6.1 Introduction

We now have a very strong sense of how the L2 contributions at the turn-level function at the micro-structural level in conversation, and how those functions contrast with those of the L1 examiner’s speech in the TLC (Chapter 2). We have also looked at how that view meshes with what we see at the macro-structural, discourse unit level (Chapters 3 and 4). We can also see how the functions of those discourse units bear some notable similarities to what L1 speakers produce when they perform the same tasks and how the link between the micro- and macro-levels are both connected to and controlled by the linguistic behaviour of the populations under study (Chapter 5). For both L2 and L1 speakers, however, we have the question of the extent to which the exam itself, and preparation for it, is a distorting factor. To some extent we have explored this by looking at the TLC L1 in the previous chapter, satisfying ourselves that grade of exam is probably the dominant reason for any distortion that we may encounter when comparing L1 performance to L2 performance. We also noted that individuals may vary from this overall pattern. However, the question of whether L1 speakers, when not bound by the exam, produce broadly similar functions as those produced by either the L2 or L1 speakers taking the exam that is the basis of the TLC and TLC L1 remains a moot point. To refine the point slightly, the goal of the exam is laudable – to prepare L2 speakers for conversational interaction with L1 speakers, with a strong focus on functional variation by task. However, what evidence do we have that the exam itself actually reflects the demands of such an interaction?

To explore that, in this chapter we turn to the Spoken BNC 2014 and conduct a short-text MDA on it. As we are working with data drawn from a context distinct from the GESE exam, we will explore the Spoken BNC 2014 short-text MDA in more detail than that of the TLC L1. The analysis that follows begins that process with a detailed discussion of a short-text MDA of the Spoken BNC 2014 data at the micro-level of discourse; in this case, the level of the turn. The chapter then proceeds to a consideration of the functions present at the macro-level (i.e. at the level of the discourse unit) before comparing the micro- and macro-structures in the data. Sections 6.2–6.8 provide a brief, detailed account of each dimension with accompanying examples. Readers who are interested in the discourse unit level, and are prepared to accept at face value the dimension labels at the micro-structural level, should skip ahead to Section 6.9.

Following this chapter, with a clear view of how the BNC data is organised at the discourse unit level, we will return to consider the question of how alike, or otherwise, the interactions in the TLC and TLC L1 corpora are to conversational L1 British English.Footnote ¹

6.2 Dimension 1: Elaborated Speech versus Discourse Management

As with previous short-text MDA results presented in this book, Dimension 1 opposes the presence of features on the positive side largely with the absence of features on the negative side (the exceptions on the negative side being positive interjections, laughter and general interjections). The results are similar to the TLC L1 results. The main difference is that there is no clear narrative component in the Elaborated Speech function. As with the TLC and TLC L1, the length of the turn tends to be the greatest predictor of the presence of features; in other words, the more words a turn contains, the more likely the turn is to exhibit the presence of features. Contrariwise, shorter turns are more likely to exhibit an absence of features. Indeed, in this case, the turn most strongly associated with the positive pole of the dimension was 656 words in length, while the turn most strongly associated with the negative pole comprised just 3 words. We included turn length as a quantitative supplementary variable, which correlates the length of each turn to its dimension coordinate. This revealed an exceptionally strong positive correlation to turn length (R = 0.86). No other dimensions were correlated to text length, meaning that turn length has been essentially controlled-for in this dimension. Dimension 1, being related to text length, is typically regarded as an effect of the method in short-text MDA, and so is disregarded from such analyses (e.g. Clarke and Grieve, Reference Clarke and Grieve2019; Clarke, Reference Clarke2022). As with other explorations of Dimension 1, we will set this dimension aside, noting only that it aligns well with the L1 exam data discussed in the previous chapter, and focus instead on the other dimensions in this analysis. However, we will return to discuss this dimension both in Section 6.9 and the next chapter.

