4.1 Introduction

In the previous chapter we established that the short-text MDA approach, coupled with the discourse unit-annotated corpus, could give us insight into discourse functions in learner and examiner language. We also saw how macro-structure and micro-structure could intersect in distinct ways. For example, in Dimension 2 we saw how there may be a link between the macro-structural discourse unit and the micro-structural turn – Information Seeking at the turn level was common in the Informative and Instructive discourse function in that dimension. Yet, for other functions explored, the link between the macro- and micro-levels was not as direct, as the macro-level appeared to be formed by an assemblage of diverse functions at the micro-level. In the case of Dimension 3 with the Realis function, the dominance of Information Seeking in turns was strongly linked to task, as the Greetings task, with its need for the transaction of known information, often comprised a synergy of the Information-Seeking function and Realis function turns, to the extent that Realis discourse units were built from them. The needs of the task, for information exchange, were key in requiring this. Therefore, task evidently plays a key role in influencing the discourse function at the turn and discourse unit levels as well as the relationships between these levels.

Recognising the varied relationship between the macro and the micro in terms of discourse functions, as well the potential for task type to influence this varied relationship, this chapter begins with the analysis of narrative as a feature of the spoken language assessment in the TLC with a view to further investigating the relationship between turns and discourse units. Narrative emerged as a feature of the micro-structural analyses of Chapter 2 and, as is presented in this chapter, is evident in the discourse unit analysis in Dimension 4. Therefore, while in the case of the Informative and Instructive function we might claim that the macro-structural function could be the result of a concatenation of Information-Seeking turns at the micro-level of discourse and the subsequent development of complementary discourse units at the macro-level, the question that opens this chapter is whether the same can be said of narrative: the Narrative function at the micro-structural level builds the Informational Narrative function at the macro-structural level. To respond to this question, this chapter explores Dimension 4 and then moves to Dimension 5 to complete our introduction of the discourse unit functions of the TLC.

4.2 Dimension 4: Informational Narrative versus Seeking and Encoding Stance

Dimension 4 is essentially shared between the turn and discourse unit view of the data. However, in the case of the discourse unit view, the view of Narrative and Non-Narrative is refined. The narratives associated with the positive side of this dimension constitute a function that we will call Informational Narratives as they are expressly oriented towards conveying information rather than, for example, being used principally to convey feelings or attitudes. On the non-narrative side of the dimension, the discourse units have the function of Seeking and Encoding Stance (Table 4.1).

There are numerous features associated with the Informational Narrative function, such as several noun types and noun modifiers which are used to produce dense noun phrases, for example, numeral and ordinal nouns, numeral and ordinal determiners (four periods), demonstrative determiners, and definite articles. There are also features associated with narrativity, such as past tense, passive constructions (can be played by piano), third-person pronouns, and existential there (there are many types). These features co-occur in discourse units that have an Informational Narrative function. One such discourse unit, from a grade 7 exam by an Argentinian student produced in the Discussion task for which they were awarded a B (file 2_7_AR_45), is given in Figure 4.1.

Figure 4.1 An example of an Informational Narrative.

By contrast, the features associated with the non-narrative Seeking and Encoding Stance function relate to stance taking and asking questions. Copular verbs, predicative adjectives, to adjective complement clauses and amplifiers, for example, are used to encode stance. Question features, such as question marks and second-person pronouns, are closely associated with seeking information. Thus, all of these features co-occur in discourse units that are functioning to seek and encode stance. Consider the following example, from an Italian student (file 2_6_IT_106) taking a grade 6 exam. This discourse unit is taken from the discussion task, for which the student was awarded a C.

This example clearly shows the orientation of this discourse unit to the Seeking and Encoding Stance function. The student takes the lead in selecting this function with their initial utterances by seeking stance. The examiner engages with that stance seeking, providing an initial response and more detail when prompted. Note that this example also shows another way in which the discourse functions are important – as a cohesive device. We saw in the discussion of the Realis function in Chapter 3 how a non-sequitur can break that cohesive device. In this case, we see a sustained Seeking and Encoding function in effect making the interchange coherent. The discourse unit is a coherent chunk of discourse precisely because both examiner and examinee are contributing to it in a way that is appropriate to the function of it.

If we look at the 100 prototypical discourse units, on either side of the dimension, what do we see? On the negative side, in the Seeking and Encoding Stance function we see the examiner dominating – there are 5,119 words in the 100 units studied, 3,073 are spoken by the examiner (60 per cent versus a proportion of 43 per cent in the whole corpus) and 2,046 are spoken by the student (40 per cent versus a proportion of 57 per cent in the whole corpus). Formulaic speech in each case indicates that there is a link to the Greeting task of the exam, as shown by expressions in the examiner speech such as how are you (fourteen occurrences) and my name is (eight occurrences). Some of these sequences are echoed by the student in that task, with nice to meet you (twenty-eight occurrences in the examiner data, fourteen in the student data) and my name is/my name’s (eight out of twelve occurrences for the examiners, one out of eight for the students) being shared. More telling for the examiner speech in this case, however, is the prevalence of markers of agreement in the speech, designed to mark agreement and to encourage the student to continue (yeah, eighty-three occurrences, okay eighty-four occurrences, mm forty-two occurrences and uhu twelve occurrences). Similarly, the interactive nature of this function is shown in the frequency of you in the discourse unit – this word, which is a feature whose presence contributes to the function according to the short-text MDA, is the most frequent word in both examiner speech (245 occurrences, 7.97 per cent of all examiner speech) and student speech (134 occurrences, 6.54 per cent of all student speech). Another notable feature of the student turns is the presence of laughter, occurring sixty-five times for the students, contributing to the presence of this feature being part of the feature set which defines the Seeking and Encoding Stance function.

