Introduction

Research articles (RAs) constitute the primary medium through which scientific knowledge is disseminated and disciplinary interaction is sustained within academic communities (Tankó, 2017). Researchers from various disciplines in different languages have been striving to carry out and publish research for knowledge development. Central to this communicative process is the abstract, which functions as a structured summary of the study and enables readers and journal editors to make preliminary evaluative decisions regarding relevance and contribution (Bhatia, 1993; Lorés, 2004). As Doró (2013) notes, abstracts render research accessible to a broader readership, while simultaneously performing a “gatekeeping function” by encouraging or discouraging further engagement with the full text (Porush, 1995: 76).

Despite the centrality of abstracts in academic publishing, constructing this section appropriately poses challenges for many researchers, whether writing in a first or second language (Johns, 1992; Ventola, 1994). These challenges stem from factors such as poor general and academic writing skills in the first or second language, limited genre awareness, unfamiliarity with disciplinary conventions, and difficulty recognizing rhetorical variation across fields. Among the structural features of RA abstracts, rhetorical moves (RMs) play a central role in shaping organization and communicative function. Deviations from conventional move patterns may result in abstracts that diverge from disciplinary expectations, potentially affecting textual coherence and evaluative reception. Consequently, the organization of abstracts in terms of RMs has become a central concern in genre-analytic research, motivating extensive empirical investigations across languages, disciplines, and academic contexts (e.g., Amnuai, 2019; Khany and Malmir, 2020; Marefat et al., 2021; Zand-Moghadam and Zhaleh, 2022, to name just a few).

Existing cross-linguistic studies of RA abstracts have typically focused on a single discipline using limited corpora and relied primarily on move-level analysis, limiting their ability to capture fine-grained rhetorical variation across epistemic contexts. Conversely, step-level investigations have tended to examine either one language or one disciplinary domain, leaving the interaction between language and discipline underexplored. To date, no study has systematically examined step-level rhetorical variation across multiple disciplines and languages within a balanced design using recent, comparable journal corpora and employing inferential tests. The present study addressed this gap by adopting a balanced cross-linguistic (English-Persian) and cross-disciplinary (Computers and Electrical Engineering – CEE, Psychiatry, and Applied Linguistics – AL) design, integrating move- and step-level analysis, and applying explicit inferential testing to a controlled corpus of abstracts published between 2022 and 2024. On this basis, and in order to operationalize these analytical objectives, the study formulated the following research questions:

Research Question One: What are the frequent RMs in the English and Persian RA abstracts from Scopus or Web of Science-indexed journals across the various areas of CEE, Psychiatry, and AL published from 2022 to 2024?

Research Question Two: What are the frequent steps of RMs used in the RA abstracts in English and Persian from Scopus or Web of Science-indexed journals across the distinct fields of CEE, Psychiatry, and AL published from 2022 to 2024?

Research Question Three: What are the common RM patterns among the three diverse domains of CEE, Psychiatry, and AL RA abstracts in English and Persian journals indexed in Scopus or Web of Science, published between 2022 and 2024?

The hypotheses of the present study are grounded in genre-based views of academic discourse, which conceptualize RA abstracts as sites where disciplinary epistemologies and rhetorical conventions intersect (Hyland, 2000; Swales, 2004). Disciplines differ in how knowledge is produced, validated, and communicated, and these differences are reflected in the rhetorical structuring of abstracts. Fields such as Psychiatry, which rely on empirical validation involving human participants and standardized reporting norms, are therefore expected to exhibit higher degrees of move conventionalization and more explicit realization of Method- and Conclusion-related steps. By contrast, engineering-oriented disciplines such as CEE, which prioritize problem-solution logics and technological contribution, may place comparatively less rhetorical emphasis on evaluative or inferential concluding functions. AL, positioned between hard and soft sciences, is expected to display greater rhetorical flexibility, particularly at the step and move-pattern levels.

In addition to disciplinary influences, cross-linguistic variation is expected to shape abstract organization. English and Persian academic writing operate within partially distinct institutional, rhetorical, and educational contexts, which may affect how research significance, methodological detail, and contribution are rhetorically compressed or elaborated in abstracts. These differences are treated not as deficiencies, but as conventionalized preferences in abstract construction. Accordingly, the hypotheses predict systematic Language × Discipline effects in move presence, step realization, and move-pattern selection, rather than uniform variation attributable to language or discipline alone. Particularly, the following hypotheses were proposed.

Hypothesis One: The Conclusion move appears significantly more frequently in Psychiatry abstracts than in CEE and AL abstracts, reflecting the discipline’s emphasis on outcome evaluation and clinical relevance, and is more prevalent in English than in Persian abstracts.

Hypothesis Two: Step realizations within the Method move differ systematically by discipline, with Instrument steps occurring more frequently in CEE abstracts and Participant steps occurring more frequently in Psychiatry and AL abstracts.

