9.1 Introduction

In this chapter we will examine narratives using the model of Labov and Waletzky (Reference Labov and Waletzky1997), in a qualitative study, to begin to explore the ways in which narratives are used by L2 speakers in the TLC. A key feature of this chapter is that we will return to an earlier stage in the book – the decision to move away from L1 background – and chart a slightly different course. In this qualitative study, where the issues of frequency (which were some of the issues that militated against a focus on L1) are set aside, we can take a slightly different approach, hypothesising about the role that L1 background may play in a focused, small-scale exploration of narrative in the TLC. Before proceeding to the analysis, however, we will first outline how the data used in this chapter was selected and annotated.

9.2 Annotating Narratives in the Data

The analysis presented in this chapter is based on annotations of narratives, following Labov and Waletzky’s approach to narrative analysis introduced in the previous chapter. We began by taking a sample of 200 texts from our data for analysis. The sample covered speakers from countries where there were at least ten files in each of grades 6–8 of the TLC. There were six national backgrounds, with sufficient numbers of files available in the TLC data analysed in this book – Argentinian, Chinese, Indian, Italian, Mexican and Spanish. Ten files per level for each national background were selected at random, with equal numbers of male and female speakers being selected where possible. In turn, this gave us access to six varied L1 backgrounds. This produced a sample of 180 files. To permit comparison to L1 English speakers, twenty files were selected from the TLC L1. These files were selected to produce a gender-balanced sample. In total, 200 files were analysed for the study presented in this chapter. In the TLC sample are speakers of Hindi and Italian as well as three varieties of Spanish (Argentinian, Mexican, Peninsular) and three varieties of Chinese.

Focusing on this sample, we began by identifying all instances of narrative by manually searching for the occurrence of at least two narrative clauses (i.e. clauses containing a verb in the simple past tense or historic past tense), as well as the compulsory structural element of Complication (as described in Chapter 8, after Labov and Waletzky Reference Labov and Waletzky1997). Qualifying texts were then read and manually annotated for the presence of the remaining narrative components outlined by Labov and Waletzky (Reference Labov and Waletzky1997): Abstract, Coda, Complication, Evaluation, Orientation and Result. We manually coded the narratives, using XML mark-up, to show where these elements occurred within the texts.Footnote ¹ These searchable annotations (which we will call tags) mark the beginning and end of specific components (e.g. Abstract begins with <abs> and ends with </abs>), as well as the beginnings and ends of narratives more widely (i.e. <nar> to </nar>). In Table 9.1, we provide an example of each narrative element in our data, to demonstrate, in very broad terms, what each one ‘looks like’ in the context of the exam. These examples are produced by L2 speakers in our TLC sample.

Table 9.1Examples of different narrative elements in our TLC narrative sample.

Let us return to a question we explored with reference to one narrative in the last chapter and briefly explore it again. Are narrative elements simply micro-structures in the same sense as explored in Chapter 2? If each narrative element was constituted by a single turn, it would raise the prospect of exploring the mapping of these elements to the micro-structural analysis in Chapter 2. In Chapter 8 we saw, in one example, that this is not necessarily the case. Unfortunately, more generally, while narrative elements may be coterminous with a turn, they may also occur below the level of the turn and at what we called the meso-structural level—that is, above the level of the turn but below the level of the macro-structure (in this case discourse unit), just as we saw in Chapter 8. Consider the following example, taken from file 2_8_ME_15, of a Mexican student taking a grade 8 exam for which they were awarded a C:

This is a single discourse unit. Within it, we see a single narrative presented. However, the narrative elements straddle the meso- and micro-levels. For example, the Result occurs in the same student turn as a preceding Complication, while the initial Complication is composed of multiple turns contributed by both speakers. Accordingly, we proceeded with an analysis of the narrative elements in their own right, as these cannot simply be viewed via our micro- or macro-structures. Nonetheless, as the discussion in this chapter proceeds, we will bring in insights from the analyses in Chapters 2–7 where relevant, demonstrating the complementarity of the approaches argued for in Chapter 8.

Of the 200 texts in our sample, in 142 cases (i.e. 71 per cent) the candidate used at least one narrative, a result consonant with the view of the data we have developed so far—that is, that it is rich in narrative. Applying the tags mentioned allows us to explore these narratives in more detail, in particular by quantifying the structural elements of the narratives, identifying patterns they form within the candidates’ discourse, and exploiting these as searchable entry points for close, qualitative analyses of narrative practices in context. This combination of quantitative insights and close, qualitative analysis will guide our exploration of the narratives in the data. The tags were searched and analysed using AntConc (Anthony Reference Anthony2020).

9.3 Findings

9.3.1 L1 Groups

We begin our exploration of the use of narratives in the data by comparing the narratives produced by the speakers when grouped according to their L1. L1 has long been recognised as a key factor that can influence language learning and language use.Footnote ² As such, focusing on Chinese, English, Hindi, Italian and varieties of Spanish (Argentinian, Mexican, Peninsular Spanish) speakers in our data, an initial exploration of L1 background and narrative production was carried out. Notably, each L1 background studied here reflects a culture or cultures that engage(s) with narratives and storytelling. Notably, owing to epistemological differences, the way in which such storytelling practices are realised within the cultures associated with these language backgrounds can differ. For example, some may show greater affinity for the composite elements of anglophone narratives, while others may construct narratives in different ways, beyond the paradigm reflected in Labov and Waletzky’s framework (which, as noted, guides our approach). So we decided to explore the hypothesis that differences in narrative production could emerge across language backgrounds, which in turn could influence the success of language learners undertaking the spoken language assessment at TCL.

Firstly, rather than consider how frequently these speakers produced narratives, we want to look merely at the presence or absence of narrative in the candidates’ speech. Presence and absence have formed significant foci in the analysis presented over the previous chapters, and the benefit of considering presence and absence of narratives here is that it allows us to distinguish, respectively, between those candidates who did and did not display this particular competency within the context of the exam. Table 9.2 shows the number of speakers who produced narratives in each of the L1 groups, also expressed in proportional terms (as a percentage of each group). For ease of interpretation, we have ranked the groups according to the proportion that did produce at least one narrative, with the group producing the highest proportion coming at the top.

Table 9.2Percentage of candidates who produced at least one narrative for each L1 group, ranked from highest to lowest.

As this table shows, L1 Spanish speakers from Argentina and L1 Hindi speakers are more likely to produce a narrative in the exam than speakers of other L1s. Interestingly, L1 English speakers were the least likely to produce a narrative, though it is also worth noting that the majority of them (60 per cent) did produce at least one narrative during the exam. L1 Chinese speakers are ranked third in likelihood to produce a narrative in the exam. However, there is a gap of approximately 9.43 per cent between L1 Spanish speakers from Argentina and L1 Chinese speakers and a gap of 8.49 per cent between L1 Hindi speakers and L1 Chinese speakers, with the latter constituting the largest gap between any of the ranked groups in Table 9.2, suggesting a swift drop in narrative use among speakers of languages other than Spanish (Argentina) and Hindi.

