Code Glosses in Moroccan Masters’ Thesis Introductions: Elaboration, Exemplification, and Definition

Code Glosses in Moroccan Masters’ Thesis Introductions: Elaboration, Exemplification, and Definition

 

Khalid Benraiss

Ibn Tofail University, Kenitra, Morocco

 

Abstract

Code glosses are interactive metadiscourse devices that help readers process academic texts by rephrasing, exemplifying, or defining propositional content. Despite their pervasive presence in academic writing, they remain among the least studied categories in metadiscourse research (Pearson and Abdollahzadeh, 2023). This study investigates code gloss deployment in 30 Moroccan master's thesis introductions (34,050 words) drawn from two institutions: the Faculty of Languages, Letters and Arts (FLLA) at Ibn Tofail University, Kenitra (16,898 words; 2011-2014) and the Faculty of Education Sciences (FSE) at Mohammed V University, Rabat (17,152 words; 2017-2019). Analysis was conducted using Hyland's (2005) framework, Prommas' (2020) marker taxonomy, and MAXQDA Analytics Pro 2020. Three rhetorical functions are identified: exemplification, paraphrasing, and definition, each with a distinct distribution across CARS moves. Corpus examples illustrate each function in context. Findings are contextualised against Pearson and Abdollahzadeh (2023), Swales (2019), and recent work on elaborative metadiscourse, with implications for understanding how Moroccan L2 writers negotiate disciplinary knowledge accessibility and for the design of genre-responsive writing instruction.

Keywords: code glosses; elaboration markers; exemplification; metadiscourse; L2 academic writing; CARS model; Morocco; move-based analysis

 

 

 

 

 

1. Introduction

Academic writing requires the management of a fundamental asymmetry: the writer possesses knowledge that the reader may not, and the degree of this asymmetry varies depending on the reader's disciplinary familiarity, the technical density of the text, and the genre's expectations regarding accessibility. Code glosses, in Hyland's (2005) terminology, are those interactive devices that help readers grasp the writer's intended meaning by reformulating, exemplifying, or defining propositional content. They are among the primary tools through which this asymmetry is negotiated in academic writing.

In the thesis introduction specifically, code glosses serve a dual function: they make technical concepts accessible to readers who may not yet be fully familiar with the specialist terminology of a field, while simultaneously signalling the writer's command of the conceptual apparatus they are deploying. The use of such as, that is, for example, and i.e. to clarify, elaborate, or define terms is not merely a pedagogical courtesy. As Hyland (2007, p. 284) argues, "code gloss signals are therefore a crucial element of a text's meaning as they work to relate a text to its context by taking readers' needs, understandings, existing knowledge, intertextual experiences, and relative status into account." Their deployment is therefore a rhetorical act that positions the writer as a knowledgeable guide through the conceptual landscape of the introduction.

Despite their pervasive presence in academic prose, code glosses occupy a relatively marginal position in the metadiscourse research literature. Pearson and Abdollahzadeh (2023) report that code glosses appeared as a focus in only a minority of the 370 studies they reviewed, with research attention concentrated instead on hedges, boosters, and self-mentions. Approximately 37% of corpus-based research followed what they term the "thin" tradition, emphasising marker frequency counts over contextually-bound interpretations. Swales (2019, p. 77) made a related point, arguing that purely lexical analyses "bore little to no benefit to students and teachers" unless connected to the communicative contexts in which markers appear. The present study responds to both calls by combining quantitative frequency analysis with move-level functional interpretation and corpus-extracted examples from the data.

Within the MENA region, code gloss analysis remains an underexamined area of inquiry, despite a growing body of research on how Moroccan and Arab L2 writers deploy metadiscourse resources across academic genres (Benraiss, 2023a, 2023b; Benraiss and Koumachi, 2023). The Moroccan context presents a particularly interesting case: writers operate across a trilingual educational landscape in which Arabic, French, and English coexist as instructional media, and the academic writing instruction available to graduate students has been characterised as limited in its explicit treatment of rhetorical and metadiscoursal conventions. Whether and how code gloss use reflects these contextual pressures, or whether it is governed primarily by genre demands that operate above the level of institutional and linguistic context, is one of the questions this study addresses.