6.3 Dimension 2: Interactive Information Exchange versus Attitudinal Descriptions

We interpret this dimension as opposing, on the positive side, turns associated with an Interactive Information Exchange (the process of figuring things out and exchanging information) with those on the negative side, which provide Attitudinal Descriptions. Table 6.1 shows the lexico-grammatical features whose presence is most strongly associated with the positive and negative poles of this dimension.

Table 6.1Features associated with Dimension 2.

The positive pole of this dimension is associated with the presence of several interactive features, such as pronouns, including first-person, second-person and object pronouns, which indicate that the speaker and hearer are ‘involved’ in the discourse. The presence of the pro-verb do also suggests that interaction is taking place, as its use implies a shared communicative context. There are also features used to form questions, such as WH-words and question marks, which suggest the exchange of information. There are features associated with encoding thoughts and stance, such as stance verbs, private verbs and complement clauses, and there are also features associated with reporting speech, such as public verbs. Possibility modals are also present, and these are used to refer to ability and possibility. These features often co-occur in turns in which the speaker is considering a particular situation, or working through a problem or situation and trying to figure something out by referring to what different people know and have said.

The turns associated with Interactive Information Exchange often include an expression of, or a reference to, epistemic stance – that is, what the speaker/hearer or an external third party knows or does not know, as well as an expression of, or a reference to, what has been said. This function also often involves a process of seeking information and enquiring about particular situations, scenarios and courses of action. These can refer directly to the listener and seek their knowledge; otherwise, the questions are reported as being asked of a third party. Overall, these features thus co-occur in turns that have an underlying Interactive Information Exchange function, as shown in the following example, from BNC file S9X9:

By contrast, the Attitudinal Descriptions function is characterised by features that are associated with a descriptive function, including stative forms (i.e. BE as a main verb and copular verb), both attributive and predicative adjectives, and amplifiers and downtoners used to reduce or increase the strength of the description provided. These features co-occur in turns that provide attitudinal descriptions of a subject. The subject of the description is often the pronoun it or a demonstrative pronoun, implying a shared communicative context. Overall, these features co-occur in turns which provide descriptions of a subject, and these descriptions are often attitudinal and express an opinion on that subject, as in the example turn that follows, taken from BNC file SLMB:

Rather than seeking information in the process of figuring things out, then, as the Interactive Information Exchange function does, the Attitudinal Description function is more concerned with providing information and descriptions.

6.4 Dimension 3: Epistemic Stance versus Informational Recounts

Dimension 3 can be interpreted as opposing turns involving epistemic stance-taking on the positive side against those which provide informational recounts on the negative side. As with Dimension 2 previously, the features present in the data that are associated with the positive and negative poles of this dimension are given in Table 6.2.

Table 6.2Features associated with Dimension 3.

In the Epistemic Stance function we find the presence of features such as private verbs and complement clauses, which are often used to encode personal knowledge and thoughts. Epistemic Stance is also characterised by stative forms, predicative adjectives and amplifiers, all of which can be used to encode a personal stance or to provide a description of a subject, as shown in the following example, from BNC file SY4E:

By contrast, the features associated with the Informational Recounts function are, understandably, much more informational in nature. For example, turns associated with this function are characterised by the presence of several noun types, including numeral nouns, proper nouns and general nouns. There are also several noun modifiers, such as demonstrative determiners, definite articles, indefinite articles, attributive adjectives, prepositions, quantifiers and markers of possession. All of these features are associated with an informationally dense style. Additionally, the Informational Recounts function is also characterised by verbs such as have as a main verb, which is used to report on possessions and experiences. Phrasal verbs and general verbs are also present in this function, and these features are used to report on activities. An example of the use of such features to produce the Informational Recount function is given in the extract that follows, from BNC file SA4W:

6.5 Dimension 4: Reveal versus Information Seeking

Dimension 4 can be interpreted as opposing turns with the function of Reveal (specifically of future plans, intentions and desires) with turns linked to Information Seeking, especially with regard to information concerning the past. The features whose presence is associated with this dimension are presented in Table 6.3.

Table 6.3Features associated with Dimension 4.