The figures in Table 9.2 reflect the proportions of the candidates in each L1 group who produced at least one narrative during the exam. However, it is also important to bear in mind that the candidates could, and often did, produce multiple narratives during their exam. It is therefore worth considering which L1 group was most productive in terms of how many narratives they produced within the exam. Table 9.3 shows this information, averaged out for each of the L1 groups considered. To guard against any undue influence of a preponderance of ‘zero counts’, we focus only on those candidates who did produce narratives. Since the lengths of the exams varied, to aid comparison across the groups, we have normalised the frequency of the narratives and expressed this figure per turns. We then rank these by most to least productive in terms of narrative.

Table 9.3Average number of turns per narrative for each L1 group, ranked from most to least frequent.

This table suggests that as well as being among the most likely to produce a narrative, L1 Hindi speakers were also most productive in terms of the number of narratives they produced, producing a narrative on average once every 78.20 turns of the exam. The gap between them and the group that ranked second on this measure, L1 Italian speakers, is large; indeed, narratives were almost twice as frequent, in relative terms, in the speech of L1 Hindi speakers compared to L1 Italian speakers. While L1 Spanish speakers from Argentina were particularly likely to produce at least one narrative during the exam, those from within this group who produced narratives did so less often than L1 speakers of Hindi, Italian and Chinese. Once again, L1 English speakers are placed at the bottom of the table, which in this case indicates that they produced narratives, on average, the least often, with an appreciable gap in frequency between this group and that ranked second-bottom (L1 Spanish speakers from Spain).

As we described earlier in this chapter, narratives, according to Labov and Waletzky’s (Reference Labov and Waletzky1997) approach, can vary in terms of the number of elements they contain. While minimally a narrative must contain the obligatory element of the Complication, the other elements can be considered optional. As well as looking at the presence and frequency of narratives, then, we can also consider their length in terms of the number of elements they contain. This information is presented in Table 9.4.

Table 9.4Average number of elements per narrative for each L1 group, ranked highest to lowest.

Firstly, we might note from this table that the range in the average number of elements per narrative is not very large. Nevertheless, we can see that although L1 Spanish speakers from Mexico ranked middling in terms of the relative frequency of the narratives they told, those narratives did contain, on average, the largest number of elements. They are followed by L1 Hindi speakers who again rank highly, coming in second place, and L1 Chinese speakers close behind in third. In the previous table we saw how although L1 Spanish speakers from Argentina were particularly likely to produce a narrative, they did not do so particularly frequently, relative to the other L1 groups, and again this group ranked middling in terms of average number of elements per narrative. While those L1 Italian speakers who produced narratives did so frequently within the exam, those narratives were not particularly lengthy in terms of average number of elements (4.25). Finally, we previously saw how L1 English speakers and L1 Spanish speakers from Spain produced narratives, respectively, least often and second-least often in the sample. Table 9.4 indicates that this pattern holds in terms of the average number of elements per narrative, with L1 Spanish speakers from Spain using the second-lowest number of elements per narrative after L1 English speakers.

Of course, it is possible – and indeed, common – for candidates to use any single element multiple times within a narrative. So another perspective we could take on narratives is to consider the presence and absence of particular elements. For example, it is possible for a candidate to demonstrate their ability to produce only some elements of a narrative in an examination. For a candidate who produces a narrative composed of Orientation, Complication, Result, Evaluation but lacking Abstract and Coda, we could therefore say that four unique elements are present in their talk. We carried out this calculation for every candidate in the sample. Table 9.5 shows the average number of unique elements produced by the candidates within each L1 group (focusing on candidates who produced at least one narrative).

Table 9.5Average number of unique elements produced by candidates within each L1 group, ranked highest to lowest.

This table suggests that L1 Hindi speakers tend, on average, to produce the largest number of unique narrative elements.

We can also zoom into the particular narrative elements. Maintaining our focus on presence versus absence, Table 9.6 shows, for each narrative element, the number of candidates within each L1 group who produced that element at least once within their exam. To facilitate comparison, we have also expressed these numbers as an overall percentage of the candidates who produced at least one narrative within each group. For example, of the twenty-two L1 Chinese candidates who produced at least one narrative, seven of these produced at least one Abstract (31.82 per cent). The information in this table is ordered alphabetically by L1 group. The names of the narrative elements are referred to as follows: ABS (abstract), COD (coda), COM (complication), EVA (evaluation), ORI (orientation) and RES (result).

Table 9.6Percentage of candidates who produced each narrative element from each L1 group.

Beginning with Abstracts, we can see from this table that most of the speakers did not produce this element at all during their exam. In fact, this is the case for every single L1 group apart from L1 Hindi speakers, where the split between those candidates who did and did not produce that element is 50/50. Just under half of the L1 Italian and English speakers produced an Abstract, then this figure drops to around a third for L1 Chinese speakers and L1 Spanish speakers from Argentina and Mexico. Of the L1 Spanish speakers from Spain, only 15 per cent produced an Abstract during their exam.

Moving on to the Orientations, L1 Hindi speakers produce this element frequently too, as all (i.e. 100 per cent) of the candidates in this group who produced at least one narrative also produced at least one Orientation. L1 Spanish speakers from Mexico also frequently produced this element (87.50 per cent), as did those from Argentina (87.50 per cent). In fact, the majority of candidates across all groups produced Orientations, with around 80 per cent of L1 speakers of Chinese, Italian and Spanish (from Spain) doing so. This figure is lower for L1 English speakers, although two-thirds of those who produced narratives from this group produced at least one Orientation.

The next column in Table 9.6 shows information for the Complication element. However, since this element is compulsory it was necessarily produced by all candidates who produced narratives. If we shift focus to the next element, Evaluation, we can see that there are only three groups where at least half of the candidates who produced narratives also produced this element: L1 speakers of Spanish from Argentina (66.67 per cent), Italian (55 per cent) and Hindi (50 per cent). Just under half of L1 Spanish speakers from Mexico who produced narratives produced Evaluations, while this figure drops to just under a third for L1 speakers of Chinese and Spanish from Spain. This figure is lowest for L1 English speakers (25 per cent).

Results were produced by almost all candidates who produced narratives across each of the L1 groups. In fact, this figure is 100 per cent for five L1 groups: Chinese, English, Hindi, Italian and Spanish (from Mexico). Meanwhile, for both of the remaining groups – L1 Spanish from Argentina and L1 Spanish from Spain – all but one of the candidates who produced narratives produced Results.

