Journal Pre-proof Metadiscourse choices in EAP: An intra-journal study of JEAP

3 Interest in language variation is a staple of English for Academic Purposes research and 4 underpins its distinctive character as a field of inquiry. It is the specific nature of 5 language use which defines EAP, yet this definition has been established almost entirely 6 on the basis of inter-discoursal studies, with comparisons of register, genre, discipline, 7 first language, etc. dominating our understanding. In this paper we take a different 8 approach and focus on variation within the field, and specifically within its flagship 9 journal, JEAP . Categorising every paper between volume 1 and 52 as principally taking 10 a textual, critical, contextual or pedagogical orientation, we explore writers’ preferences 11 for metadiscourse use. The differences which emerge can be attributed to the 12 argumentation preferences of sub-fields and their knowledge-making practices. The 13 findings offer evidence of intra-disciplinary variation in discoursal preferences and 14 hopefully contribute to our understanding of both the journal and our field.

Interactive resources allow the writer to manage the information flow to establish his 215 or her preferred interpretations. They include: 216 • transitions -devices (mainly conjunctions) used to mark additive, contrastive, and 217 consequential relations (in addition, but, thus, however). 218 • frame markers -refer to text boundaries or structure, including items used to 219 sequence, label stages, announce discourse goals and indicate topic shifts (finally, 220 to conclude, my purpose is). 221 • endophoric markers -make additional material salient to help the reader recover 222 the writer's intentions by referring to other parts of the text (noted above, see Fig,223 in section 2). 224 • code glosses -restates information (for instance, in other words).

Data and analysis 248
We created a JEAP corpus by gathering all empirical articles published in the journal 249 from volume 1 in 2002 to volume 52 in July 2021. The collection included papers in 250 all regular and special issues but excluded editorials, book reviews and brief reports. 251 This produced a corpus of 441 articles of over 3 million words. As noted above, we 252  (1) First, despite its macrostructure, the RA should have a separate 294 Discussion section. Second, the first available drafts had to have been 295 written by the students (the first authors) on their own. 296 (2) the sample encompassed the first, second, third, and fourth year 297 differences of normed frequency were statistically significant. The spreadsheet allows 305 a comparison between more than two corpora, so the raw frequencies and total words 306 of each corpus were entered with the resulting LL and significance (p) values as output 307 indicating whether the differences were significant. 308

332
We were surprised at the extent of the variation in the use of metadiscourse types as 333 we had assumed there would be greater similarities in such an apparently 334 homogeneous field. One possibility which occurred to us was the potential 335 influence of NNS authorship on the use of metadiscourse as Author 1 and Author 2 336 (2021a), for example, found that most papers in EAP are authored by Asian and 337 European authors and these may favour textual papers. However, there is little 338 evidence for this and reviewing and editing processes involved in a paper's journey to 339 publication appear to rule out the influence of NNS authorship on rhetorical choices 340 (Englander, 2006; Author 1, 2015). The journal itself is relatively non-directive in its 341 recommendations regarding style, with the guide to authors simply specifying the 342 need for inclusive language. But while writers have a certain leeway in creating 343 their connections between ideas and with readers, their choices appear to be 344 constrained by the topics they select, most obviously in the difference between12 345 textual and pedagogical papers. In the following sections we explore these choices in 346 more detail. 347 348

Interactional metadiscourse in different strands 349
Interactional metadiscourse represents the writer's intrusion into the text to comment 350 on what is being discussed or directly address the reader. It concerns the participants 351 of the interaction. Table 2   We also see in the table that textual and critical papers, which typically offer tentative, 374 and often even speculative, explanations for the language patterns they report in texts, 375 are most heavily hedged. Thus trying to explain the decline of a language feature over 376 time can involve relatively tortuous prose: Possibly as a result of changing audiences with less knowledge of specialist 380 techniques.
(Textual) 381 Reporting pedagogical research, on the other hand, allows for greater assurance in 382 explaining results: 383 (5) Indeed, the answers to questions prompting the participants to 384 elaborate on measures they took to enhance validity showed that most 385 were unfamiliar with the validity evidence they needed to attend to in 386 order to address the types of validity that concerned them most. 387 (Pedagogical) 388 389 Pedagogical papers also contain significantly fewer self-mentions. This was surprising 390 as we had assumed that many of these studies would involve the writers' personal 391 experiences of classroom practice, but this does not seem to be the case. Instead, we 392 find discussions of textbook materials, tasks, programmes, curricula and classroom 393 methods which are not directly related to the reporting authors' experiences. This can 394 be illustrated by these brief extracts from two presentations of methods: 395 (6) A content analysis method was used to study the feedback comments. 396 Each sampled feedback comment was coded into a set of evaluative 397 statements relating to the Task Fulfilment and Organisation category, with 398 each coded evaluative statement assigned to a sub-category. All coding 399 was performed by the author, using Microsoft Excel. audience, tends to be less studied than stance and is always far less frequent in 442 research articles. Despite this, it is a key means by which writers present themselves as 443 sharing, or perhaps failing to share, attitudes with readers and how they manage 444 solidarity and affiliation. Another feature which stands out in Table 4  (Pedagogical) 504 Because directives seek to engage and position readers, they carry strong connotations 505 of unequal power, claiming greater authority for the writer (Author 1, 2001). This 506 seems to be most apparent with necessity modals which seem to impose far more on 507 the reader than imperatives, yet despite this, modals remain a common rhetorical 508 option in our corpus. 509 510 Asides and questions are far less common in JEAP papers, although questions do crop 511 up in critical and pedagogical papers. Questions, of course, are at the heart of all 512 academic inquiry, but only occasionally surface in research papers. Their appearance 513 invites readers into the text by addressing them as having an interest in an issue and 514 the good sense to follow the writer's response to it (Author 1, 2016). 515 516 The use of questions seems to be particularly attractive to those working in critical and 517 pedagogical areas. In the former questions not only help capture readers' curiosity 518 (Chang & Schleppegrell, 2011), but also provide an orientation for them; a frame 519 where they can be led to the writer's viewpoint (Author 1, 2002). 520 (20) Are language studies journals therefore accepting the use of we as 521 a means for single author self-reference? Should they do so? How about 522 sentences such as "This paper thinks/believes? (Critical) 523 In pedagogical papers, the questions posed appear to be more involving, posing 524 issues that readers might ask when considering their own classroom practices: 525 (21) How do participants in advanced academic writing courses learn 526 to analyze genre examples when they are introduced to the genre 527 analysis framework outlined in Swales (1990)? The answer to this 528 question seems unclear in Swales (1990).
(Pedagogical) 529 So questions, while relatively rare in these JEAP papers, can play a useful rhetorical 530 role for writers in some areas. 531 532