The Reveal function is characterised by the presence of features that are used to talk about future plans, expectations and desires. These features include prediction modals, stance verbs, infinitives and time adverbs. Perfect aspect is also present in turns associated with the positive side of this dimension, and this is used to talk about experience or to suggest a possible course of action. First-person pronouns and object pronouns are used to mark that the future plans being discussed are the speaker’s plans, as well as that these plans in some way impact the speaker or other people. Example 58, from BNC file SXRR, shows this function.

The features associated with the Information-Seeking function include WH-words, question marks and stative forms. Several features mark a third-person subject, such as third-person singular verbs forms, proper nouns, demonstrative pronouns and definite articles. Finally, past tense co-occurs with the other features, often with the function of asking questions about the past. The example from BNC file SBX7 that follows illustrates the use of this function.

Note that in Chapter 2 we also saw, at the turn level, the Reveal function in L2 English. The features which constitute the Reveal function for L2 are very similar to those for Reveal in L1. The major difference relates to person – in the L2 data, the focus is on object pronouns and second-person pronouns. Here, the focus is on subject pronouns and first-person pronouns. This difference in emphasis aside, the function is almost identical.

6.6 Dimension 5: Narrative versus Non-Narrative

We now move onto Dimension 5, which we interpret as opposing the Narrative discourse function with Non-Narrative turns. The features whose presence is associated with this dimension are given in Table 6.4.

Table 6.4Features associated with Dimension 5.

The Narrative function is associated with the presence of features which indicate narrativity, including public verbs used to report on speech, progressive aspect, past tense verbs, object pronouns and third-person personal pronouns. It also includes proper nouns used to introduce a subject into the narrative, as well as predicative adjectives used to describe a subject, and WH-words and complement clauses which are used to elaborate and extend the narrative description. Other features of Narrative are associated with interactivity, including both positive and general interjections. The following example of this function is from BNC file SKJ3:

By contrast, Non-Narrative is characterised by have as a main verb used to describe possessions, as well as the auxiliary Do and analytic negation, used to negate an action. Other features of Non-Narrative are associated with descriptions, such as downtoners and attributive adjectives. The presence of indefinite articles and general nouns suggests the function may introduce a particular referent that has not been mentioned previously. Second-person pronouns are also present, but these are often used as the generic you, while the pronoun it is used to refer to general situations. These co-occur in turns that are Non-Narrative and which often describe general situations and scenarios. The following example, from BNC file S3SA, shows the function in use:

6.7 Dimension 6: Opinionated Narrative versus Situation-Dependent Commentary

Dimension 6 opposes turns with the discourse function of Opinionated Narratives against those which function to comment on the particular situation in which the interaction is taking place, Situation-Dependent Commentary. Table 6.5 gives the features whose presence is associated with this dimension.

Table 6.5Features associated with Dimension 6.

Opinionated Narrative is marked by the presence of features that are associated with expressions of stance and opinions, including downtoners, quantifiers, amplifiers, stance verbs, private verbs and complement clauses. Contrastive conjunctions are also present in this function, suggesting that a contrast in ideas is being made in turns associated with this function. Finally, the function also includes features associated with recounting, such as past tense verbs and object pronouns. The following example from BNC file SH79 shows this function in use:

Situation-Dependent Commentary is marked by the presence of features that are associated with a shared communicative context, such as the pronoun it, demonstrative pronouns and question marks. Also present in Situation-Dependent Commentary are features associated with a more present continuous and future tense, such as perfect aspect, prediction modals and progressive aspect. This example from BNC file SBCZ illustrates the use of this function:

6.8 Dimension 7: Advisory versus Personal Narrative

The final dimension, Dimension 7, is interpreted as opposing turns with an Advisory discourse function with those that constitute Personal Narratives. The features whose presence is associated with this dimension are given in Table 6.6.

Table 6.6Features associated with Dimension 7.