Of all the narrative elements, the final one – Coda – was produced by the smallest proportion of candidates who produced narratives. There is not a great deal of variation here, with the highest proportion being just 21.05 per cent (L1 Spanish speakers from Mexico) and the lowest being 15 per cent (L1 Spanish speakers from Spain), with all other groups sitting somewhere in-between. Hence, Coda is a generally infrequent element in the data.

If we pause our exploration of the data at this point to take stock of what we have found, an emerging trend seems to be that L1 Hindi speakers are particularly adept at producing narratives both in terms of frequency and the diversity of elements within the narratives. For example, L1 Hindi speakers were among the L1 groups most likely to produce a narrative during the exam (second only to – and marginally behind – L1 Spanish speakers from Argentina). L1 Hindi speakers also produced narratives most frequently, when quantified in terms of narrative per turn, and only L1 Spanish speakers from Mexico produced narratives containing more elements (on average). Furthermore, as a group, L1 Hindi speakers ranked highly in terms of the proportions that produced each narrative element (including ranking joint first for three of the five non-obligatory elements). Of course, other groups rank highly on particular measures. However, L1 Hindi speakers seem to rank consistently highly across all the measures we have considered.

Naturally, attributing storytelling ability to speakers of a specific language or culture can risk overgeneralisations. However, there might be certain cultural and historical factors that play a role – or at least a partial role – in explaining why L1 Hindi speakers perform so well across the various measures of narrative considered here. The activity of storytelling is deeply ingrained in Indian culture, with roots tracing back to historic epics such as the Ramayana and Mahabharata, which have profoundly influenced contemporary methods of storytelling. India also boasts a large and diverse body of folk tales that are unique to each state. Importantly, India also has a well-documented oral tradition, which means that such stories, fables and parables are passed down through generations not only through writing but also through speech (Blackburn et al., Reference Blackburn, Claus, Flueckiger and Wadley1989). Moreover, today, this storytelling culture is reflected in the narrative style of Bollywood movies (Ganti, Reference Ganti2004), as well as in many festivals and rituals that are intertwined with stories which recount historical events and convey moral lessons (Sutherland, Reference Sutherland1989). This hypothesis is interesting to consider and, importantly, touches on contextual factors that are likely to have a bearing on how the candidates use language within the exam. Yet, while such factors are important, we should also bear in mind that other rich storytelling cultures are, of course, represented by the candidates within our sample.

Another difference in the groups studied may impact upon our results: the status of English in India, where it is a language of government and an official language across a number of states. This could mean that the L1 Hindi speakers construct their stories in a way that more closely resembles canonical stories in English (or, at least, find it easier to construct them in this way for the purposes of the exam). With that said, in the case of Hindi, and of the variation in narrative use in different varieties of Spanish we have discussed, it is plausible to consider culture as a key factor in producing the variations that we have seen. The choice of when to deploy a narrative is as much about whether it is considered, in a speech community, to be appropriate in context as much as it is about the affordances of narrative. The acquisition of such norms is culturally bound – it is learnt within a speech community within which a speaker must learn

to produce structured units of extended discourse … that the audience may know little to nothing about the events being told … and … to consider the possibility that the listener may interpret the story events differently than they do.

(Carmiol and Sparks, Reference Carmiol, Sparks and Matthews2014: 280)

The key to this process is social – children acquire the ability to introduce elements of narrative from their interlocutors (Peterson and McCabe, Reference Peterson and McCabe1992) and through such interaction they acquire the pragmatic competence to use narrative effectively in their speech community. Given that narrative as a macro-structure is universal to a degree, yet may vary in the specifics of the linguistic devices and discourse strategies that are associated with it according to the cultural background of the speaker (Minanmi, Reference Minanmi2000),Footnote ³ we should expect to see some cultural variance in the precise use and form of narrative across a population of L2 learners. In terms of previous work in the area, our focus is upon the fourth level of L2 narrative competence introduced by Berman (Reference Berman, Extra and Verhoeven1999, Reference Berman, Verhoeven and Strömquist2001) – level one relates to the grammatical competence to produce a narrative, level two to access to the vocabulary needed to craft a narrative, level three to the ability to use narrative across discourse functions, and level four to the ability to abide by contextual, genre and cultural conventions relevant to the speech community. Earlier in the book we looked at narrative, through the short-text MDA technique, at levels one to three, primarily. However, the differences we are seeing here relate mainly to level four. Issues from this level, as much as the others, may have pedagogical consequences because ‘L2 learners’ narratives show that learners experience difficulties in acquiring language-specific patterns … and that their choices are affected by universal and language-specific factors, such as L1 influence’ (Pavlenko, Reference Pavlenko, Byrnes, Weger-Guntharp and Sprang2006: 114). The presence of differences in our analysis of three varieties of Spanish in this chapter would mean that we would agree with this statement, while noting that L1 culture, as much as language, is likely to be important. Viewed through this lens, the link between the Narrative function in discourse units and attainment discussed in Chapter 4 takes on a different complexion – it may be the result of some learners having to adapt their narrative styles to match the British English norms by which they are being assessed. Rather than being a matter of the acquisition of grammatical forms alone (level one in Berman’s terms), for example, it is also about the journey of the student as they adapt the macro-structure they are familiar with to the different expectations of the speech community they are aspiring to communicate with (more rooted in Berman’s levels 3 and 4). One may hypothesise that, where the changes between the students’ L1 narrative style and that of British English are greater, then ‘becoming proficient in a second language may require greater resources for processing incoming narrative information, for example, in conceptualizing and retrieving syntactic information and word meanings based on phonological, semantic, and syntactic cues’ (Gutiérrez-Clellen, Reference Gutiérrez-Clellen2002: 190). This, in turn, increases the cognitive burden for such students and increases the possibility that accommodating to the British English version of the narrative macro-structure may impact on attainment.

By acquisition in context, differences in the use of narrative in conversation arise – while ‘all children learn that representing past events in conversation … is a culturally valued activity’ it is also true that ‘the way they learn to formulate narrative discourse in response to new pragmatic understanding is culturally specific’ (Carmiol and Sparks, Reference Carmiol, Sparks and Matthews2014: 289). Focusing on studies looking at Latino and Anglo-American English speakers: (i) Rodino et al. (Reference Rodino, Gimbert, Pérez and McCabe1991) found that Latino children used more Evaluation elements in their narratives than Anglo-American children; (ii) Carmiol and Sparks (Reference Carmiol, Sparks and Matthews2014: 290) found that description was used to introduce information about their families by Latino children, a tendency not shared by Anglo-American children; and (iii) Carmiol and Sparks (Reference Carmiol, Sparks and Matthews2014: 289) found differences in the flow of narratives between Latino and Anglo-American children, with Anglo-American children following a linear path in a narrative as opposed to the topic-oriented, relatively diffuse organisation of the Latino children. Given that narrative is a resource that a speaker can carry from their L1 to their L2, and given the different ways in which speech communities license narrative to be used in interaction, the type of culturally bound differences in narrative production that we are observing is understandable, though potentially easy to overlook, as may be its impact on language assessment.