On the positive side, in the Informational Narrative function the picture is sharply reversed – it is very much a student-dominated discourse function. The prototypical discourse units for this function are also longer than on the other side of the dimension, with the 100 units being composed of 8,337 words, 1,144 from the examiner (13.72 per cent versus a proportion of 43 per cent in the whole corpus) and 7,193 from the students (86.27 per cent versus a proportion of 57 per cent in the whole corpus). The relative passivity of the examiner is also shown by the words most frequently uttered by the examiner – backchannels make up the majority of the examiner speech, with mm (332), yeah (45), okay (51), uhu (35), oh (24), ah (15), yes (20) and right (30) making up 8 out of the 11 most frequent examiner words, representing 48.5 per cent of all examiner speech in these examples. There are no examples of formulaic speech that meet our criteria in the examiner speech in prototypical discourse units with this function. The contrast with the student data is stark; the top 5 most frequent words contain two filled pause markers, er (450) and erm (173)—a clear sign that the student is working to hold the floor in the interaction. Other words of interest form part of the feature set that define this dimension – the definite article (the, 467 examples), a past tense verb (was, 159 examples) and the third-person pronouns (he, 98 examples, they 79 examples). The orientation to the past is not only apparent in the verbs used – formulaic speech also orients to the past with expressions relating to dates (in two thousand and, four occurrences), the linking of existential there to the past (there was a, six occurrences) and comments relating to time (when he was, six occurrences).

Overall, a thumbnail sketch based on this exploration of the prototypical examples shows Seeking and Encoding Stance to be co-constructed. Informational Narratives see students engaged in long stretches of speech in which they actively seek to hold the floor and orient their conversation to past events. This is clearly in line with the discussion of the functions so far, though it adds the insight that one function seems to be led by the examiner and the other by the examinee.

While these prototypical examples may show the examples which are most typical of use, it does not imply exclusivity – examiners may use the narrative function and students may use Seeking and Encoding Stance, in principle. To better understand both functions, it is worth unpacking them further.

In the Seeking and Encoding Stance function, questions emerge as an important feature. Notably, this is not the first time questions were identified as an important feature of the discourse discussed in this book as, in Chapter 3, their use in the Irrealis function was noted. We may assume that their use is not identical as the bundle of features that make up the Irrealis and Seeking and Encoding Stance functions appear linked but sufficiently different. In fact, there are only two present features which are shared in both functions – the use of auxiliary do and negative interjections. In terms of absence, there is one feature, the use of contracted forms, which is absent in Irrealis, but present in the feature set of the Seeking and Encoding Stance function. Therefore, the link between the two seems slight in terms of form,Footnote ¹ though we might be able to argue for a stronger connection in terms of function.

This is a useful point at which we may be reminded that the discourse units analysed here have values on all of the dimensions we are analysing. So, we can examine the hypothesis that the Irrealis and Seeking and Encoding Stance functions are, in fact, linked. Can we see evidence in the discourse units that we have analysed that the two functions are strongly linked? For example, does a high score on one imply a high score on the other? Consider the last example we looked at, from file 2_6_IT_106. This was a student using the Seeking and Encoding Stance function. An argument could be made for the example that while it is Seeking and Encoding Stance, it also clearly has elements of Irrealis to it – the student is expressing an opinion about something which they are unsure of. They do not couch their statements in terms of secure knowledge—that is, with the Realis function. If we consider the scores assigned to the discourse units for Seeking and Encoding Stance, and correlate those with the scores received for Irrealis, we should see a systematic relationship between the two if this is a consistent feature linking the two functions, that is, if the Seeking and Encoding Stance is systematically bound to Irrealis. When that is done, a correlation may be observed. For example, for the Discussion task, the presence of a negative score on Dimension 4 (Seeking and Encoding Stance) correlates meaningfully with a positive score on Dimension 3 (Irrealis)Footnote ² – ther score for the correlation is −0.121, showing a weak relationship (after Cohen, Reference Cohen1988). This is in line with the observation made about the minimally shared, but far from identical, feature set linking the two functions – while very weakly related, the Irrealis and Seeking and Encoding Stance functions are not the same.Footnote ³ Distinct forms here are giving an indication of distinct functions. So the functions, while related, are distinct.

If we switch our focus to the issue of the preference of the examiner or the examinee for a function, we can also see how the two functions on this dimension interact. Let us begin by considering an Informational Narrative that matches the prototypes well. The Informational Narrative function is used to convey information about an event, typically in the past. These events are factual rather than imaginary. Consider the following example, from the Conversation task in a grade 7 exam. The student is from Moldova and they were awarded a B for the Conversation task of which this discourse unit is a part (see file 2_7_MOL_3):

This shows nicely the nature of the Informational Narrative. This is a successful narrative – the student selects the function and performs it well. The narrative is not co-constructed and the role of the examiner is to use backchannels to indicate their understanding and to encourage the student to proceed. So this stands as a good example, both in terms of examiner and examinee behaviour, of the type of interaction that was implied by our analysis of the 100 prototypical examples of this function in the data.

Moving away from the prototypes, as was shown in the discussion of Dimension 3 in the previous chapter, the fit of the discourse function chosen to the needs of the conversation is key to understanding the performance of the student. This is the case in this dimension also.Footnote ⁴ Conversation has a preference for the Seeking and Encoding Stance function, while for Discussion the Informational Narrative function is preferred. However, in both tasks students scoring D are associated with discourse units with coordinates that are more strongly related to the function preferred by that task than the students receiving higher grades. But is the driver for the selection of the function the same in each case; that is, do the factors at work in driving the selection of the function in interactions with students of lower proficiency in Conversation match those which drive their choices in the Discussion?