On the one hand, Advisory turns are advisory and informative. Features such as prediction and possibility modals co-occur in the function in order to provide suggestions and recommendations about a possible course of action. Other features are associated with references to specific entities, such as the use of demonstrative pronouns and demonstrative determiners. These are used to refer to a particular problem or situation in order to inform and to provide advice. Nominalisations are also present and are used to refer to abstract scenarios and situations so as to give advice and to recommend a particular course of action. The following example, taken from BNC file SPXW, demonstrates this Advisory function, with the speaker suggesting that the hearer needs to obtain financial information on a particular industry before going into a particular meeting in order to be prepared:

On the other hand, the Personal Narrative function is not associated with the discussion of future action but is instead marked by the presence of features such as past tense verbs, third-person pronouns, third-person singular verb forms and first-person pronouns. Proper nouns are also associated with this function, indicating particular proper referents in the narratives. Finally, all types of negation co-occur in turns associated with the Personal Narrative function, including analytic, synthetic and negative interjections, suggesting that the personal narratives provide clarification about what did and did not happen in relation to a specific event. The following example, from BNC file SHX2, shows the function in use:

6.9 The Discourse Unit Level View

What happens to these patterns if we vary the view of the data, from the micro- to the macro-level? Looked at from the perspective of discourse functions, the results are remarkably stable. While the polarity of some of the interpretations flips, the set of labels that we can apply to the dimensions, and the number of dimensions that can be analysed, are, for the first six dimensions, similar, as shown in Table 6.7.

Table 6.7The discourse unit view of discourse functions in the BNC 2014, functions with reversed polarity relative to the turn-level view are marked with an asterisk.

Nine functions – Discourse Management, Elaborated Speech, Informational Recounts, Information Seeking, Narrative, Non-Narrative, Opinionated Narrative, Reveal and Situation-Dependent Commentary – are shared between the micro- and macro-structural levels. The functions which sit exclusively at the macro-structural level in this data are the two functions, Informative and Instructive and Seeking and Encoding Stance. The functions Advisory, Epistemic Stance, Interactive Information Exchange and Personal Narrative are exclusively micro-structural functions.

A brief exploration of two functions may help to show how the levels interact. Discourse Management becomes more functionally elaborate at the macro-level in contrast to its realisation at the micro-structural level. At the micro-level, Discourse Management includes a range of features: hesitations, false starts, backchannels and filled pauses, inter alia. However, at the macro-level, these are brought together into an assemblage with a function largely designed to help two speakers manage interaction. Consider the following discourse unit from the BNC data, taken from file S8J6:

Prior to this, Speaker B has been talking about how they celebrate their ‘work birthday’. Speaker A has interrupted Speaker B to point out where a drink is and what ensues is partly focused on that information, but is in fact largely about mitigating the interruption through reciprocal laughter and ending the interruption, through backchannels from A, to permit Speaker B to continue. In other words, Speaker A makes it clear that, other than communicating where the drink is, they have no intention of taking the floor. Accordingly, in the following discourse unit Speaker B starts again to explain their ‘work birthday’ and Speaker A encourages them to hold the floor and to elaborate, providing positive backchannels (‘Oh yeah?’) in response to the comments about the work birthday. On their own, these short sequences are unremarkable, though they are clearly a form of discourse management. Brought together with a macro-structural function to perform, they help to manage the discourse beyond simple backchannelling. In this case the micro-structures are combined to allow, at the macro-structural level, an interruption, and a related conceding of the floor back to the interrupted to occur.

The first example showed us how a function at the micro- and macro-level may interact to permit the function at the macro-level to perform tasks it would be hard to account for at the micro-structural level alone. A second example relates to the Informative and Instructive function—a function which occurs only at the macro-structural level. Consider the following discourse unit from BNC file S2RD:

This is Informative and Instructive precisely because it is a mixture of statements providing information, for example, ‘I don’t even know’, and demands on the hearer, for example, ‘why does she heart Alan’. At the micro-level these are separate, yet the assemblage of turns at the macro-level realises the Informative and Instructive function.