Can the short-text MDA be of help in assessing whether there are affinities between narrative and the different L1 backgrounds of the L2 speakers? As noted in Chapter 1, the number of observations that we can make of the intersection of L1 background and narrative at the discourse unit level is such that the reliability of the results may be doubted. However, at the micro-structural levels there is a much larger number of observations that can be made and, if we use the analysis in Chapter 2, we can see that the Narrative function may be approached, albeit at the turn level, if we look at Dimension 5, which is Narrative (positive) versus Non-Narrative (negative). If we introduce L1 background as a supplementary variable for this analysis and test the relationship of L1 background to the two poles of this dimension, for our six L1 backgrounds we find on the positive side (most marked to least marked) Hindi, Argentinian Spanish and Mexican Spanish, while on the negative side (most marked to least marked) we find Chinese, Peninsular Spanish and Italian. There is some agreement between this analysis and that shown in Table 9.2. The L2 speakers with what we might call a narrative style (on the positive side of the dimension) are those ranked at 1, 2 and 4 in Table 9.2, while those with a non-narrative style (on the negative side of the dimension) are ranked 3, 5 and 6. So even though the ranking in Table 9.2 is in terms of the use of a macro-structure, a narrative matching the model of Labov and Waletzky, the overall picture it paints of the relative propensity for speakers from different linguistic and national backgrounds to call upon narrative is a close match for that which we can see at the micro-structural level.

The weaving together of the quantitative and qualitative analysis so far has largely shown that the two are mutually supportive. This, in turn, allows us to make another observation that was apparent in the previous paragraph – culture, as much as L1, has an impact on the linguistic choices of learners. One of the advantages of having three varieties of Spanish in our sample is that it allows us to see that something other than a learner’s L1 can drive choice, in this case of narratives. Our hypothesis is that culture is the factor weighing upon the choice of the Argentinian, Mexican and Peninsular Spanish L1 speakers when they decide to use a narrative in L2 English speech. Attempting to falsify this hypothesis is beyond the scope of this book – we would need access to the equivalent of the Spoken BNC 2014 for each of the varieties of Spanish, for example, to begin to meaningfully explore the different preferences for narrative use for each variety. However, what is apparent in both the qualitative and quantitative results presented so far is that the three varieties of Spanish clearly vary with regard to their use of narrative in L2 English speech. Exploring the role of culture in the selection of macro-structures like narrative is clearly a fruitful avenue for future research.Footnote ⁴

We shall move on at this point and consider whether there may be other factors influencing narrative production in our sample. Further variables that are known to influence language learning and language use pertain to other demographic factors, such as age. Age is well established as a key variable that influences language learning, whereby younger learners, for example, are often touted as being more effective language learners, often owing to factors underpinning the ‘critical period’ hypothesis (Singleton and Lengyel, Reference Singleton and Lengyel1995), as well as their position within education systems and their openness to and availability for learning. Likewise, there is potential for the examination itself to influence narrative production. The GESE test is designed in such a way to be valid and fair in testing spoken language proficiency, a claim we feel is borne out by the analysis in Chapter 7 in particular. Learners across the world undertake the same examination and this occurs by design, as test developers wish to ensure that all learners accredited with certificates of language proficiency have comparable speaking skills. However, there is one variable that is more challenging to control in the deployment of international assessment: the examiner. Our research has already shown that examiners can influence the success of a conversation and the construction of narratives. As such, it may be the case that the examiners’ practices are influencing the production of narratives. Therefore, to further unpack the factors influencing narrative production beyond the L1, the following two sections investigate the relationship between narrative production, age and examiner practices.

9.3.2 Age Categories

We began our exploration of age simply by looking at the average age, age ranges and age groupings of the different L1 groups represented in our sample. Table 9.7 gives the average age and age range of the candidates in each L1 group within the sample. For ease of interpretation, we have ordered the rows by average age (from lowest to highest). The accompanying Table 9.8 presents this information differently. The candidates are also grouped into five broader age-related categories: young (8 to 15 years); adolescent (16 to 19 years); young adult (20 to 35 years); middle adult (36 to 50 years); older adult (50 years plus). The latter table shows the proportion of each age-related category across each L1 group in the sample.

Table 9.7Average age and age range for each L1 group, ordered from youngest to oldest (on average).

Table 9.8Proportion of age categories in each L1 group, ordered from youngest to oldest (on average; see Table 9.7).

From these tables, we can see that the L1 Hindi speakers are the youngest, and when considered in terms of the age categories they are assigned to, 100 per cent of the candidates in this group belong to the ‘Young’ category, with the oldest speaker in this group being just fourteen years old. This is the only group for which this is the case. If we look at Table 9.7 in concert with Table 9.2, we might infer that younger speakers are more inclined than older speakers to produce narratives in the exam. In fact, the order of the L1 groups in each table is almost identical (barring L1 Hindi and L1 Spanish speakers from Argentina, who occupy the top two slots but in reverse order). To explore the potential influence of age, we repeated the calculations we carried out for Table 9.2, this time looking at the different age categories. The resultant Table 9.9 shows the proportion of candidates who produced at least one narrative within each age category (among the L1 groups considered to this point).

Table 9.9Percentage of candidates who produced at least one narrative for each age category, ranked from highest to lowest.

The information in this table suggests that there is not a straightforward relationship between a candidate’s age and whether or not they produce a narrative during the exam. All of the Older Adults in the sample produced a narrative. However, there are only four candidates in this category, so we need to treat this figure of 100 per cent with a great deal of caution. For the other categories, for which we have appreciably more data, we can see that Young candidates were most likely to use a narrative and Adolescents the least likely (though more of these did produce a narrative than did not). One possible interpretation we could draw is that L1 Hindi, Spanish (from Argentina) and Chinese speaking candidates were particularly likely to produce at least one narrative because these L1 groups each contain a much higher proportion of Young candidates than the others; respectively, 100 per cent, 79.31 per cent and 73.33 per cent, whereas the next highest is L1 Spanish from Spain with just 32.14 per cent. Meanwhile, although Middle Adults were almost as likely to produce a narrative as Young candidates, there were far fewer of these in the sample, and as such they constituted a much smaller proportion of most L1 groups (the highest being L1 English, but only 30 per cent of this group were Middle Adults).