The study is organised as follows. Section 2 reviews the relevant literature on code glosses in metadiscourse theory, L2 academic writing, and the CARS model. Section 3 describes the corpus and analytical procedures. Section 4 presents the frequency and functional findings. Section 5 discusses the results in relation to prior research. Sections 6 and 7 address limitations and pedagogical implications respectively, followed by a conclusion.

2. Literature Review

2.1 Code Glosses in Metadiscourse Theory

The concept of metadiscourse refers broadly to language that comments on the discourse itself rather than advancing propositional content. Vande Kopple (1985) and Crismore et al. (1993) introduced early taxonomies of metadiscoursal devices, treating elaboration markers as part of a broader textual apparatus for guiding readers. Hyland and Tse (2004) reconceptualised this apparatus along interpersonal lines, distinguishing between interactive resources that organise the text for the reader and interactional resources that engage the reader in dialogue. Code glosses belong to the interactive category alongside frame markers, transitions, endophoric markers, and evidentials.

Hyland (2005) defines code glosses as "devices that supply additional information by rephrasing, specifying the meaning of, or providing examples of adjacent material" (p. 87). The primary forms identified include: parenthetical brackets (...), exemplification markers (such as, for example, e.g., including, like), paraphrase markers (that is, in other words, i.e., or rather), definitional phrases (called, defined as, known as, refers to), and appositive constructions, that is, nominal restatements without explicit marker. Their defining characteristic is their reflexive function: they comment on the code itself, signalling that the writer recognises that language requires supplementation to guarantee comprehension. In Hyland's (2007, p. 266) terms, writers who use code glosses are "applying a gloss" to their own discourse, demonstrating audience awareness in the act of writing.

Theoretically, code glosses perform what Halliday and Matthiessen (2014) term the "elaborating" logico-semantic relation. They do not add new propositional content but restate or specify existing information in a new register or through concrete instantiation. This elaborative function distinguishes them from other interactive devices: transitions add new propositional steps, frame markers signal rhetorical structure, and evidentials attribute sources, while code glosses concern themselves exclusively with ensuring that the content already stated is accessible to the reader. This narrow but important function makes code glosses a distinctive index of the writer's orientation toward the reader.

Swales (2019) raised important questions about parenthetical brackets specifically, noting that their functions vary considerably across disciplines and text types. While Hyland's framework treats brackets consistently as code glosses, Swales observed that in humanities writing they frequently serve as scholarly asides, hedged speculations, or allusions to expert readers, functioning more as signals of disciplinary membership than as reader-oriented elaboration. This concern is directly relevant to the present study, given that the disciplinary context is applied linguistics, a field positioned between the social sciences and humanities. Not all brackets in this corpus will serve the same reader-oriented function, a consideration that informed the manual verification step in the methodology.

Ädel (2023) has further argued that metadiscourse research needs to move beyond a purely marker-based approach toward a more contextualised analysis of the discourse functions performed by metadiscourse devices. This call resonates with Swales' (2019) critique and motivates the move-integrated functional analysis adopted in the present study.

2.2 Code Glosses in L2 Academic Writing

Research on code gloss use in L2 academic writing, though limited in volume relative to studies of hedges or boosters, has produced consistent findings. Studies comparing L1 and L2 writers consistently show that L2 writers use exemplification markers at rates broadly comparable to expert writers but rely on a narrower inventory of forms (Mauranen, 1993; Crismore et al., 1993; Dehghan and Chalak, 2016; Barabadi and Aghaee, 2021). The default forms in L2 writing are typically such as, for example, and e.g., while expert L1 writers deploy a richer range that includes appositive constructions, nominal restatements, and multi-word exemplification frames. This pattern suggests that L2 writers acquire the prototypical forms of elaboration early but do not develop the stylistic range that characterises expert academic prose.