To begin to understand the relationship of the grade D students to the conversation task, consider the following discourse unit, produced by a Spanish student taking a grade 7 exam who was awarded a D for the Conversation task (file 2_7_SP_3):

The examiner starts the turn and does so in a way that orients to the Seeking and Encoding Stance function, asking for the student’s attitude to having a dog. The student then begins to reveal a stance, and the examiner leaves space for the student to self-repair. However, the examiner has to scaffold the interaction by making the student’s response relevant to the question with ‘so you wouldn’t be able to have any pets’. The seeking of stance is very much from the examiner, the student is required to self-repair and the relevance of what they have to say to the examiner is resolved by the examiner, not the student. So while the discourse unit is squarely related to Seeking and Encoding Stance, the seeking is exclusively done by the examiner and the examiner leads the interaction. So here we have a clear example, and motivation, for the use of this function by the examiner, which helps to explain why the function links to grade D students – it is examiner-led and repair-oriented.

Contrast that with this discourse unit, from the Conversation in a grade 6 exam with a Mexican student awarded an A (file 2_6_ME_25):

While the examiner initiates the Seeking and Encoding Stance function, the student then drives the interaction, with the examiner providing nothing but backchannels. While the student does code-switch to Spanish at one point, and the examiner either responds with a backchannel or a correction, the student does very clearly respond appropriately to the initiation of the Seeking and Encoding Stance function by making their feelings abundantly plain, with auxiliary do, predicative adjective, contracted forms and pronoun it, all of which are drawn from the set of features that constitutes this function. So the student’s performance is successful at the level of discourse because it is relevant to the function initiated. But importantly, this also shows that the student using Seeking and Encoding Stance may in fact effectively interact. So while the function is linked to examiner repair strategies, it is also possible for the function to be used in the task felicitously by the student.

If we look at the dimensions of the examiners’ turns, do we see any further evidence in favour of this hypothesis: that it is examiner behaviour that is driving the placement of discourse units involving D grade students in the Conversation task further into the Seeking and Encoding Stance function than students awarded higher grades? The evidence is the same as we have discussed previously – where examiners’ turns draw least upon the Information-Seeking function in the turn-based short-text MDA, the student is likely to receive a higher mark. This pattern is most pronounced at grade 6, as shown in Table 4.2. These Information-Seeking turns fit well into the Seeking and Encoding Stance function when we explore discourse units. It is the elevated level of these Information-Seeking turns for the D grade students that sees the discourse units associated with these students more strongly associated with the Seeking and Encoding Stance function than for students at higher grades.

So the micro-structure and macro-structure intersect again here. Where the examiner builds a Seeking and Encoding Stance macro-structure using Information-Seeking micro-structures in tasks where the students are supposed to be interacting and leading discussion, it is a sign of repair and it is indicative of low proficiency on the student’s behalf. So function, role in discourse, micro-structure and macro-structure interact here. It is not simply that Seeking and Encoding Stance indicates poor performance. As shown in Example 42, the function may be student-led and be a sign of successful interaction. Likewise, the simple presence of the function with the examiner initiating it need not be a sign of poor performance – as we have seen, there is a strong link of this function to the Greeting task, irrespective of the level of attainment of the student. The selection of the discourse function in the Greeting is driven by the communicative purpose of that task. However, where certain tasks and roles come together with this function, for example examiner selection of the function in Conversation, then the function is likely to be a repair strategy indicating problems with proficiency. So while the prototypical usages are useful guides to begin our discussion of a function, cases which vary from those examples do so with purpose – for example, the examiner may most often use the Seeking and Encoding Stance function in the Greeting, but their use of that function in other tasks is, arguably, even more important for our understanding of the interaction of the L2 and L1 speakers in this case. In our next example, we will see, in an exploration of the Discussion task, more clearly how this may cast light on how the two functions on this dimension may interact.

What of the students’ performance in Discussion? As noted, like Conversation, the students awarded a D are different to the others – but rather than being more strongly associated with Seeking and Encoding Stance, Discussion is linked to the Informational Narrative function, and the students graded D are associated with discourse units that are more markedly associated with that function than other students (see Table 4.3). At grade D, we can find narratives which, while marked by notable disfluencies, are the students’ own. However, we can also find narratives which are essentially co-constructed, in which the examiner produces the Informational Narrative discourse unit. Example 43 from the Discussion task of an Indian student, marked D for the task in a grade 6 exam (file 2_6_IN_100), shows this. The student has chosen to talk about women’s safety. The discourse units of the task are heavily laden with examiner speech as the student provides short responses to the questions of the examiner. The examiner tries to encourage the student with backchannels, but is also obliged to encourage the student to try to keep the conversation going – thus at the micro-structural level there is an eruption of Information-Seeking turns into the Informational Narrative macro-structure. In this example, through those questions, a story builds of children being abducted for ransom in India. That leads to the examiner producing an Informational Narrative discourse unit with minimal input from the student. This is clearly scaffolding to help a weak student who is struggling to tell the story – they have built up to the point where a summary or conclusion of some sort is needed, but the discourse unit opens with the student struggling to produce it. The examiner then presumes the student was trying to select the Informational Narrative function and proceeds to produce one, via Information-Seeking turns, that makes the student’s point for them

So, we might hypothesise that the pronounced preference for discourse units with an Informational Narrative function in low-scoring students in the Discussion task is driven by the use of examiner-led narratives like this, by micro-structures drawing on the Information-Seeking function as a scaffolding device, either in whole or part constructing the narrative to sustain the conversation. Is there evidence to support this hypothesis if we examine the short-text MDA of examiner turns from Chapter 2? The key here is Dimension 7 of the examiner turns: Narrative (positive) versus Stance Seeking (negative). When the distribution of this dimension across the examiner turns is considered with regards to task and grade (see Table 4.4), we find that Conversation elicits more Stance turns from the examiner, while Discussion elicits more Narrative turns.

