPs) as they are conventionalized in this position and thus create “marked” turn endings (Sidnell Reference Sidnell2010: 154) or endings of units of talk. Such elements therefore play a crucial role in the organization of sequentiality and conversational interaction. The different expressions in the final field form a temporary subsystem of grammar (Hopper Reference Hopper and Tomasello1998: 158) that is continuously reorganized and thus in flux, as the development and variation in the use of tags (e.g. tendencies toward invariant innit?), general extenders (see Pichler & Levey Reference Pichler and Levey2011), or final particles (Kim & Jahnke Reference Kim and Jahnke2011; Haselow Reference Haselow2013; Traugott Reference Traugott2016) shows.

The following analysis of the expressions in Table 4.7 in their individual contexts of use seeks to identify the core functions of the final field, which are indicative of the communicative tasks that speakers typically deal with at the end of an emergent unit of talk.

4.5.2.2 Linking Adverbs/Final Particles

The elements in this category comprise the lexemes actually, anyway, but, even, so, then, and though used at the end of a unit of talk. They have been variously categorized as conjuncts (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 631–647), linking adverbials (Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 889–892), connective adjuncts (Huddleston & Pullum Reference Huddleston and Pullum2002: 779), pragmatic markers (Traugott Reference Traugott2016), or final particles (Mulder & Thompson Reference Mulder, Thompson and Laury2008; Haselow Reference Haselow2012a, Reference Haselow2013; Izutsu & Izutsu Reference Izutsu, Izutsu, Hancil and König2014; Hancil, Haselow & Post Reference Hancil, Haselow and Post2015). The preference for the term final particle (FP) is based on the observation that they do not form part of a microgrammatical unit and therefore do not conform to the class of adverbs/adverbials in the categorical system developed within sentence-grammatical approaches in the strictest sense, as they are not related to another constituent with which they maintain a morphosyntactic and semantic relationship and whose properties they modify. Moreover, in final uses these lexemes have no propositional content and no effect on the semantic content and the truth-value of the preceding unit as they are deletable without a reduction in meaning and of the validity of the unit they accompany. They have a peripheral relationship with the unit they accompany and they can connect units of different size, from lexical units and clauses to larger chunks of text. Analyzing FPs as linking elements or connectors stresses their function of establishing a relation between two units of discourse. However, FPs also interact with the illocutionary force of an utterance, e.g. as mitigators of other-repair (final though, Couper-Kuhlen & Thompson Reference Couper-Kuhlen, Thompson, Hakulinen and Selting2005) or intensifiers of information-seeking requests (e.g. anyway, Lenk Reference Lenk1998: Ch. 4), which justifies the label pragmatic marker (Traugott Reference Traugott2016). It is, however, debatable if these elements operate in the pragmatic domain only and, if so, where to draw the boundary between grammar and pragmatics with respect to spoken discourse (see Ariel Reference Ariel2008; Haselow Reference Haselow2013).

All FPs have a lexical counterpart in non-final use. However, in final use they share a set of functions that justify a classification of the respective lexemes as a category of their own. Outside the final field, these lexemes are typically conjunctions (but, so), adverbs (e.g. then, actually), or scalar focus particles (even) (see Haselow Reference Haselow2013 for an overview).

The categorical value of FPs in spoken English is that they mark a unit of talk as reactive to a preceding one, which has usually been produced by the interlocutor. However, the two adjacent units they link may also be produced incrementally by the same speaker, e.g. when speakers correct inferences that could mistakenly be derived from what they have just said (e.g. I know her. I can’t remember her name though.) or with utterances expressing afterthoughts (e.g. I’m glad she’s gone. I didn’t like her anyway.). In any case, FPs indicate a particular kind of sequential relationship between two subsequent units of talk, namely the one they accompany and the one preceding it, as illustrated in (52).

1. (52)

   a.	320	A:	ZARA’s a Marks and Spencer’s person °aren’t you° (.)
   	321	B:	well it’s easy to buy clothes and (.)
   			[rubbish °like that° .           ] (..)
   	322 →	C:	[they’re all right ↓then.]
   	323	B:	I only buy my (.) from there. <<laughs>>
   (ICE-GB S1A-017)
   b.	212	B:	d’you want me to put anything like cheese on toast on.
   	213		°or anything you’ve had lunch already.°
   	214 →	A:	I wouldn’t mind some coffee actually.
   	215	B:	coffee.                (ICE-GB S1A-045)