Hyland (2004), in a large-scale study of metadiscourse in 120 doctoral and 120 master's theses by Hong Kong students across soft and hard disciplines, found that soft-discipline writers use more interactive metadiscourse overall, with code glosses appearing as a regularly deployed resource for managing terminological complexity. The study documented differences between master's and doctoral writers in the depth and variety of their code gloss use, suggesting a developmental trajectory in elaboration sophistication. Prommas (2020), in a cross-disciplinary and cross-cultural analysis of 116 master's theses by New Zealand and Thai students, found that English Language Teaching thesis writers used more interactive metadiscourse than their Business Administration counterparts, with code glosses constituting a consistent feature of the more discursive ELT genre.

Vijayakumar (2024) examined exemplification specifically in a learner corpus of student essays, finding that student writers use exemplification frequently but in limited syntactic configurations, favouring sentence-initial for example over more integrated appositive and relative clause forms. The study argued that EAP courses pay relatively little explicit attention to exemplification despite its high frequency, leaving students to develop elaboration practices through exposure rather than instruction. This finding resonates with Pearson and Abdollahzadeh's (2023) observation that metadiscourse instruction remains underdeveloped relative to the research on its importance.

Jiang and Hyland (2025), comparing metadiscourse in student essays and AI-generated texts, found that student writers are more likely than the AI model to use code glosses, precisely because students draw on a genuine sense of audience to identify where readers need elaboration. As the authors note, reformulations "signal that the writer recognises the original formulation did not achieve its purpose," a kind of metacommunicative monitoring that reflects audience awareness developed through real communicative experience. This finding underscores the interpersonal dimension of code gloss use that Hyland's framework foregrounds: elaboration is not merely a cognitive act of clarification but a social act of reader acknowledgement.

For thesis writing specifically, the genre imposes additional constraints on code gloss deployment. The master's thesis requires writers to operationalise key terms, define the scope of their investigation, and establish conceptual clarity for a reader who is at once an expert evaluator and a potential sceptic. This operationalisation function requires definitional code glosses of a particular precision, quite different from the casual exemplification that might suffice in a review or discussion section. Al-Harthi et al. (2022), in a cross-disciplinary study of interactive metadiscourse in academic articles, found that code glosses were consistently deployed to manage the tension between disciplinary insider knowledge and accessible exposition, a tension that is especially acute in thesis writing where the writer's competence is simultaneously on display and under evaluation.

2.3 Code Glosses in the CARS Model

The Create-a-Research-Space (CARS) model, developed by Swales (1990) and revised in 2004, describes how research introductions typically organise their rhetorical work across three moves: establishing a territory (Move 1), establishing a niche (Move 2), and occupying the niche (Move 3). Each move has a distinct communicative purpose, and it follows that the metadiscourse resources deployed in each move should reflect those distinct purposes. Code glosses in particular would be expected to vary in type and function across the three moves, since the knowledge management demands of establishing disciplinary context, identifying a gap, and presenting one's own research are quite different.

In Move 1, writers survey established disciplinary literature and claim the significance of their topic. The primary code gloss function here is exemplification: writers give concrete instances of the phenomena they are reviewing to make abstract claims about the field tangible. This is consistent with Hyland's (2005) observation that exemplification is the dominant elaboration strategy in sections where writers introduce their disciplinary territory. In Move 2, where writers identify gaps, limitations, or problems in existing research, code glosses continue to serve exemplification but may also shift toward paraphrasing: writers clarify what the identified gap means or restate it in different terms to ensure the reader grasps its significance. In Move 3, where writers present their own research design and objectives, definitional code glosses become essential: writers must operationalise key terms, specify the scope of their investigation, and establish the conceptual precision that the research requires.

Del Saz-Rubio (2011), in her meta-move analysis of agricultural science research article introductions, found that code glosses appeared across all three CARS moves but were most densely concentrated in Move 1, where terminological clarity was most urgently required. She noted that exemplification was particularly frequent in sections where writers surveyed the established disciplinary landscape, while definitional devices clustered in sections where writers made specific methodological or theoretical commitments. The present study tests whether this distribution holds in the contextually different genre of master's thesis introductions in applied linguistics.

Cao and Hu (2014), in a comparative study of interactive metadiscourse across research paradigms in applied linguistics, found that writers in different research traditions deployed code glosses differently: qualitative researchers used more exemplification and definitional glossing than quantitative researchers, reflecting the greater need to define constructs and illustrate phenomena when the research approach is itself interpretive. This disciplinary and paradigmatic variation is relevant to the present study, given that the FLLA and FSE corpora represent programs with somewhat different theoretical orientations within applied linguistics, a difference that may help explain the marker-level variation identified in the results.