Do these contributions from examiner and student to the discourse units in this dimension become clearer when we consider grade, mark and task together?

The analysis of examiner turns seems to suggest that for the Informational Narrative function at least, grade 6 may represent a point at which the development of this function might be at a tipping point. We will begin by considering Conversation and then return to consider the role of narrative in Discussion.

For Discussion, the pattern is quite consistent with regards to examiner turns, with a preference for Narrative turns. Yet that preference is consistently muted for students scoring a D, suggesting that, as shown, other functions such as Information Seeking and Stance Seeking are used to help the weaker students co-construct a narrative. The choice of Informational Narrative (positive Dimension 4) is appropriate for this task and we may see that differences in proficiency of students marked A, B and C are apparent: as the grade at which those scores are awarded increases, so the orientation of the discourse units to the Informational Narrative function increases, as can be seen in Table 4.5.Footnote ⁵ At grade 6 students marked D seem to be associated with discourse units linked to Informational Narratives more than students scoring A or C, and at grade 7 there is a pronounced peak in such an association for these students. Does the behaviour of the examiners as co-creators of narratives in such cases provide us with an explanation for this patterning?

From grade 7 onwards, the use of the Narrative function in Discussion at the micro-level by the examiner generally declines as the mark attained by the student increases, until, by grade 8, the Narrative function is only slightly marked in terms of usage for students awarded an A, as shown in Table 4.4. To some extent, the relationship between grade and the use of the Narrative function is observable at grades 7 and 8 for the examiner turns also, but the pattern is reversed – the examiner produces less Narrative function turns at higher grades with students marked A–C, as can be seen by looking at grade 8 in Table 4.4. This is evidence that the burden of narrative production in the interaction is being taken up by the students. The examiner does not need to co-construct the narrative with the student as the latter’s proficiency in the use of Informational Narratives increases. This is a sign that the use of Narrative turns by the examiners in Informational Narrative discourse units links to scaffolding. However, the sustained reliance of students marked D on Narrative turns from examiners is a sign that, beyond grade 6, such students are still having issues with the formation of Informational Narratives in this task, as illustrated earlier in this discussion.

The most striking finding with regards to Conversation, presented in Table 4.6, is that, for students marked A–C, the discourse units associated with them shift over the grade. While at grade 6 all students are associated with Seeking and Encoding Stance discourse units, at 7 and 8 they are associated with Informational Narratives instead.

This result is congruent with what is happening with Discussion. As students progress, they are much more successful at performing the Informational Narrative function and the supporting questions of the examiner are a less marked feature of the conversation. However, as discussed earlier, the link of the students marked D to Seeking and Encoding Stance is a sign that there is a proficiency issue and there is the use of scaffolding turns from the examiner. When we consider again the association of the examiner turns with Dimension 2 in the Conversation according to the learners’ mark and grade (Table 4.2), we find support for the view that at grade 6 examiners selecting Information-Seeking turns are driving this behaviour for all marks, though the behaviour is most notable with students receiving a D, indicating again a link to the use of this function by examiners as a form of scaffolding. This intensity of engagement with Information Seeking with lower grade students lessens as students progress through the grades. Eventually the examiner’s performance at the micro-level begins, for example with students receiving an A at grade 7, to be more associated with the other side of the dimension – a Descriptive function.

So a clear pattern has emerged: the interaction with students is driven, to an extent, by the selection of Seeking and Encoding Stance discourse units by examiners. These discourse units contain Information-Seeking micro-structures used to scaffold the performance of students. This behaviour is marked for students at lower grades in general and with lower marks in particular. In turn, this suggests that, as students progress in their language learning, as indicated by grade and/or mark, their performance requires less scaffolding by the examiner. It is the development of competence in the selection and performance of discourse unit functions, as appropriate to the task, that is the factor controlling our view of discourse unit functions on this dimension.

To conclude the discussion of this dimension, let us summarise the key findings that have emerged. Firstly, the prototype exploration was helpful in terms of giving us strong, typical examples to focus on. But, as we saw, usage which went against those examples was motivated by choice and communicative purpose; for example, the use of Seeking and Encoding Stance as a scaffolding device by the L1 speaker outside of Greeting. We also saw that the Seeking and Encoding task function is typically examiner-led, while the Informational Narrative task is typically student-led. The relationship between the two thus relates, in part at least, to roles in social interaction – the L2 speakers, in the tasks where they use the Informational Narrative function, are licensed to do so. It represents a choice of discourse function appropriate to the task they are performing. The same is true of examiners as their role is licensed too and the Seeking and Encoding Stance task helps them to discharge that function. This may be related to task (specifically Greeting) but it also relates to their role in the conversation – they are a cooperative conversational partner and they may use that function to scaffold the interaction of L2 speakers also. Especially in the context of scaffolding, we see some of the functions at the micro-structural level discussed in Chapter 2 come into play for the examiner – their use of Narrative and Information-Seeking turns can be key resources for them in helping the L2 speaker, and an element of co-construction of discourse units, in particular Informational Narratives, may ensue.