9.3.3 The Intersection of Age and L1

It is possible, then, that the patterns we have observed relating in particular to L1 Hindi, Spanish (from Argentina) and Chinese speakers is an effect of the composition of our data and the fact that Young speakers, who are particularly likely to produce a narrative, make up the majority of these L1 groups. To explore this further, we repeated some of the procedures we carried out previously, this time focusing on the Young candidates in our sample in order to see if there are differences between the different L1 groups making up this age category. This can help us to get a sense of the extent to which age and L1 might influence both how and how often narratives feature within the candidates’ speech. For this part of the analysis, we were not able to include L1 English speakers (whose sample contained no Young candidates) and we decided to exclude L1 Italian and Spanish (from Mexico) candidates, since these groups did not have a sufficient number of candidates for us to credibly see any patterns in their data (having just three Young candidates each). To begin, Table 9.10 shows the number of Young candidates who produced narratives in each of the L1 groups, also expressed in proportional terms (as a percentage of the total number of Young candidates within each L1 group). To aid comparison, we have also provided this information for the remaining age categories (combined) for each L1 group.

Table 9.10Percentage of young candidates who produced at least one narrative for each L1 group, ranked from highest to lowest.

We then re-ran the analysis from Table 9.3 on this same sample, with the resultant Table 9.11 showing the frequency of narratives across each group, expressed in terms of average number of turns per narrative.

Table 9.11Average number of turns per narrative for young candidates in each L1 group, ranked from most to least frequent.

Building again on the presence and frequency of narratives, we also considered the lengths of the narratives in terms of the average number of elements they contained (as we did earlier for Table 9.4). The results are presented in Table 9.12.

Table 9.12Average number of elements per narrative for young candidates in each L1 group, ranked highest to lowest.

As earlier, we also considered the number of unique elements that the Young candidates in each L1 group demonstrated they could use. We thus repeated the analysis presented in Table 9.5 but for the four Young L1 groups under focus here (see Table 9.13).

Table 9.13Average number of unique elements produced by young candidates within each L1 group, ranked highest to lowest.

Finally, we repeated the element-by-element analysis we undertook for Table 9.6. Table 9.14 shows the proportion of Young candidates in each L1 group who produced each of the narrative elements (see Table 9.14, abbreviations as per Table 9.6).

Table 9.14Percentage of young candidates who produced each narrative element from each L1 group.

If we take the results presented across Tables 9.10 to 9.14, then we can see that there is not a neat relationship between age and the production of narratives in the exam. By some measures, the Young candidates are more likely than their counterparts from other age groups to produce narratives and produce narratives more frequently and with more elements, including drawing on a wider range of unique elements. However, in other cases, the reverse is true. At the very least, this suggests that the strong performance overall of L1 Hindi speakers – and to a lesser extent, L1 Spanish speakers from Argentina – cannot be explained by the younger age of their respective samples alone, relative to the other age groups.

The foregoing analysis suggests, then, that age is likely to be at least one of a number of factors bearing on the how and how often the candidates produce narratives during their exams, and while there are likely to be other factors at play, age does at least seem to have a clearer relationship to narrative production than candidates’ L1 does. The generally stronger performance of younger candidates regarding the production of narratives might be because these learners are of school age, and narratives and storytelling are commonly used pedagogical devices in schools, used by teachers to teach all kinds of subject matter. This more regular exposure to narratives in their everyday lives might put school-age candidates at something of an advantage, then. Yet at the same time, age alone cannot account for the patterns we have observed, and as noted age is likely to be one of a number of variables that come together to influence exam performance. When looking at L1 in combination with age, the patterns in our data begin to make more sense. However, the picture is not fully clear still, and another factor we can consider is how the exam itself might influence the production of narratives. With that in mind, in the next section we explore the role of the examiner in supporting the candidates’ production of narratives.

9.4 The Role of the Examiner

At this point, we switch our focus to the language produced by the examiner administering the exam. In particular, we consider how the examiners’ linguistic contributions might serve as ‘prompts’ to support the candidates in developing their narratives. So far we have concluded that the narratives produced by the examinees are on the their own initiative. Nothing in our exploration of the various forms of narrative in our short-text MDA suggested otherwise. While we did see narratives being scaffolded by the examiner, our exploration of the various prototypical examples of narratives did not seem to suggest that the narrative was initiated at the behest of the examiner. Does a more qualitative exploration of our narrative sample confirm or alter that view?

Of the 142 candidates who produced at least one narrative, 54 (i.e. 38.03 per cent) received at least one prompt to produce a narrative. On average, those candidates received 2.11 prompts per exam (range: 1 to 9). Compared to candidates who produced at least one narrative without a prompt, those who received at least one prompt produced more narratives (average: 2.98 versus 2.17, per exam) and produced narratives more frequently (average: once per 123.09 turns versus once per 204.87 turns). To compare the influence that the prompts had on the complexion of the narratives themselves, we can compare candidates who produced one narrative with a prompt against those who produced one narrative but without a prompt (focusing on cases where candidates produced just one narrative ensures that the narrative was actually influenced by the prompt(s)). When we do this, we find that the narratives from candidates who received at least one prompt tend to contain more elements (average: 4.5 versus 3.9) and more unique elements (average: 3.9 versus 3.25). In other words, we might infer that narratives influenced by prompts tend to be longer (structurally) and exhibit a wider range of elements than those that are not supported by prompts. This finding modifies our view of narratives in the TLC analysis in Chapters 2–7. While there we saw that, for the prototypical narratives explored at least, narratives could be produced spontaneously, our exploration of the narratives annotated according to Labov and Waletzky’s scheme shows that the complexity and frequency of narrative in L2 output in general in our data could be linked to examiner behaviour.

Prompts, then, are likely to help candidates to produce more narratives, and to produce narratives that are more complex in terms of the number and range of elements they contain. However useful, this overview is somewhat broad-brush – some of the prompts given could occur within narratives, perhaps scaffolding a narrative, as we saw in Chapter 4, rather than initiating a narrative. To differentiate these from prompts which elicit a narrative from an L2 speaker when no narrative is in train, we placed the examiner prompts linked to narrative into one of two broad categories. Firstly, we have those prompts which occur before the onset of a narrative. We refer to these as initiating prompts, as they initiate the onset of a narrative. These can occur either following the conclusion of a separate narrative or following a non-narrative clause, but they are always followed by the onset of a narrative. The majority of the prompts we identified were initiating, accounting for 71 of the 114 prompts in the 54 files in which at least one prompt occurred (i.e. 62.28 per cent of all prompts). An example of an initiating prompt is given as follows, where this prompt leads into an Orientation at the beginning of a narrative.