In both cases, the FP retrospectively relates the unit of talk it accompanies to a preceding one. In (52a), the unit of talk they’re all right is conditionally related to A and B’s utterance, from which C derives a particular conclusion. More precisely, C’s utterance can be analyzed as pragmatically expressing an inference that awaits confirmation and that is drawn from the content of A’s utterance and the illocutionary goal of B’s utterance, which are pragmatically rank-shifted to conditional protases whose premises are not hypothetical, but given (premise conditionals). Thus, final then triggers a reconfiguration of C’s utterance as the apodosis: “if, as is the case, U1_A (=‘Zara is a Marks & Spencer’s person’) and U1_B (=B provides another argument for buying at this store): U2 (=‘the clothes are all right’).” This way, C justifies the production of her utterance by marking it as the second part of an implicit conditional construction. In (52b), A’s response diverges from the premises of the request produced by B: while B refers to food for lunch in the first way, A is asking for a beverage, marking the divergence from an expected response by means of final actually. Note that in both cases the syntactic features of the FPs are not determined by any structural constraints holding within the unit to which they are attached. In this sense, FPs are clearly outside microgrammar, representing a final field. They do not participate in the internal organization of the structural unit to which they are attached, but create a two-place relation in discourse, relating the unit to which they are attached to prior talk.

FPs are strictly backwards-oriented and can be conceived of as components of a schematic pattern of the form “U1–U2+FP”: the pattern consists of a unit of talk (U1) that is followed by a second, reactive unit of talk (U2) that includes an FP, which establishes a particular kind of relationship between U1 and U2 (Haselow Reference Haselow2013: 384). U1 and U2 are typically adjacent and form a pair in which the content of U2 has been motivated by and is conceptually dependent on the content of U1. The retrospective linking function of FPs is represented in the abstract discourse schema in Figure 4.2.

Figure 4.2 Abstract discourse schema for final particles

FPs are paradigmatically organized and thus exhibit internal subdifferentiation, as shown in the example in Figure 4.2: replacing one FP by another one attached to a unit like he is right changes the type of rhetorical relation holding between two subsequent discourse segments. Final actually indicates that the content of the second unit of talk U2 (he is right), which expresses a confirmation of the hearsay knowledge expressed in the preceding unit U1, is potentially unexpected for the addressee; final anyway (if stressed) marks U2 as valid irrespective of U1, that is, as a kind of general truth that he is right no matter what has been said or implied before; final but indicates that U2 is dissonant to another, implied proposition that weighs heavier and thus assigns it a lower communicative value; even marks the assertion he is right as ranking low on an implied scale of expectability and signals that, on top of what he said (that she is a bad teacher), the fact that he is right is especially remarkable and noteworthy; so points to an implied conclusion related to U2; then indicates that he is right is a conclusion inferable from U1 and conditionally related to it; though signals that U2 is true in spite of an implied opposite view. This kind of paradigmaticity is one of the core features of grammatical categories (Diewald Reference Diewald2011) and an important indicator for the fact that FPs form a category of their own with clearly definable core functions and formal properties that determine category-membership. FPs have a framing function that spans two adjacent units of talk, U2 expressing information that is contingent to the information provided in the preceding unit of talk U1. They play an important role in the creation of coherent discourse produced in real time where syntactic disconnectedness, caused by the successive production of structurally isolated units as a result of incremental, real-time speech production and speaker change, is ubiquitous. The unavailability of structural embedding with speaker changes or the successive planning and production of smaller syntactic segments leaves linear-temporal sequentiality and the use of conventionalized linking devices as the only possibility for structural linkage. In this sense, FPs, like other macrogrammatical expressions, repair much of the structurally unintegrated character of spontaneous spoken discourse.

An important cognitive aspect that motivates the use of FPs in real-time speech production is thus that they allow a speaker to establish a relationship between two subsequent discourse units without prior planning of this relationship. In same-speaker turns, speakers are not required to have two ideas in mind before they produce an utterance, and to syntactically embed one of the ideas into the other, but may integrate later information into information expressed before in a post-hoc way. This view is also implied in Biber et al.’s (Reference Biber, Johansson, Leech, Conrad and Finegan1999: 851) explanation for the much higher occurrence of concessive final though in spoken texts compared to all written text types:

[a] possible factor at work is that considerable forward planning is required to construct a sentence with a concessive adverbial clause. The speaker has to have two propositions in mind, together with a realization that one runs counter to the other, before starting to speak. This is not easy to manage in the online production of speech.

Limitations in speech planning thus seem to be responsible for the use of final though in spoken language, as its use allows speakers to produce two propositions, both expressed as structurally independent units, subsequently in a linear sequence, and to mark the dissonant relationship between both propositions after they have been produced. This strategy is, for instance, relevant when speakers need to modify a statement already made, e.g. restricting its validity or canceling a possible inference derivable from it. This is the case in (53), where speaker A tells C that he wants to change the format of his CV, which currently extends over two pages, to have it on one page. C offers A to help him changing the format by converting the document on his computer.