The shifting of functions up and down dimensions, or the flipping of polarity within a dimension, are indicative of the relative change in the rank ordering of these functions. We will look at the issue of functions switching dimensions shortly, as it is not present in the micro- and macro-analyses of the BNC. However, we do see polarities flip – a good example is Narrative, which is positive Dimension 5 at the micro-level, but negative Dimension 5 at the macro-level. This is an indicator that, of the paired functions, Narrative is slightly more dominant at the micro-level in this dimension, but Non-Narrative is slightly more dominant at the macro-level within this dimension. So, in the flipping of polarities, we are seeing the balance of the shared functions play out between the micro- and macro-levels.

Setting aside changes in polarity, the repertoire of functions at the micro- and macro-levels in the Spoken BNC 2014 is quite similar. What may be the cause of this similarity? As we saw in Chapters 2–4, some stability is to be expected when shifting between the turn and discourse unit level view. However, the stability here is striking. Two factors may plausibly explain this. Firstly, the Spoken BNC 2014 is much larger than the TLC – it may be the case that some of the turbulence in the shift of view between Chapter 2 on the one hand and Chapters 3 and 4 on the other is attributable to data sparsity. Secondly, the Spoken BNC 2014 is composed of a much more homogeneous set of speakers than the TLC. While there are sociolinguistic variables in the BNC data that may well be a source of variation – age, sex and social class, for example – the TLC has these variables too, and adds other variables which, as we have seen, do have an impact on the use of discourse functions – proficiency, grade of exam, task and L1 background, for example. So, the BNC data might be assumed to suffer less from any issues of data sparsity and is not impacted by a range of variables which can cause significant variation within the dataset. This observation should allow us to reassess the variation viewed between the analyses in Chapter 2 versus Chapters 3 and 4, inclining us towards interpreting it as being the result of scale of, and diversity in, the data.

At this point we might ask, why does Dimension 7 not appear in the discourse-level view? The most obvious answer to this question relates to the meshing of micro- and macro-level views. Specifically, the functions identified in Dimension 7 are of significance only at the micro-level; they are not constituted at the macro-level. To explore this further, let us see how Dimension 7 at the micro-level contributes to the macro-level. We can do this through a short qualitative study looking at turns coded as being at the extremes of the distribution of Dimension 7, and then by seeing where these appear in discourse units. Our goal in doing so is to find if there is an indication of what type of contribution Dimension 7 at the micro-level makes to functions at the macro-level.

To begin, we can focus on the 100 turns in the BNC which are most strongly associated with the Advisory function in the turn-based view of discourse functions. By this, we mean that we look at the 100 turns which have the highest dimension coordinates for positive Dimension 7. We then look at the discourse units in which they occur and for the function with which they are most strongly associated by looking for the highest dimension coordinate—that is, the coordinate which is placed furthest into either the positive or negative side of the dimension.Footnote ² For example, the turn most strongly associated with the Advisory function has a dimension coordinate of 1.268. It is in a discourse unit with a set of dimension scores (from the first to the sixth dimension) of −0.404, −0.018, 0.174, 0.105, 0.069 and 0.103. So, in this case, we would say that the Advisory turn is appearing in a discourse unit that for the purposes of this experiment we would categorise as Elaborated Speech (negative Dimension 1), as the most negative value (−0.404) is placed further into the negative side of the dimension than the most positive value (0.174) associated with Informational Recounts (positive Dimension 3).

Taking this approach, if we look at how the micro-structural level (Advisory turns) intersects with the macro-structural level, we see that the Advisory turns, all of which are produced by different speakers, are strongly associated with Elaborated Speech function at the discourse unit level. Ninety-seven of the prototypical examples explored occurred within a discourse unit with this function. The other three examples appeared in Informative and Instructive (one) and Reveal (two) discourse units. However, these examples hardly seem worth discussing – it is clear that these prototypical Advisory turns appear as part of a longer stretch of monologic speech, Elaborated Speech, but this function at the macro-structural level is not principally concerned with the Advisory function as realised at the micro-structural turn level. Hence the Advisory function, while closely linked to that discourse function, does not define it.