So, overall, at an abstract level the variation we see in the data has a common root in discourse in that the same factors are driving choice and variation – the role of the speaker in the interaction, their proficiency, the task in question and the appropriateness of the discourse resources they have to the goal they are pursuing. With regards to the Interactive task, there is a lack of clear patterning around the task, with the functions themselves not being diagnostic of proficiency.

Before leaving the discussion of this dimension, we must return to an important question posed at the start of this chapter and consider the interaction of narrative functions at the macro- and micro-structural levels. We see an important feature of the Informational Narrative that is implied by the discussion so far – it is composed of narrative elements. We have seen that Narrative is present at the micro-structural level and that these turns can build to create a macro-structural Informational Narrative. What is the process governing the composition of such micro-structures to compose a macro-structure? May we be able to find, in essence, intermediate structures between the turn and the discourse unit? May higher level macro-structures be present which, in turn, are composed of discourse units? We will return to focus on these questions in Chapter 8. For now, we will focus on concluding our discussion of the short-text MDA of the TLC discourse units by considering Dimension 5.

4.3 Dimension 5: Persuasion versus Information Seeking

The final dimension that characterises the discourse units produced in the TLC sees Persuasion on the positive side of the dimension and Information Seeking on the negative side of the dimension. The features which constitute both are given in Table 4.7.

The Persuasion function is constituted by features, such as suasive verbs and prediction modals, that are associated with trying to bring about some future action. There are also many features used to encode stance such as comparatives, to + adjective complement clauses, predicative adjectives and stance verbs and infinitives. These features co-occur in discourse units that are often suggesting/advising/guiding a particular course of action or sharing their opinion in order to persuade and encourage some desired future action. The following example shows an Indian student, taking grade 7 (file 2_7_IN_47), trying to persuade an examiner, in the Discussion, of the virtues of a footballer. The student was awarded an A for this task. Note that the examiner engages in exclusively phatic communication in this interaction. The discourse unit also has many of the features which are present in the Persuasion function, including contracted forms (e.g. I’d), stance verbs (e.g. like) and comparatives (e.g. better).

By contrast, Information Seeking is characterised by several question features, such as WH-words, question marks and WH-clauses. There is also a past tense orientation through past tense verbs and have as a main verb, which is often used to talk about experience. Time subordinators and adverbs of usuality also suggest that time is important in these discourse units. By looking at the discourse units most associated with Information Seeking, it is possible to see that these linguistic features co-occur to realise the underlying Information-Seeking function. The function is especially related to the past. The following discourse unit, from a grade 6 Discussion task with an Italian student (file 2_6_IT_50) awarded an A for this task, is an example of Information Seeking. In this case, the behaviour is mutual – the examiner initiates the function by asking the student for information, and in turn the student builds upon the Information-Seeking function initiated by the examiner by asking the examiner for information.

What do the prototypical discourse units for each of these functions show? For Persuasion, it shows that the examiner draws heavily on this function when introducing tasks – here it takes the form of providing an explanation to the student of what they are to do in a task. This is shown by the highly formulaic nature of examiner speech in the discourse units marked most strongly for this function, including lengthy, frequent fixed expressions such as need to keep the conversation going (eighty-five examples) and after four minutes I’ll end the conversation (sixty-six examples). By contrast, student speech is reduced to communicating agreement, with okay and yes being the most frequent words spoken by the student in these discourse units (occurring 169 and 117 times respectively) and I’m ready and yes ma’am being the most frequent multi-word units (occurring 18 and 16 times respectively). So these prototypical uses are clearly examiner-led and relate to the introduction of tasks. The dominance of the examiner is again apparent when we look at the contributions – the 100 discourse units most strongly associated with persuasion are composed of 6,707 words: 5,703 by the examiner (85.03 per cent) and 1,067 by the student (14.97 per cent).

For Information Seeking, the contributions from examiner and examinee are more evenly balanced, with the discourse units being composed of 7,675 words: 3,765 from the examiner (49.06 per cent) and 3,910 from the student (50.94 per cent). The examiner speech is characterised by a wide range of multi-word expressions, mainly oriented towards asking questions of the student, such as do you have (twenty-three occurrences), what do you (thirteen occurrences), do you know (twelve occurrences), did you have to (seven occurrences) and do you think (seven occurrences). These are directed towards the student, contributing to you being the most frequent word uttered by the examiner in these discourse units (257 occurrences). By contrast, the student talks from a personal perspective, with I being the most frequent word uttered by the students in these discourse units (241 occurrences). The students use filled pauses to hold the floor in the interaction, making er and erm the second and third most frequent words spoken by the students (160 and 125 times, respectively). The first-person singular orientation of the student speech is reflected in the multi-word expressions used by the students, which include I don’t know (eighteen), no I don’t (seven), when I was (seven), I didn’t (five) and I think that (five). So the indications are that these discourse units are interactive in nature and jointly constructed, though in terms of Information Seeking the weight of questions leans towards the examiners, with the students responding.

If we consider task, how do the Discussion and Conversation tasks relate to the functions of Dimension 5? Table 4.8 presents the association of the discourse units in the different tasks according to the marks the learners received. To begin with Discussion, the task is linked overall to Information Seeking, with the lowest grade, D, being most pronounced in linking discourse units to that function.

A possible explanation for this may suggest itself to the attentive reader – as discussed in the previous section, the function Information Seeking is also present as a feature of examiner speech at the micro-structural level, suggesting that the increased salience in discourse units of the Information-Seeking function is caused by examiners intervening with that function at the micro-structural level to support weaker students. A good example of that can be seen in the following Discussion example (from file 2_6_IN_59). This discourse unit involves an examiner and an Indian student taking a grade 6 exam who was awarded a D for the discussion task. The discourse unit that follows is the sixth of nine discourse units produced in the task.