It is worth noting that the CARS model is not always realised in a canonical sequence. Research on L2 writers in various contexts has documented move recurrence, co-occurrence, and non-canonical ordering (Benraiss and Koumachi, 2023). When Move 2 and Move 3 co-occur or alternate repeatedly, as found in Moroccan thesis writing, the functional boundaries between move-specific code gloss uses become less clear-cut. The present study takes this structural flexibility into account by analysing code gloss functions in terms of the actual rhetorical purpose served in context rather than by move label alone.

3. Methodology

This study examines code gloss deployment in a corpus of 30 master's thesis introductions (34,050 words). Fifteen introductions were drawn from the FLLA program at Ibn Tofail University in Kenitra (16,898 words; 2011-2014) and fifteen from the FSE program at Mohammed V University in Rabat (17,152 words; 2017-2019). All writers are Moroccan L2 English speakers enrolled in Applied Linguistics and TEFL programs. The corpus was constructed to enable comparison across institutions, programs, and time periods while controlling for genre (master's thesis introduction), language (English), and broad disciplinary domain (applied linguistics).

Code gloss markers were identified through MAXQDA Analytics Pro 2020 using the Prommas (2020) taxonomy. Each identified instance was manually verified against Hyland's (2005) operational definition: only markers functioning to elaborate or explain adjacent propositional content were coded as code glosses. Parenthetical brackets were included only where they served definitional or exemplificatory elaboration, not where they functioned as bibliographic references or precision markers. This distinction was informed by Swales' (2019) concern about the multifunctionality of parentheticals in academic prose and required case-by-case contextual judgement.

Each code gloss instance was additionally assigned to one of three functional categories: exemplification (markers introducing concrete instances to illustrate general claims), paraphrasing (markers restating propositional content in alternative terms), and definition (markers specifying the meaning of key constructs as part of the writer's own research framing). Each instance was also assigned to the CARS move in which it appeared, following Swales' (2004) three-move model and the move identification procedure established for the corpus. Unmarked appositive elaborations were excluded from quantitative analysis owing to annotation reliability constraints. Normalised frequencies (Nf per 1,000 words) and log-likelihood (LL) statistics were computed. For a 10% inter-rater reliability sub-sample, Cohen's kappa (Cohen, 1960) was kappa = 0.89.

4. Results

4.1 Overall Frequencies

Sub-corpus	Raw Count	Nf (per 1,000 words)	LL	p
FLLA	98	5.79	0.03	>.05
FSE	102	5.94

Table 1. Overall code gloss frequencies by sub-corpus

The two sub-corpora show virtually identical code gloss frequencies, with a log-likelihood value of 0.03 and p > .05. This near-identical rate is the most striking quantitative finding in the dataset. It indicates that the tendency to elaborate and clarify propositional content is consistent across the two institutions, two programs, and the eight years separating the corpora. The shared recognition of the need to deploy code glosses appears to operate above the level of institutional and temporal variation, suggesting that the genre itself, the master's thesis introduction, imposes consistent elaboration demands on writers regardless of where or when they are writing.

4.2 Marker-Level Distribution

Rank	FLLA Marker	FLLA Nf	FSE Marker	FSE Nf
1	(...)	1.59	such as	1.51
2	such as	1.24	that is	1.05
3	that is	0.95	(...)	0.81
4	specifically	0.47	specifically	0.52
5	namely	0.41	namely	0.46
6	i.e.	0.36	i.e.	0.41
7	in other words	0.30	in other words	0.35
8	for example	0.24	for example	0.35
9	e.g.	0.12	e.g.	0.29
10	defined as	0.11	defined as	0.19

Table 2. Top 10 code gloss markers by sub-corpus (Nf = per 1,000 words)

While overall frequencies converge, the leading marker differs between sub-corpora in a revealing way. In FLLA, parenthetical brackets (...) lead (Nf = 1.59), suggesting a preference for embedded terminological clarification: writers place technical terms or alternative formulations in brackets alongside their main expression, allowing the elaboration to proceed without disrupting the syntactic flow of the sentence. In FSE, such as leads (Nf = 1.51), indicating a preference for explicit exemplification as the primary elaboration strategy. The marker inventory used by both groups is narrow, concentrated in the top three to four forms, a pattern consistent with the general finding in L2 academic writing that novice writers rely on a limited range of high-frequency defaults rather than the fuller range available in academic English (Hyland and Milton, 1997; Adel and Erman, 2012). Eight of the top ten markers are shared between the two sub-corpora, confirming the convergent character of the category overall even as the preference ordering differs.