The remaining forty-three prompts (i.e. 37.72 per cent of all prompts) occurred within narratives. We refer to these as mid-narrative prompts, and these occur following either a narrative element or a non-narrative clause occurring within a broader narrative. Mid-narrative prompts are always followed by a narrative element. An example of a mid-narrative prompt is given as follows. This instance follows (in fact, interrupts) a Complication and prompts the candidate to produce an Orientation.

In terms of what we have said about narratives in the previous chapters, we note that the majority of the narratives in our sample did not contain a prompt, with most L2 speakers who produced a narrative doing so without any prompt at all (eighty-eight candidates). Of the fifty-four candidates who produced at least one narrative associated with a prompt, we can see that a majority of these narratives were produced as a result of an initiating prompt. Less commonly, they were produced in a way that scaffolded the narrative via a mid-narrative prompt. So, the account of narrative provided in Chapters 2–7, while given nuance by these findings, at a high level stands – most of the time, learners do not produce narratives as the result of a prompt from the examiner. The direct elicitation of a narrative by an examiner when one has not been initiated by an L2 speaker is the minority case.

Another point requires some attention. In Chapter 4 we saw the existence of mid-narrative prompts. Their role in scaffolding was identified in our discussion of the Informational Narrative function in Chapter 4 (see, for example, the discussion of Example 43). We saw the examiner using micro-structures with an Information-Seeking function to scaffold narratives. However, in principle, either initiating or mid-narrative prompts could constitute scaffolding. Is there any evidence, in our qualitative study, that they do? To explore this, we will now explore each type of prompt to provide a descriptive account of each type (i.e. initiating and mid-narrative) in order to understand how each seems to influence the narratives in the data. As this section will demonstrate, the location of the prompt has a large bearing on the form the prompts take, as well as on the effects they produce in the candidates’ narratives.

Overall, in the 54 files containing narrative-related prompts, there were 114 prompts in total. As noted, seventy-one of these were initiating prompts while forty-three were mid-narrative prompts. Initiating prompts tended overwhelmingly to take the form of questions (n = 61; 85.92 per cent). As we have noted, these could occur on the back of the conclusion of another narrative or on the back of a non-narrative clause. Initiating prompts framed as questions almost always occurred on the back of non-narrative clauses (n = 54; 88.52 per cent). This was the case for Example 78. Conversation analytic research has demonstrated how questions placed at the end of turns constitute the primary means through which interlocutors hand over the conversational floor and encourage others to speak (see, for example, Neumaier, Reference Neumaier2023: 59). Such questions might thus also represent one of the main ways in which examiners facilitate candidates’ talk. Cases of initiating prompts following the conclusion of a separate narrative, meanwhile, were rare (n = 7; 14.08 per cent). We have exemplified this as follows, where one narrative about childhood games concludes with an Evaluation, before a prompt from the examiner invites the candidate to begin a new narrative specifically about one of the games the candidate mentioned, hide-and-seek, beginning with an Orientation which provides information concerning where that game was played.

In the majority of cases (n = 43; 70.49 per cent) questions functioning as initiating prompts were followed by Orientations, as demonstrated in Examples 78–80. In the remaining cases, these prompts were followed by narratives which started with a Complication (ten cases), an Abstract (seven cases) and an Evaluation (one case). Perhaps unsurprisingly, the elements of Coda and Result never followed these prompts, as these tended to occur towards the end of the narratives, whereas Abstracts and Orientations tend to occur towards the beginning and Complication and Evaluation can be more flexible in terms of where they occur.

As well as questions, in a minority of cases (n = 10; 14.08 per cent), initiating prompts took the form of requests. These included straightforward requests for tellings (Example 81; which leads to the beginning of a narrative with an Abstract), but also requests for elaboration on detail which then resulted in the beginning of a narrative (Example 82; which follows a non-narrative passage discussing the candidate’s band, and leads to a narrative beginning with an Orientation).

Whatever type of request was issued, these prompts consistently followed non-narrative clauses, as opposed to the conclusion of separate narratives—that is, the function of narrative was elicited where it had not been used in the immediate context. By virtue of their less frequent occurrence (relative to initiating prompts taking the form of questions), we have smaller numbers on which to base our observations about the use of requests. This point notwithstanding, we can see some parallels between these types of initiating prompts as, similar to questions, requests tended to bring about the use of Orientations most often (five cases), followed by Abstracts and Complications (each two cases), with a single use of an Evaluation. Also like the use of questions, requests were not followed by the use of Results or Codas, which again is likely to be due to the tendency for these elements to occur towards the ends of narratives (thus being unlikely to be brought about by initiating prompts). Thus, for initiating prompts, the form taken does not seem to influence the element that follows it, which might suggest that the location of the prompt is more influential. To test this hypothesis further, we can consider prompts located mid-narrative.

So what of the forty-three mid-narrative prompts? Like initiating prompts, mid-narrative prompts also tended to take the form of questions (n = 35; 81.40 per cent), as seen in Example 79 earlier. Questions functioning as mid-narrative prompts could follow any narrative element (except for Coda, which occurs at the end of narratives), and most often followed Orientations (thirteen cases), followed by Evaluations (ten cases), Complications (six cases) and Abstracts (two cases). Although Results tend to occur towards the end of narratives, they could also be followed up with this type of prompt (two cases), and in as many cases these prompts followed non-narrative clauses that occurred within the wider context of narratives.

Switching our focus now to the elements that followed mid-narrative prompt questions, we can see that these types of prompts seem to have a similar effect to the initiating prompts analysed earlier. In particular, in the majority of cases mid-narrative prompt questions were followed by Orientations (twenty cases), and then by Complications (seven cases) and Evaluations (three cases), but never by Codas. There are also differences, though, as the mid-narrative prompt questions did not lead to Abstracts but could be followed by Results (as exemplified in the following dialogue). This is clearly down to the location of these prompts, as the Abstracts occur naturally at the start of narratives, while Results tend overwhelmingly to occur towards the end. In Example 83, an examiner’s mid-narrative question follows a Complication and prompts the candidate to produce a Result of the narrative.

The remaining eight mid-narrative prompts all took the grammatical form of declaratives. This marks another difference between these prompts and those occurring in the initiating position. These declaratives constitute statements which relate in some way to the story being told. Their utterance could be viewed as functioning as a clarifying question, since these statements typically recapitulate some aspect of the narrative immediately preceding them. While the questions analysed take the syntactic form of interrogatives, these declarative statements seem to function in a similar way to questions – as noted, to clarifying questions.Footnote ⁵ Interestingly, while the prompts analysed so far all seem to attract Orientations in terms of what follows them – and, in the case of the mid-narrative prompts, what they follow – mid-narrative prompts taking the form of declaratives tend to follow Complications (four cases), following Orientations twice and a non-narrative clause once.