In this exchange it is the examiner who introduces the Information-Seeking function and they sustain it. The examinee does not participate in the information seeking. If we look at the other discourse units in this task, Information Seeking predominates, with seven of the nine discourse units in the task having that function. When we examine the Information-Seeking discourse units, it is almost exclusively the examiner seeking the information. Only in the first discourse unit does the student produce turns which are Information Seeking (the student asks a series of questions). Thereafter the student abandons this function and responds to questions rather than asking them. All Information Seeking is then initiated and sustained by the examiner, as we saw in the example just given. This is in spite of a prompt from the examiner in the seventh discourse unit of the task for the student to engage in information seeking– they initiate an Information-Seeking discourse unit by saying to the student ‘okay alright and erm can you ask me a question about ice cream?’. The student fails to do so. So it seems that the association of this function to low proficiency is indicated by an increased incidence of Information Seeking by the examiner and a related low incidence of Information Seeking by the student. Information Seeking itself, as in the example from the seventh discourse unit, can itself become a scaffolding function, where the examiner explicitly uses Information Seeking to see whether the student is able to perform this function.

If we return to the discussion of Dimension 3 in the previous chapter, we see the same explanation in operation here in the Discussion as we saw in the Conversation there. At all grades, salience of the Information-Seeking task is linked to poor student performance; more specifically, the production of turns with an Information-Seeking function by the examiner is most pronounced with students scoring a D. This is for the reasons outlined – it is because the ability of the student to produce that function seems limited, leading to the balance of Information Seeking shifting back to the examiner, who produces these utterances both to sustain the discussion and, on occasion, to attempt to expressly prompt the student to produce this function.

For the Conversation task, a similar argument seems to present itself. Students scoring A and B are associated more with discourse units relying on Persuasion, while students scoring C and D are linked to discourse units engaging in Information Seeking, very markedly so in the case of students scoring D (see Table 4.8).

The increased reliance on the Information-Seeking function for those scoring a D can, once again, plausibly be ascribed to the link with examiner behaviour at the micro-structural level. At all three grades for the Conversation task, the examiner turns draw markedly on the Information-Seeking function, and the function declines in use as the attainment of the student increases, as shown for grade 6 in Table 4.2. It should also be noted that when students scored an A at grades 7 and 8, the examiner turns are actually associated with the Descriptive (positive Dimension 2 at the micro-structural level) function, as opposed to Information Seeking, suggesting that there is an interaction between grade and function for this task, as will be explored shortly.

Before leaving this point however, it should be noted that increased usage is, once again, generally indicative of examiners being obliged to use Information Seeking as a strategy to sustain an interaction. Yet the function itself need not necessarily align with a low score if, for example, it is used appropriately by a student. Consider the following discourse units, both drawn from the Conversation task and both associated with the Information-Seeking function. The first, from file 2_6_IN_106, is from a grade 6 examination of an Indian student who was awarded a D for the task. While initially Information-Seeking turns by the examiner are used to encourage responses from the student, the laconic nature of the replies finally forces a more explicit request from the examiner for the student to continue their response in the seventh turn and arguably in the ninth turn also.

The second is from file 2_6_AR_37, and is, once again, a Conversation discourse unit. This Argentinian student was awarded an A for their performance. Relative to the previous example, this one shifts the weight of the interaction towards a balance of seeking and revealing of information, with much of both being framed by the past in the first, second and fifth turns.

The difference between the two examples is stark – in the former, the examiner is using the Information-Seeking function to sustain a conversation with a student who briefly responds to the questions given, placing the burden of sustaining the conversation on the examiner. By contrast, the roles are reversed in the second example. The A-graded student understands how to lead the conversation by an appropriate use of the Information-Seeking function and initiates and constructs discourse units with that function. The student effectively carries the burden of directing the course of the conversation. Similar examples have been presented for other functions discussed so far – poorer grades for students are linked to functions selected by the examiner who is trying to encourage the student to proceed with an interaction where the student is struggling to initiate and/or construct discourse functions appropriately to perform a task. However, once mastered, the use of those functions by the student may betoken a successful acquisition and use of that discourse function.

For Discussion, our view of the dominance of Information Seeking becomes refined when we consider task, grade and score together, as presented in Table 4.9. There is a clear link with proficiency which reinforces the view of Information Seeking being a sign of scaffolding – grade 6 sees the most marked use of Information Seeking for all students, irrespective of the score they received. However, students scoring C and D are more strongly linked to the function than those marked A and B. Also, as proficiency builds, the highest scored students become linked with Persuasion – starting at grade 7, students scoring A or B become linked to Persuasion rather than Information Seeking.

Below is a discourse unit, from the discussion task of a Portuguese student taking a grade 8 exam, who received a score of A. The example is taken from file 2_8_POR_1.

There is one Information-Seeking turn from the examiner, who otherwise only encourages the student to continue through the use of backchannelling. The bulk of the discourse unit demonstrates the student confidently using the discourse unit function of Persuasion, trying to encourage the examiner to understand their view on the use of technology in planes. This particular Discussion runs over five discourse units and all of them are linked to the Persuasion function and, once the introduction to the task is completed, the interaction is very much led by the student, as the example just given illustrates.