4.3 Functional Distribution Across CARS Moves

Function	Primary Markers	Move 1	Move 2	Move 3
Exemplification	such as, for example, e.g., including	High	Medium	Low
Paraphrasing	that is, in other words, i.e., or	Medium	Medium	Medium
Definition	defined as, called, namely, (...)	Low	Low	High

Table 3. Code gloss functions across CARS moves

The functional distribution shows that different types of code gloss predominate in different parts of the introduction. Exemplification is most concentrated in Move 1, where writers survey the disciplinary field, and remains present in Move 2, where writers identify research gaps. Paraphrasing is distributed relatively evenly across all three moves, serving a general comprehension-support function throughout the introduction. Definitional code glosses are rare in Moves 1 and 2 but spike in Move 3, where writers establish their own research parameters and operationalise key constructs. The following corpus examples illustrate each function in context.

Exemplification in Move 1 (illustrating disciplinary phenomena through concrete instances):

"L2 reading is a complex process that involves the interaction of a series of processes and knowledge. It includes decoding skills (e.g., letter recognition), higher level cognitive skills (e.g., inferencing) and interactional skills (e.g., agreeing or disagreeing with the author's point of view)." (FSEMA-AL10)

Exemplification in Move 2 (specifying the nature of an identified gap or problem):

"... it has to deal with limitations such as inadequacy of time and exam-oriented language programs (Richards, 2015). In addition to this, other obstacles can be found. For example, in major public Moroccan universities, there are crowded English language classes and interrupted schedules." (FSEMA-AL3)

Paraphrasing distributed across moves (restating a proposition to ensure comprehension):

"... we take it for granted that speaking is much more important and difficult than listening, the latter involves a lot of hidden thinking processes. Since speaking is overt, it appears to be more active than listening. However, this is not true as listening is an act that requires a number of subskills such as predicting, guessing, associating, understanding, recognising discourse markers, remembering and the like." (FLLAMA-AL12)

Definition in Move 3 (operationalising key constructs for the study):

"For the purposes of this study, metadiscourse refers to those linguistic devices through which writers signal their awareness of the ongoing text and their relationship to the reader, namely, frame markers, transitions, evidentials, and code glosses." (FLLAMA-AL4)

These examples show that code gloss deployment is not random: writers select different elaboration strategies depending on what the move requires. In Move 1, examples give concrete form to abstract disciplinary claims. In Move 2, examples make the gap visible and specific. In Move 3, definitions provide the conceptual precision that the reader needs to evaluate the research on its own terms. Paraphrasing, meanwhile, serves as a continuous monitoring function, offering alternative formulations wherever the writer judges the original statement may be insufficiently clear.

5. Discussion

The virtually identical frequencies across the two sub-corpora (LL = 0.03, p > .05) is the most striking finding in the dataset. It indicates that code gloss use is governed primarily by the requirements of the genre rather than by institutional, temporal, or program-specific factors. Writers at FLLA and FSE, writing at different times and in different institutional environments, converge on essentially the same rate of elaborative marking. This finding aligns with the theoretical expectation that interactive metadiscourse is calibrated to the communicative needs of the genre (Hyland, 2005), and it extends that expectation specifically to code glosses, suggesting that the thesis introduction imposes a consistent elaboration imperative that overrides contextual variation.