If we carry out the same exercise on the negative side of Dimension 7, looking at the 100 prototypical Personal Narrative turns, and exploring their distribution in the discourse unit level analysis, we see differences and similarities. In terms of difference, while the Elaborated Speech is still the largest category of discourse unit in which turns of this sort appear (forty-two examples) we see a much wider range of discourse unit functions calling upon turns marked as Personal Narrative, including Situation-Dependent Commentary (fifteen examples), Information Seeking (thirteen examples), Discourse Management (eight examples), Narrative (six examples), Informative and Instructive (four examples), Attitudinal Descriptions (four examples), Informational Recounts (three examples), Seeking and Encoding Stance (two examples) and Reveal (two examples). The similarity is that we do not see a micro-structural function simply translating into a macro-structural function and in this case the possibility of it existed. It may have been, for example, that Personal Narrative at the micro-structural level is always contained in a Narrative discourse unit. But this is not the case – Personal Narrative in these prototypical examples is present much more frequently in discourse units which are not identified as having, overall, a Narrative function.

Accordingly, one can conclude that, while the Dimension 7 turn functions are coherent at the turn level, at the discourse unit they contribute to a range of discourse functions, though on the positive side there is a strong link to Elaborated Speech. But in neither case is the micro-structural function essentially the same as the macro-structural function, even where this is a clearly possible outcome (as we saw with Personal Narrative). To put it simply, Personal Narrative may be interspersed in a range of functions at the discourse unit level without becoming the dominant function of that discourse unit. Consider Figure 6.1. This shows a single Situation-Dependent Commentary discourse unit. The underlined turn is coded as Personal Narrative at the turn level.

Figure 6.1

A Situation-Dependent Commentary discourse unit from Spoken BNC 2014 file SXRR.

In this figure, we see a turn which may be viewed as a Personal Narrative, but it does not define the discourse unit.

Before concluding this discussion, however, we can consider one way in which the Advisory function may be similar to the Elaborated Speech discourse unit function. The Advisory turns seem to be associated with lengthy discourse units. Is this length association reflected in the turn-level analysis? If we compare the length of the 100 most strongly associated Advisory turns to the length of the 100 most strongly associated Personal Narrative turns, we find that Personal Narrative turns are shorter on average (average turn length of 19.23 words) than the units on the advisory turns (average turn length of 149.01 words). Indeed, if we correlate each turn’s Dimension 7 coordinate (in the whole corpus) to the length of the turn in word tokens, there is a slight positive correlation to length (r = 0.15). This indicates that turns associated with the Advisory function (positive Dimension 7) are generally longer than personal narrative turns. So, the association of the long Advisory turns is notable in that it means that those turns are, in essence, so substantial that they make a strong contribution to the Elaborated Speech function of Dimension 1 of the discourse unit-based analysis. However, the association of advisory turns to negative Dimension 1 of the discourse unit-based analysis is not so strong that the Elaborated Speech function of the discourse unit analysis becomes co-terminus with the turn-level Advisory function – there is greater diversity in the Elaborated Speech function at the discourse unit level than that.

Hence the fit between the micro- and macro-approaches to the discourse functions in the BNC 2014 is now clearer. There is a close fit between the functions discovered at turn level and the discourse unit level. The exception is Dimension 7 at the turn level. These turns contribute to the realisation of a wide range of functions at the discourse unit level. However, their associations to any one dimension at the discourse unit level is not strong enough for the same functions to occur as the Dimension 7 functions at the turn level, though in the case of Elaborated Speech the contribution of the Advisory function from the turn level is strongest.

With the micro- and macro-structures of the Spoken BNC in mind, as discovered using short-text MDA, we will proceed in the following chapter to micro- and macro-structural functions across the three corpora we have analysed – the TLC, TLC L1 and the Spoken BNC 2014. For the discussion that follows we will use the discourse-level view of the discourse functions for purposes of comparison. However, we believe that the alignment of the turn-level functions discussed so far is a sound guide to the nature of those functions, where they occur, at the discourse unit level.