What about Conversation? Again, the combination of factors refines our view of the data. As with Discussion, the Information-Seeking function is dominant for all students at grade 6, though the prevalence of discourse units with this function increases with declining proficiency (see Table 4.10). However, unlike Discussion, all students bar those awarded D are associated with Persuasion discourse units for grades 7 and 8. The association of the students awarded D with Information Seeking continues across grades 7 and 8, though it declines as the grade of exam increases, suggesting that there is still an increase of proficiency across grades even for these students.

For both Conversation and Discussion, the impression of a declining incidence of Information Seeking as proficiency increases, either between grades or as measured within a grade, is reinforced by looking at examiners’ production of Information-Seeking turns. As noted, this is true of Discussion and it is also true of Conversation. In Conversation at grade 6 the association between examiner turns and the Information-Seeking function increases as student proficiency declines. At grade 7, students scoring C and D are associated with Information-Seeking turns by the examiner; those marked D most notably so. At grade 8, irrespective of student score, the examiner turns are not Information Seeking, they are Describing; that is, the function paired with Information Seeking on Dimension 2 of the analysis of examiner turns in Chapter 2. However, some degree of sensitivity to proficiency is visible, with students scored A being most strongly associated with the Descriptive function in examiner turns.

In the Interactive task, there is a clear link between proficiency and the Persuasion function. Persuasion is an important function for the task; the higher the proficiency of the student, the higher their average score for this function is on the positive side of this dimension (see Table 4.8). The Interactive task often hinges on Persuasion – indeed, one of the prompts used in the task in the corpus is ‘my friends have been trying to persuade me to join a social networking site’. The A-graded students pursue the Persuasion function in response to this, though for the weaker students, Information Seeking is apparent. Consider the following discourse units (Examples 50 and 51) from the Interactive task. Both were produced as part of the task by grade 8 students when the prompt regarding social media was used. The first is a Brazilian student marked D (file 2_8_BR_1). The interaction is brief, covering just three discourse units. The second discourse unit follows.

The student, throughout the interaction, largely gathers information. The examiner repeatedly signals that they want to be persuaded – but with the exception of ‘Twitter can be a good thing’, the student neither tries to understand the motives of those trying to persuade the examiner nor tries to use the information either to suggest that the friends are wrong or to persuade the examiner to change. The discourse goal of the examiner – wanting to be persuaded one way or another – does not match well the discourse strategy of the student, as realised. Contrast this with this Argentinian student, scored A (file 2_8_AR_4). The discussion begins with the student asking for the examiner’s thoughts about social media and their friends’ motives, moves through an assessment of both and concludes with a discourse unit (Example 51) where the student engages in the Persuasion function to advise the examiner to both join social media and remain engaged face to face with friends. The examiner reflects back that the student has understood their position and the Persuasion reaches a conclusion – they agree on a position.

4.4 Conclusion

This chapter and the one preceding it have shown that the conversations that L2 speakers engage in with their L1 examiners are influenced by social context as well as proficiency. Given what has been seen, it may also be argued that proficiency itself, in the exam context, represents an important part of the social context and, hence, that social context is the prime force influencing the form and function of language in this corpus, whether that be at the micro- or the macro-structural level. The social context is the examination, and that social context proceeds through a range of tasks. The social context in its broader sense controls discourse – the examiner has a role, relative to the student, that puts them in a position where they are warranted to engage in certain behaviours which the student is not. So, for example, in the Greetings task we see the examiner checks the student’s personal details to make sure that they are the person who is to be examined and that the level of the exam is right. The social context does not allow the student to do the same. At the same time, as the examination itself moves through the tasks, the social context alters and the roles alter with them, to a degree, as the task demands that the student leads in some tasks, for example.

It is with that broad social context that proficiency intersects – the student’s ability to engage in verbal interaction which is appropriate to the social context varies by proficiency. As proficiency improves, we see the L2 speakers able to perform their role in the interaction better. Where this is not the case, we see the role of the examiner flex – rather than simply eliciting a performance from the student, the examiner alters their stance to be a source of support, aiding the student to continue with the interaction and guiding the conversation. The examiner’s choice to act in this role may be triggered by proficiency, but it is still part of the social context of the examination in two important ways. Firstly, and perhaps from the student’s perspective most importantly, it is central to the ethos of the examination – the exams are designed as ‘one-to-one, face-to-face assessments of English language speaking and listening skills with a Trinity examiner, who encourages the candidate to show what they can do through prompts and authentic interactive dialogue’ in a way which builds student confidence.Footnote ⁶ So the behaviour of the examiner should match the expectations of the examinee – this is what the exam is about. However, there is a second way in which the examiner’s behaviour is part of the expected social context – this is what conversation is like. The exam is seeking to assess spoken interaction in English through authentic conversations. Part of conversation is not to act as though one is in a rather harsh exam, correcting an interlocutor continuously or simply halting an interaction if the other interlocutor cannot perform well enough. Rather, an important part of interaction is a key feature of pragmatics – the cooperative principle. Grice (Reference Grice, Cole and Morgan1975: 45) introduced the idea by saying that in speech you should:

This controls the interaction in the TLC – it is the principle by which both the exam and conversation are organised. Students who fail to do this, as we have noted, require the examiner to effect a repair or scaffold. That leads to a judgement that their proficiency is not high. Students who can abide by the cooperative principle elicit behaviours from the examiner focused on participation in the task at hand rather than a repair or some such and, accordingly, are rated as being more proficient. Hence the exam itself, by not artificially constraining the speech of the examiner, actually permits a major organising principle of discourse into the conversation – the cooperative principle. Some of the principles that are related to the cooperative principle have certainly been evidenced so far, in particular the maxim of relevance – that one’s contributions should be relevant. Again, we have seen in Example 36 in the previous chapter that the examiner intervenes where students violate this maxim, in the example cited challenging a non-sequitur.