The differential preference for brackets in FLLA versus such as in FSE is subtle but theoretically meaningful. Brackets allow writers to include alternative terminology without explicitly foregrounding the choice, absorbing terminological ambiguity passively without interrupting the main discourse flow. Such as, by contrast, actively introduces exemplification as a rhetorical move, directing the reader's attention to the concrete instances being provided. The FSE preference for such as may reflect a more explicit elaboration style, consistent with the more structured academic writing pedagogy that characterises the applied sciences education available to FSE students. It may also reflect disciplinary differences within applied linguistics: FLLA writers, trained in literary and linguistic theory, may be more accustomed to parenthetical precision as a genre convention, while FSE writers, oriented toward language teaching practice, may default to explicit illustration. The Cao and Hu (2014) finding that writers in qualitative traditions use more exemplification than those in quantitative traditions provides a partial parallel, though the direction of the effect in the present data is not straightforwardly interpretable in paradigmatic terms.

The concern raised by Swales (2019) regarding the multifunctionality of parentheticals is worth addressing directly in relation to the FLLA data. The high frequency of brackets in FLLA raises the question of whether all these instances function as reader-oriented code glosses or whether some serve instead as scholarly asides addressed to expert readers. The manual verification step in the methodology aimed to exclude non-code-gloss bracket uses, but the possibility of residual ambiguity remains, particularly given that applied linguistics writers routinely use brackets to signal theoretical allegiance or to hedge claims in ways that shade into scholarly signalling rather than pure elaboration. Future research using more fine-grained discourse-functional annotation would help clarify this.

The dominance of definitional code glosses in Move 3 confirms the genre-theoretical expectation that this move is where writers commit to specific conceptual and methodological frameworks. Del Saz-Rubio (2011) found a similar concentration of definitional devices in the equivalent move in agricultural science research articles, and the present data extend that observation to the master's thesis introduction genre in applied linguistics. The present data also reveal that exemplification is not confined to Move 1: it persists into Move 2, where writers use concrete examples to make their identified gaps visible and persuasive. This exemplification-in-Move-2 pattern suggests that writers use code glosses not merely to illustrate established phenomena but to make the case for their research: by grounding the gap in specific, observable instances, they strengthen the argument that the gap is real and worth addressing. This observation enriches the existing CARS-based account of Move 2 rhetoric.

Viewed against the broader landscape established by Pearson and Abdollahzadeh (2023), the present findings demonstrate the value of combining quantitative precision with move-level functional interpretation. The systematic review showed that approximately 37% of metadiscourse research follows what they call a "thin" approach, reporting frequencies without contextual interpretation. The functional analysis presented here goes further, showing that the same markers serve different rhetorical purposes depending on where they appear in the introduction. Such as in Move 1 illustrates disciplinary terrain; such as in Move 2 specifies the nature of a research gap; neither function is captured by a frequency count alone.

The multilingual educational context of Morocco further contextualises these findings. Writers at both institutions operate in a trilingual environment where Arabic, French, and English coexist as instructional media. The genre-level convergence in code gloss rates suggests that the rhetorical pressure to elaborate and clarify is strong enough to operate across this multilingual context, overriding whatever differential elaboration norms writers may carry from Arabic or French academic writing traditions (Chibi, 2021). At the same time, the narrow marker inventories documented here, with writers concentrated in three or four default forms across a much larger available toolkit, suggest that genre recognition is not the same as genre mastery. Writers know that elaboration is required; they are less certain about how to elaborate richly.

6. Limitations and Future Research

The present study is limited by its focus on two institutions within a single country and a single broad discipline. The convergence finding, while robust within this dataset, should be replicated across broader geographic and institutional samples, including STEM disciplines, before being accepted as a general principle of L2 academic writing. Within applied linguistics itself, the present corpus represents only two sub-disciplines; whether writers in other areas of the field show the same marker preferences and move distributions remains to be investigated.

The exclusion of unmarked appositive elaborations from the quantitative analysis likely underestimates total elaborative marking, particularly in FSE writing. Future research should develop annotation protocols for appositive code glosses using explicit syntactic criteria to enable reliable identification. The concern raised by Swales (2019) regarding the multifunctionality of parenthetical brackets also points to a methodological limitation: finer-grained discourse-functional annotation distinguishing reader-oriented elaboration from scholarly precision and hedged speculation would produce a more nuanced account of bracket use specifically.