This type of prompt was followed by Complications (three cases) and Evaluations (three cases), and in one instance each by an Orientation and a Result. This is the only case so far that we have seen of the form of prompt seemingly having a bearing on the type of element that follows it. This seems to be the case because these kinds of prompts occur following the telling of Complications that relay events on which the examiner then seeks clarification. This is perhaps the case because those Complications are, unlike the relatability and relative mundanity of the details told through Orientations, unusual (hence, their tellability). This invites a response from the examiner which either seeks to clarify the Complication, which in turn can invite a response which develops the narrative further (as in Example 84), or which joins in with the appraisal of the events and further contributes to the construction of the story’s tellability (through an Evaluation, as in Example 85).

We are now getting a sense, then, of some of the ways in which prompts can support the development of narratives. Broadly, we have seen that the location of the prompt, relative to the narrative, can have a bearing on the form the prompt takes as well as the kinds of elements that it gives rise to. In the case of prompts occurring mid-narrative, we also saw how these could take the form of declarative statements, in which case both the location and the form of the prompt seemed to culminate to create a particular kind of prompt which tended to occur after – and support – the further narrative development or tellability of Complications. One feature of note for both initiating and mid-narrative prompts is their close relationship to Orientations. Initiating prompts most often elicit Orientations and mid-narrative prompts are often preceded and followed by them. This is of note because there is an echo here of the scaffolding behaviours that adults engage in with children as they are acquiring narrative – Peterson and McCabe (Reference Peterson and McCabe1992), in a study of spontaneous conversations between parents and children, explored questions used as prompts to children during a narrative that were, in essence, Orientations.

Reflecting back on our analyses earlier in the book, we can now address the question of what the prompts are achieving in the interaction. Both initiating and mid-narrative prompts are dominated by questions, exactly the type of information-seeking micro-structure that we observed in examiners scaffolding narratives in the Informational Narrative function in Chapter 4. So, at the micro-structural level, we are seeing the Information-Seeking function used by the examiner, a micro-structural function of examiner speech introduced in Chapter 2. As noted, this is exactly what we saw in the scaffolding of narrative in the discussion of Dimension 4 in Chapter 4. The examiner prompts can be viewed largely as scaffolds, irrespective of whether they are placed initially or mid-narrative. The examiner, following the cooperative principle, is building the conversation in such a way as to permit the student to display a wider knowledge of how to use narrative and to sustain their contribution. This interpretation is also in line with what we have seen regarding the overall effect of the prompts – the students who are prompted are those who produce the longer and more complex narratives. The analysis of the contributions the prompts make to the narrative is strongly suggestive of the points in the macro-structure of narrative where scaffolding is necessary – in the case of mid-narrative prompts, we see that these most often follow Orientations. With initiating prompts, it is the narrative function itself which the examiner is facilitating the L2 speaker to select, and the initial substructure within the narrative which most often starts these elicited narratives is, once again, Orientation. So, the examiner may be argued to either be guiding some students towards an element of narrative which is problematic for the L2 speaker, or intervening after this feature where the L2 speaker fails to show the ability to use that feature.

Of course, this discussion assumes that such prompts are not part of narrative in conversations between L1 British English speakers. Fortunately, we can test this by asking how many of these prompts occur in the TLC L1 data in which narratives have been identified. If we find narrative prompts behaving the same in that data, then we would have to conclude that it was not proficiency that accounts for the prompts – rather, prompts would simply be a feature of narratives given the tasks the speaker is required to undertake in the TLC. However, of the twenty files of the TLC L1 annotated for narrative, only one includes narrative prompts. File 20 includes two narrative prompts, occurring in separate narratives. There are two narratives in total in the file and each starts with an initial prompt from the examiner. What we might conclude here is that prompts are typically not needed for L1 speakers – they produce narratives, and the need for the examiner to elicit, or scaffold, a narrative is limited. In the case of the speaker in file 20, the examiner chooses to elicit two narratives from a student who does not use narrative. However, once elicited, the narratives are produced successfully without further need of support. So, apart from noting that eliciting a narrative may be a tactic used by the examiner to test the student’s ability to produce a narrative, the points made so far stand – the use of prompts in narrative by the examiner relates, to a degree at least, to examinee proficiency.

At this point, we can switch our focus somewhat to consider whether prompts play a role in contributing to the kinds of trends relating to L1 background reported earlier in the chapter. In the next part of our analysis, we therefore consider how the use of prompts by the examiner might support the development of narratives, and whether the use of prompts might help to explain the patterns we have reported so far in this chapter. To begin, Table 9.15 shows what proportion of the candidates who produced a narrative in each L1 group also received at least one prompt. For ease of interpretation, we have ordered this from highest to lowest.

Table 9.15Proportion of candidates who received at least one prompt out of those who produced at least one narrative, grouped by L1 and ranked from highest to lowest.

If we view this table in concert with Table 9.7 presented earlier, we can see that, generally speaking, there seems to be a relationship between age and the proportion of candidates within each L1 group that received prompts, with the L1 Hindi, Spanish (Argentina) and Chinese samples being the youngest on average and having the highest proportion of candidates in the Young category. The L1 Italian and English samples were the oldest on average and contained the smallest number of candidates belonging to the Young category. There seems to be a relationship between candidates’ age and the issuing of prompts, with younger candidates being much more likely to receive prompts than older candidates. This trend is demonstrated by Table 9.16, which shows the proportion of each age category who did and did not receive at least one prompt. Note that, to guard against zero effect, we once again only considered those candidates who produced at least one narrative during their exam. As this table shows, not only were Young candidates the most likely to receive at least one prompt from the examiner, but this is the only age group in which the majority of the candidates received at least one prompt.

Table 9.16Proportion of candidates within each age category who did and did not receive at least one prompt, expressed as percentages of overall number of candidates who produced at least one narrative within each category.

It appears, then, that candidates’ age is likely to drive whether or not examiners use prompts to support the development of narratives. This is an interesting finding as it is in line with previous research on the acquisition of narrative. As noted in the discussion of the Berman model of L2 narrative acquisition, broader questions than the narrowly linguistic come into play at the third and fourth levels of the model. That includes general cognitive development because

So a hypothesis arising from our findings could certainly be that what we are seeing in this data are younger speakers who have yet to achieve proficiency at levels 3 and 4 of narrative production in their own language carrying that across to their L2. This, in turn, guides examiner behaviour.