Interactive metadiscourse use in different strands 533
Interactive features are those which organise a text to help readers recover the writer's 534 intentions, creating surface cohesion and influencing understandings of propositional material. They function to link material, offer elaborations, signal text stages and refer 536 to information elsewhere in the text. As such, they not only help glue the text together 537 but also represent an internal dialogue with readers, reflecting the writer' assessment 538 of what needs to be done to present information in the most comprehensible and 539 convincing way for particular readers. 540 541 Table 2 shows significantly different frequencies of interactive features across the 542 strands, with critical and pedagogical papers containing substantially fewer forms. 543 Writers of the textual papers, it seems, often go to considerable lengths to spell out the 544 connections between ideas when trying to explain reasons for their findings: Here we see the authors pressing frame markers, transition signals, code glosses and 554 endophorics into service to account for their results. 555 556 In contrast, this extract of similar length taken from an interpretive passage in a critical 557 paper presents an argument with a sparse use of interactive forms. We see only an 558 endophoric 'here' referring back to a previously discussed student text, a contrastive 559 marker and an evidential. 560 (23) Zohra's writing decisions here do not stem from a critical incident 561 or serve as an example of linear transfer often highlighted by studies she gives us an insight into the uptake that guided her ultimate transfer: 564 offering an expanded example of the metaphor for transfer that "adaptive remediation" (p. 35). (Critical) 568 Clearly this is a very different kind of discussion with less reflexive signalling and overt 569 structuring. The interpretation is focused on a single case supported by a reference without 570 the need for greater elaboration. 571 572 We find similarly low frequencies of interactive forms in the pedagogical papers, 573 which, like many of the critical papers, are focused on particular case study subjects, 574 as here: 575 (24) Lee joined a group with two NSE students due to seating proximity. 576 She was a little nervous as she knew at least Diane was a very strong 577 student. "She's very serious to her study, and always has great points." 578 (weekly chat) Lee's group decided to write a listening test for level III 579 students at the ELI. The entire test consisted of two parts focusing on 580 improving listening skills such as paraphrasing, inferencing, 581 understanding the main idea and listening for details. (Pedagogical) 582 This narrative-style account of a student's participation in a teacher education writing 583 group contains just one evidential, citing the 'weekly chat' record as the source of the 584 quote and a code gloss expanding 'listening skills'. Despite the lack of reader 585 assistance, however, the extract is readable and easy to follow. The differences in individual interactive features can be seen in Table 5. While all 609 differences are significant, we note that textual and contextual papers contain more 610 code glosses, frame markers and transitions and that writers of critical and contextual 611 papers used fewer endophorics and evidentials. 612  and 'comment' were used to report academics' views.
(Textual) 622 (26) Quoted segments consisting of less than a one T-unit were marked as 623 "phrasal" and then coded according to their phrasal structure: "verb phrase," 624 (defined as one or more inflected verbs plus any modifiers, objects or 625 complements accompanying them), "noun phrase," (defined as one or 626 more nouns plus any modifiers accompanying them) or "adjective phrase" 627 (defined as one or more adjectives plus any modifiers accompanying them). 628 We can see here a very different argument style than that in (31). Instead of pointing 686 outside the text for supporting testimony for an argument, these writers lay out an 687 account of connected events in a linear way. In both cases information is presented 688 almost as a story. In (32) concessive connectives predominate as the writer holds our 689 interest by confounding our expectations while in (33) there is a preference for 690 additive markers with a concessive used to shift the focus of the argument. 691 692 Finally, frame markers. These extend transitions by helping to organise the text and 693 reduce the readers processing effort by explicitly marking the structure of the served similar functions in their quantitative and qualitative RAs and, once again they 696 are most frequent in contextual papers and least common in pedagogical ones. Announcing the goal of the research is a function used in all four areas of study, an 700 unsurprising finding as the need for clear objectives is crucial to encourage readers to 701 engage with the paper and evaluate its effectiveness. 702 (Pedagogical) 718