This chapter shifts the analysis to the macro-structural (discourse unit) level. Short-text MDA reveals five dimensions at the discourse unit level (ten distinct functions). This chapter deals with the first three dimensions. The analysis begins with a brief discussion of the first dimension before exploring in depth the second and third dimensions. Throughout the analysis is guided by an exploration of prototypical discourse units – those discourse units most strongly associated with either side of a dimension. This allows an exploration of the roles of the L1 and the L2 speakers in the use of the functions as well as the interaction between discourse unit function and task, level of proficiency and attainment in the examination. These early studies show that that discourse unit functions are sensitive to task in particular and that the role of the examiner in the examination may be seen to vary through discourse unit functions as the proficiency of the L2 speaker increases. The chapter also remarks on links between micro-structural discourse functions and those at the macro-level.

In this chapter the final two dimensions of the TLC are analysed. The four discourse unit functions within those dimensions are once again approached via prototypical discourse units, and task, level of examination and grade of exam are considered as potential sources of variation. Importantly, Narrative emerges in this chapter as a function at the macro-structural level. The analyses show variation by task, level of exam and attainment, and show clearly how the scaffolding behaviour of the examiner influences the selection of micro-structural discourse functions that have an impact on the macro-structural functions present. The chapter argues for the salience of the cooperative principle from Gricean pragmatics as a key organising principle in the discourse observed.

In this chapter the macro-structures in the TLC, its L1/L1 counterpart and the spoken BNC 2014 are compared. The results broadly divide into three groups: discourse unit functions, which are shared across all three corpora; task-specific discourse unit functions; and a number of discourse unit functions unique to individual corpora. The overall findings are that the construct used in the test in the Trinity corpus is a good match, in terms of discourse unit functions, for everyday conversational English, but also that some apparent differences, especially in Dimension 1, are illusory. The analysis of the BNC and the L1/L1 Trinity corpus leads to a revision of the Dimension 1 data for the L2/L1 Trinity corpus, which has the effect of making all three corpora more similar functionally. The chapter also explores the possibility of meso-structures within the discourse units and uses the concept of face to explain some of its findings. Throughout, the presence of narrative is so salient in all three corpora that the chapter concludes with a decision to explore narrative in more detail.