Having considered the presence of prompts in this broad sense, it is now useful to adopt a more granular perspective on the prompts used with each L1 group, in terms of some of the features of prompts explored to this point (i.e. the location of the prompts, the form the prompts took and what preceded and followed them). Earlier, we found that there were important differences between prompts that initiated a narrative versus those that occurred mid-narrative. Table 9.17 shows this information but for each L1 group. Note that from this point we do not consider the L1 English speakers as this group did not receive a sufficient number of prompts for us to be able to reasonably identify patterns.

Table 9.17Proportion of prompts that were initiating or mid-narrative for each L1 group, expressed as percentages.

As this table shows, for most of the groups the majority of the prompts occurred in the initiating position (as we reported to be the case earlier when taking a whole-sample perspective on the use of prompts). Interestingly, the only group for which this was not the case was L1 Hindi speakers, for whom prompts tended to be mid-narrative rather than initiating, albeit with a fairly even split between the two. This more even balance in the location of the prompts given to L1 Hindi speakers might help to explain why members of this group did particularly well when we considered what proportion could produce each narrative element (i.e. Table 9.6). To test this hypothesis, it is necessary for us to look more closely at the influence of the prompts on the narratives produced by members of each L1 group. To adopt this perspective, we looked again at the elements that followed the prompts provided to each L1 group, beginning with initiating prompts. This is shown in Table 9.18.

Table 9.18Elements following initiating prompts for each L1 group, expressed as percentages.

This table provides some support for the hypothesis set out, inasmuch as the balance between the frequencies of individual elements following initiating prompts are arguably most evenly balanced for L1 Hindi speakers. While Orientations were the element that followed initiating prompts most commonly for all groups, it was least ‘dominant’, in this sense, for the L1 Hindi group (i.e. it accounted for the smallest proportion of elements following this kind of prompt, relative to the other L1 groups). Initiating prompts were followed by a wider range of elements for L1 Chinese and Italian speakers. However, the range between the most and least frequent elements occupying this slot – respectively, Orientations and Complications – is the narrowest, by far, for L1 Hindi speakers (16.67 percentage points; compared to 61.54 (Chinese), 33.33 (Italian), 47.06 (Spanish from Argentina) and 42.86 (Spanish from Spain)). The prompts are the most evenly distributed, then, among L1 Hindi speakers, which suggests that the initiating prompts they received produced a more balanced set of elements.

It is worth bearing in mind, though, that the majority of the prompts given to L1 Hindi speakers occurred in a mid-narrative position. So, we now ask the same question of this type of prompt (Table 9.19).

Table 9.19Elements following mid-narrative prompts for each L1 group, expressed as percentages.

Looking over Table 9.19, we can see similar trends to those we observed in Table 9.18, which again points to a more even balance in terms of the kinds of elements that followed mid-narrative prompts with L1 Hindi speakers relative to the other groups. Firstly, this is the only group for whom mid-narrative prompts were followed by four different types of narrative element. The range between the most and least frequent type of element to follow the mid-narrative prompts for this group was again relatively narrow (26.67 per cent), suggesting something of a balance but a clear preference for Orientations. The most evenly distributed group in this regard were L1 Spanish speakers from Mexico (33.33 per cent each for three separate elements). However, there were only three cases of mid-narrative prompts for this group, compared to fifteen cases for L1 Hindi speakers.

9.5 Conclusion

In this chapter, we have explored the candidates’ use of narratives in their examinations. We took an exploratory approach, wherein from the outset we found that the candidates’ L1 seemed to have a bearing on how and how frequently they produced narratives within their discourse. We found that L1 Hindi speakers performed well across a number of measures of narrative frequency and complexity, which we hypothesised could be related to the cultural significance of storytelling in India, as well as the status of English in the country.

We then used the metadata available to us to try to understand the observed differences between the L1 groups. This led us to another factor, age, as something which distinguished L1 speakers of Hindi from the other groups (being the youngest group on average, comprised exclusively of candidates in the ‘Young’ category). This analysis revealed that there was not a neat relationship between candidates’ ages and the frequency or complexity of the narratives they produced. Thus, the patterns we observed in relation to the L1 Hindi speakers (and, to a lesser extent, the L1 Spanish speakers from Argentina and the L1 Chinese speakers) could not be explained simply by the younger ages of these groups relative to the others, though age could of course be one of a number of factors, including ones not visible to us with the data available.

In the final part of the chapter we considered the context of the exam itself and switched our focus to how the narratives in our data might be influenced by the language produced by the examiner. In particular, we analysed the examiners’ uses of prompts and found that these could vary in terms of their location and form, with the location seeming to be most influential in terms of the particular narrative elements that were produced by the candidates as a result of these. Prompts could initiate a narrative, invite more detail regarding the circumstances surrounding the narrative (through elicited Orientations), advance the narrative’s development (through elicited Complications) and contribute to the construction of the narrative’s ‘tellability’ (through elicited Evaluations). In these ways, examiners can co-construct narratives with candidates, leading to the development of stories that are longer, contain more elements, are more holistic in the range of elements they contain, are more detailed and are more evidently ‘tellable’. While there is evidence that examiners’ prompts could also support the use of Abstracts, which are generally infrequent, we cannot be sure whether or not the candidates would have used these elements without a prompt. Moreover, notably, prompts did not support the use of Codas, which were the least frequent narrative element across the entire sample. Narratives that involved prompts were more complex, then, across a number of measures. While they could not necessarily make up for ‘gaps’ in the candidates’ narrative competence, they served to accentuate competencies that are already there.

When we looked at the intersection of different factors (i.e. L1, age and examiners’ prompts), we saw that the youngest candidates were most likely to receive a prompt from their examiner, that L1 Hindi speakers were the only group that received a higher proportion of prompts mid-narrative than in the initiating position, and that the prompts they received produced a wider and more balanced range of elements than for other groups. If we take L1 Hindi speakers as a case study, then, we can see how both factors tied to the exam context and factors originating outside of that context are likely to combine to influence the use of narratives in this context. In this case in particular, factors such as the prominent role of storytelling in Indian culture, the status of English in India and the age of this sample (with school-aged children perhaps being exposed most regularly to stories as a pedagogical tool) all combined in some way, with the result being that this group produced narratives not only frequently but also with relative complexity. This group was among those that were most likely to receive a prompt (as a proportion of the group overall), which might be explained by the group’s relative youth. L1 Hindi speakers were also the only group to receive a larger proportion of prompts mid-narrative than in the initiating position, and the prompts they received not only produced – but also occurred following – a more diverse and evenly balanced set of elements compared to the other groups. This could be evidence of different factors combining; examiners are more inclined to support younger candidates, and since the L1 Hindi-speaking candidates’ narratives tend to be longer and employ a diverse range of elements, this gives the examiners a larger number and wider range of opportunities to contribute to the co-construction of the narratives through their prompts.