The FLLA corpus (2011-2014) and the FSE corpus (2017-2019) span eight years. While temporal differences did not emerge as salient in the quantitative analysis, longitudinal designs tracking individual cohorts would better isolate temporal from institutional effects. Future research should also investigate the relationship between code gloss use and writing quality: if richer code gloss repertoires correlate with higher-rated thesis introductions, this would provide direct empirical support for corpus-based code gloss instruction. Longitudinal research tracking individual writers from undergraduate to postgraduate writing would illuminate the developmental trajectory of elaboration strategy acquisition in the Moroccan L2 English context.

7. Pedagogical Implications

The convergence of code gloss frequencies across sub-corpora, combined with the functional differentiation across CARS moves, suggests that Moroccan master's writers have internalised the genre imperative to elaborate technical content, but that they may benefit from instruction in diversifying their elaboration repertoire. The high frequency but narrow inventory pattern documented here, in which writers converge on a small set of default markers while underusing the fuller range available, points toward a specific pedagogical gap: students know they need to elaborate but have limited tools for doing so.

Instruction should make explicit the distinction between exemplification strategies, most appropriate in Moves 1 and 2, and definitional strategies, most appropriate in Move 3. Students who currently rely on such as and for example across all three moves may be applying a single elaboration strategy where genre convention calls for a more differentiated approach. Corpus-based instruction using authentic examples from published research articles in applied linguistics, selected to show high-elaboration and low-elaboration versions of Move 3 passages, could help writers understand when and how to shift from exemplification to definition as the introduction progresses.

Specifically, writers should be introduced to the full range of definitional code glosses available in academic English, including appositive constructions, nominalisations, and extended definitional phrases, beyond the default such as and that is. Writers who default to "Metadiscourse is important, such as in thesis introductions" can be guided toward "Metadiscourse, those interactive devices through which writers signal their stance and guide reader comprehension, plays a central role in thesis introductions." The latter uses an appositive construction to integrate definition into the sentence flow. This move-specific instruction, grounded in authentic corpus data, should follow the cyclical model of introducing, familiarising, and practising that corpus-based EAP pedagogy recommends (Hyland, 2004; Vijayakumar, 2024).

8. Conclusion

Code glosses in Moroccan master's thesis introductions represent the most institutionally stable metadiscourse category documented in this study, with virtually identical frequencies across FLLA and FSE writers. This convergence suggests that elaborative marking is a genre-level imperative that overrides institutional and temporal variation. Below the surface of this frequency convergence, however, meaningful differences in preferred markers, brackets in FLLA versus such as in FSE, and move-specific functional strategies, exemplification in Moves 1 and 2, paraphrasing across all moves, definition concentrated in Move 3, reflect distinct elaboration approaches shaped by disciplinary orientation and pedagogical context. Corpus examples drawn from the data give empirical grounding to these move-level functional distinctions.

These findings respond to Swales' (2019) call for contextually grounded analysis of what metadiscourse markers do at specific communicative points in a text, and to Pearson and Abdollahzadeh's (2023) call for "thick" functional analysis that goes beyond frequency counting. The finding that students' code gloss choices, while limited in range, are functionally purposeful challenges simple deficit narratives about L2 academic writing. Writers at both institutions deploy elaboration strategies that match the rhetorical demands of each section of the introduction. What they lack is not the instinct to elaborate but the breadth of tools with which to do so. This is a precise and actionable finding for writing instruction, and it constitutes the main contribution of this study to the understanding of interactive metadiscourse in L2 academic writing in the MENA region.

References

Adel, A., and Erman, B. (2012). Recurrent word combinations in academic writing by native and non-native speakers of English: A lexical bundles approach. English for Specific Purposes, 31(2), 81-92.

Ädel, A. (2023). Adopting a "move" rather than a "marker" approach to metadiscourse: A taxonomy for spoken student presentations. English for Specific Purposes, 69, 4-18.

Al-Harthi, M., Alshahrani, H. J., Hamed, D. M., and Ibrahim, W. M. A. (2022). Evidentials, code glosses, hedges and boosters in academic articles: A cross-disciplinary study. Eurasian Journal of Applied Linguistics, 8(3), 181-200.