However, the conversations are not entirely authentic – people are not generally rated on their performance in conversations, nor do such conversations usually proceed through a fixed set of tasks. Also, the exams do lead the interlocutors away from Gricean maxims in an obvious way. It is almost certainly the case that the maxim of quality is disregarded – there is no need for the examiner or the examinee to speak the truth, avoid falsehoods or make statements which are not well supported with evidence. For the examiner, working to prompts given to them, we may reasonably expect that when they say, as a prompt at the start of the Interactive task ‘my teenage nephew has decided that he wants to take up boxing’ (occurring in five different files) that this is not true. There is no such nephew. It is simply a prompt used by the examiner to stimulate a conversation. If we check the metadata on the examiners, we discover that five different examiners say this, stretching our credulity further. Likewise, the scenarios discussed, and the opinions expressed by the students, may or may not be true. These are unimportant for the purpose of the exam, which is very much about assessing the ability of the students to interact in a range of tasks requiring a mastery of different functions in discourse. The truth of the propositions presented by either of the interlocutors, to the extent that this does not disrupt the coherence of the interaction, is unimportant.

The question arises, however, of the degree to which both student and examiner abide by the cooperative principle. Might it be that this principle is language-specific and, in effect, introduces a variable into the exam which should be considered? Some theorists are clear on the matter – the cooperative principle is language-independent: Leech (Reference Leech1983: 10–11) notes that while we may think of general pragmatics, and place something like the cooperative principle within that, the cooperative principle in practice may be influenced by local conditions. So, abstracted from context, one might conceive of the cooperative principle as an ideal which operates across languages. However, in use, the realisation of the principle may vary – for example, somebody who has committed a crime and is questioned by the police about it may be well motivated not to abide by the cooperative principle. On the other hand, an innocent person interviewed in the same circumstances may be highly motivated to be as cooperative as possible. The influence of local conditions on the cooperative principle is clear but, of course, depends on a degree of shared knowledge – another crucial aspect of local conditions may be an awareness of those conditions and the realisation of those conditions being clear to both interlocutors. This is crucial because it explains why we might expect the cooperative principle to apply across our data – because the interlocutors are aware of the local conditions that will shape its use. Hence our view of the cooperative principle here falls clearly within what Leech would call social pragmatics – this is less abstract than general pragmatics as it is bound to local conditions – it is the ‘sociological interface of pragmatics’ through which something like the cooperative principle may ‘operate variably in different cultures or language communities, in different social situations, among different social classes, etc.’ (Leech, Reference Leech1983: 10). In the case of the exam, the social context is one for which both the student and examiner have prepared – the student has been made aware of the nature of the exam, which is rooted in the cooperative principle, and the examiner is tasked with delivering an exam in which they play the role of a cooperative interlocutor. Additionally, as each task starts, the examiner makes the cooperative nature of the interaction clear by alerting the student to the nature of the task and acquainting the student with their role in it. In that context, while some authors have sought to identify ways in which local context in non-English-speaking cultures may vary with regard to the interaction between local conditions and the cooperative principle (e.g. Keenan, Reference Keenan1976), this does not really change or challenge Leech’s argument – the cooperative principle exists at a level of general pragmatics and is influenced by, but not negated by, local conditions in socio-pragmatics.Footnote ⁷ So in our socio-pragmatic context, the student has been prepared for the local context of the exam, the examiner reinforces those expectations and, throughout, the cooperative principle is in effect in the interactions.

Of course, this argument stands at a level of abstraction – it is possible to see that the socio-pragmatics of the interaction should be such that the cooperative principle should be applied irrespective of the cultural background of the speaker. But is there a way of seeing whether this is the case? While we might look to see whether certain discourse units marked with functions associated with scaffolding are present with speakers from one background rather than another, our attempt to study this is confounded by the co-constructed nature of the discourse units – without going into the data and coding the discourse units carefully to say whether the function of the discourse unit is scaffolding or not in each case, and then determining whether the discourse unit is student- or examiner-led, approaching this question becomes problematic. However, at the micro-structural level the approach seems clearer – if we look at examiner speech and consider the cultural background of the learner and their proficiency, then we may see that micro-structures associated typically with scaffolding by the examiner are more commonly associated with learners from one background rather than another. This would be an indication that the examiner is having to do more scaffolding with students from certain cultural backgrounds, which may, in turn, suggest that the local conditions vary by student background.

With the data at hand we may hypothesise that task, social, and perhaps cultural, context have an impact upon the data, but it is hard to explore the hypothesis critically. For example, if we consider for a moment the top-down discourse unit coding and look for examples of discourse units relating to conflict, we find that there are only nine discourse units in the corpus where conflict is the primary code of a discourse unit. It is very little data to work with – we might note that the units are produced by nine different students: five from India, three from Mexico and one from Spain. Likewise, we may note that those students are linked to a range of proficiencies, with the students being scored as A (once), B (three times), C (four times) and D (once) in a range of tasks (Conversation, Discussion, Interactive). But while that allows us to dismiss some absolute statement we could make – for example, it is not true that students producing a conflict discourse unit will never be graded A – it does not really allow us to explore why this discourse unit code is dispreferred. To begin to do that, we need a different type of data – we can switch to contrasting the TLC data discussed in this chapter and the previous one with data where the examiner and examinee share the same L1 and the examinees have not been trained to take the exam; Rather, they are simply responding to the tasks as L1 speakers. With that data, a clearer view of the factors at play in L2 performance in the corpus may be achieved. That is the goal of the next chapter.