1. (53)

   178	A:	I would like it done on Wednesday if possible. (.)
   179	C:	mm.
   180	A:	please.
   181	C:	.h well=if you give me it tomorrow I might be able to do some
   		tomorrow [morning before I go to rugby. ]
   183	A:	[that would be excellent.        ]
   182 →		I can only do an hour though.                 [ICE-GB S1A-038]

In (53), speaker C produces an assertion (he might be able to do some [converting] tomorrow morning), which might lead A to derive a particular inference or conclusion from it, namely that, in case C is dealing with it tomorrow morning, he will have no particular time limit for doing it, apart from the fact that he will do it before he goes to rugby. The example illustrates that speakers may become aware of possible, unwanted inferences during or after the completion of an assertion: in the next unit of talk (line 182), C limits the time frame available for dealing with A’s CV to an hour and links the unit to the preceding one by means of final though, thus canceling the possible inference that he can deal with the CV the entire morning.

FPs exemplify the importance of a dualistic approach to grammar as they are not integrated into the binary, hierarchical relationships holding with the structural unit they accompany. As shown above, both their functional description and their assignment to a particular grammatical category appear difficult on the basis of sentence-based grammars. Given the principles of macrogrammar discussed in Section 3.3 – all of which apply to the use of FPs – they can be treated as a category of their own in a grammatical domain that deals with the organization of language outside binary morphosyntactic and semantic relationships.Footnote ⁹

4.5.3 Expressions in the Final Field as Macrogrammatical Elements

The different kinds of expressions in the final field do not qualify as constituents in the traditional sense of sentence grammar: they are not morphosyntactically integrated into hierarchical relations with other lexical, phrasal or clausal units, they do not receive their morphosyntactic and semantic properties from any other form or assign them to another form, and they do not relate clausal constituents to one another. Expressions in the final field tend to form a relatively closed class and are characterized by the following features, which coincide with those of expressions in the initial field and which justify their assignment to the macrogrammatical domain:

1. (i) they are used to organize spoken text rather than creating microgrammatical relationships,

2. (ii) they are morphosyntactically autonomous,

3. (iii) their internal structure is not or only weakly compositional,

4. (iv) they do not contribute to, change, or restrict the propositional content of the unit they accompany; their semantic content is minimal;

5. (v) they serve as processing cues, framing or fine-tuning a preceding message and retrospectively integrating it into a sequential context.

Note that expressions in the final field can occur in combination. A tag question can, for instance, be followed by a general extender, as in (62a), or an FP by another FP, as in (62b).

1. (62)

   a.	112	B:	↑well he has this stupid girl he falls in love with
   			↓°doesn’t he or something°.
   	113	A:	NO::	[ICE-GB S1A-006]
   b.	175	B:	we should=maybe just leave a message here saying head over. (.)
   	176	A:	she won’t bother coming then though.
   				[ICE-GB S1A-039]

The combinability of expressions in the final field is not unconstrained: some combinations seem to be excluded or at least do not occur in the ICE-GB and intuitively appear unnatural, such as an independent if-chunk followed or preceded by a tag question. Moreover, the serialization of units in the final field is often constrained: an FP can, for instance, be followed by a parenthetical, but it is usually not preceded by it (e.g. *He’s not invited I think then). This indicates that the final field has a syntax of its own that requires further investigation.

4.5.4 Functions of the Final Field

The functions of the different kinds of expressions in the final field are very similar, which suggests that speakers deal with recurrent communicative tasks at the potential end of a unit of talk in the linear flow of time. These tasks are related to the fine-tuning of epistemic or pragmatic aspects of a message or a “focus of consciousness” just expressed in order to facilitate the interpretation of a unit of talk, the organization of the further course of interaction (e.g. by prompting the addressee to engage in ongoing interaction), and the integration of a unit of talk just produced into the discourse context. They provide a last procedural cue guiding the addressee’s interpretation of a message before (potential) turn transition or continuation with a next unit of talk. The tasks performed by speakers in the final phase of the production of a unit of talk are summarized in Table 4.8.

The final field is a resource that is used when the interactive past has not already contributed to a solution of the tasks listed above or when the speaker considers the conditions required for the successful processing of a unit of talk-thus-far as not fulfilled. For instance, speakers may notice unwanted effects on different dimensions (e.g. on the content or the interpersonal level) that they wish to remove. From a dynamic, temporal perspective on utterance production the final field is the last opportunity to make whatever adjustments are necessary according to the speaker’s intuition or contextual demands before possible turn transition or continuation with a new idea.