Barabadi, E., and Aghaee, E. (2021). The differential use of reformulation markers in three sub-corpora: L1 English, L2 English, and L1 Persian. Teaching English as a Second Language Quarterly, 40(1), 1-32.

Benraiss, K. (2023a). Interactional metadiscourse in applied linguistics research article abstracts of Moroccan L2 writers of English: A small corpus investigation. International Journal of English Language Studies, 5(1), 23-31.

Benraiss, K. (2023b). Establishing a niche in applied linguistics master thesis introductions of Moroccan L2 graduate writers of English. Eastern Journal of Languages, Linguistics and Literatures, 4(2), 49-60.

Benraiss, K., and Koumachi, B. (2023). Rhetorical moves in applied language research article introductions of Moroccan L2 writers of English: A generic move analysis. Eastern Journal of Languages, Linguistics and Literatures, 4(3), 1-13.

Cao, F., and Hu, G. (2014). Interactive metadiscourse in research articles: A comparative study of paradigmatic and disciplinary influences. Journal of Pragmatics, 66, 15-31.

Chibi, M. (2021). Collectivism and individualism as cultural aspects in Arabic/English argumentative writing by Moroccan students. International Journal of English Literature and Social Sciences, 6(4).

Cohen, J. (1960). A coefficient of agreement for nominal scales. Educational and Psychological Measurement, 20(1), 37-46.

Crismore, A., Markkanen, R., and Steffensen, M. (1993). Metadiscourse in persuasive writing. Written Communication, 10(1), 39-71.

Dehghan, M., and Chalak, A. (2016). Code glosses in academic writing: The comparison of Iranian and native authors. Research in English Language Pedagogy, 3(2), 21-29.

Del Saz-Rubio, M. M. (2011). A pragmatic approach to the macro-structure and metadiscoursal features of research article introductions in Agricultural Sciences. English for Specific Purposes, 30(4), 258-271.

Halliday, M. A. K., and Matthiessen, C. M. I. M. (2014). Halliday's introduction to functional grammar (4th ed.). Routledge.

Hyland, K. (2004). Disciplinary interactions: Metadiscourse in L2 postgraduate writing. Journal of Second Language Writing, 13(2), 133-151.

Hyland, K. (2005). Metadiscourse: Exploring interaction in writing. Continuum.

Hyland, K. (2007). Applying a gloss: Exemplifying and reformulating in academic discourse. Applied Linguistics, 28(2), 266-285.

Hyland, K., and Milton, J. (1997). Qualification and certainty in L1 and L2 students' writing. Journal of Second Language Writing, 6(2), 183-205.

Hyland, K., and Tse, P. (2004). Metadiscourse in academic writing: A reappraisal. Applied Linguistics, 25(2), 156-177.

Jiang, F. K., and Hyland, K. (2025). Rhetorical distinctions: Comparing metadiscourse in essays by ChatGPT and students. English for Specific Purposes, 79, 17-29.

Liu, Y., Tang, R., and Lim, F. V. (2023). The use of code glosses in three-minute thesis presentations: A comprehensibility strategy. Journal of English for Academic Purposes, 1-14.

Mauranen, A. (1993). Cultural differences in academic rhetoric: A textlinguistic study. Peter Lang.

Pearson, W. S., and Abdollahzadeh, E. (2023). Metadiscourse in academic writing: A systematic review. Lingua, 293, 103561.

Prommas, P. (2020). Metadiscourse markers in academic writing: An updated categorisation. Journal of Applied Linguistics, 37(1), 56-78.

Swales, J. M. (1990). Genre analysis: English in academic and research settings. Cambridge University Press.

Swales, J. M. (2004). Research genres: Explorations and applications. Cambridge University Press.

Swales, J. M. (2019). The futures of EAP genre studies: A personal speculation. Journal of English for Academic Purposes, 38, 75-82.

Vande Kopple, W. J. (1985). Some exploratory discourse on metadiscourse. College Composition and Communication, 36(1), 82-93.

Vijayakumar, P. (2024). Exemplification in student essay writing: A study of learner corpus of essay writing (LCEW). International Journal of Applied Linguistics, 34, 1514-1532.