Hypothesis Three: Persian abstracts contain a higher proportion of Definition steps and a lower proportion of Gap steps than English abstracts, reflecting language-specific rhetorical preferences in establishing research significance.

Literature Review

Move Analysis in RA Abstracts

RA abstracts have long been recognized as a highly conventionalized academic genre, characterized by recurring rhetorical functions that enable writers to present complex studies concisely and persuasively (Doró, 2013; Lorés, 2004). Genre-analytic research has consistently shown that abstracts tend to follow a limited set of macro-structural patterns, typically realized through a sequence of RMs such as contextualizing the research, stating its purpose, outlining methodological procedures, reporting findings, and drawing conclusions (Pho, 2008; Santos, 1996). Across disciplines, these core functions appear relatively stable, suggesting that abstracts serve a shared communicative purpose within academic publishing.

At the same time, the relative prominence and optionality of individual moves are not uniform across research contexts. While studies in AL and related fields have consistently reported high frequencies of Purpose, Method, and Product moves, the Introduction and Conclusion moves are often optional or variably realized (Santos, 1996; Tseng, 2011). In contrast, research in scientific and medical domains tends to exhibit more complete move realizations, particularly where methodological transparency and outcome evaluation are central to disciplinary standards of knowledge validation (Ramadhini et al., 2021; Samraj, 2005). These findings suggest that, despite a shared communicative core, abstract structure is shaped by contextual factors that extend beyond genre conventions.

Cross-Disciplinary Variation

Building on this observation, disciplinary variation has emerged as a key explanatory factor in abstract organization. Comparative genre studies have shown that even closely related fields may differ systematically in how they prioritize rhetorical functions and sequence moves (Samraj, 2005; Suntara and Usaha, 2013). Disciplines grounded in experimental and empirical traditions typically foreground methodological detail and result reporting, whereas fields with interpretive or applied orientations allocate greater rhetorical space to contextualization or implications (Nagano, 2015). From this perspective, disciplinary affiliation does not merely modulate move frequency but reflects deeper epistemological orientations that shape how research is framed, justified, and evaluated in abstract form.

Such differences are not simply stylistic but reflect how disciplines construct and legitimate knowledge. In engineering-related fields, for instance, abstracts often emphasize problem specification, instrumental design, and product outcomes, aligning with a problem–solution epistemology. In contrast, medical and clinical disciplines frequently foreground participant characteristics, procedures, and evaluative conclusions, reflecting their reliance on human-subject research and evidence-based practice. AL, as a hybrid field, often exhibits rhetorical flexibility, combining empirical reporting with contextual and pedagogical framing (Pho, 2008). These disciplinary tendencies highlight the importance of analyzing abstract rhetoric in relation to domain-specific communicative priorities. Supporting this view, Tseng (2011), in a cross-disciplinary analysis of abstracts from applied sciences and humanities, found that experimental disciplines exhibited significantly higher frequencies of Method and Result moves, whereas soft disciplines showed greater variability in Introduction and Conclusion moves. Similarly, Doró (2013) reported that disciplinary membership accounted for substantial variation in move sequencing and optionality in abstracts, suggesting that abstract organization is closely aligned with discipline-specific norms of knowledge presentation.

Cross-Linguistic Variation

In addition to disciplinary influences, language-specific rhetorical conventions play a significant role in shaping abstract structure. Cross-linguistic studies comparing English with other languages have shown that writers may differ in how explicitly they articulate research gaps, define key concepts, or evaluate findings, even when targeting international journals (Johns, 1992; Karimah et al., 2023; Li, 2020). For example, abstracts written in languages other than English have been found to rely more heavily on definitional framing or implicit contextualization, while English abstracts often prioritize explicit gap articulation and outcome evaluation.

Comparative research on English and Persian academic writing consistently reports systematic differences in rhetorical preferences, particularly in the realization of introductory and concluding functions. Across disciplines, Persian abstracts have been shown to rely more heavily on definitional framing and general macro-moves, while English abstracts tend to foreground explicit problematization, methodological specification, and result reporting (Ghasempour and Farnia, 2017; Marefat et al., 2021; Marefat and Mohammadzadeh, 2013; Zand-Vakili and Fard Kashani, 2012). Importantly, these studies converge in showing that such differences persist even in abstracts published in indexed journals, suggesting that language-specific rhetorical conventions are not fully neutralized by globalized publishing norms.

Step-Level Analyses

While move-level analyses have provided valuable insights into abstract macro-structure, they often obscure finer-grained rhetorical distinctions within moves. Step-level analysis addresses this limitation by identifying more specific communicative functions, such as defining concepts, specifying participants, describing instruments, or articulating implications. Studies employing step-level frameworks have shown that steps within the same move can vary substantially in frequency and function across disciplines (Khany and Malmir, 2020; Li, 2020). In a qualitative, cross-disciplinary study, Relawati and Basthomi (2021) investigated the moves and steps of 48 abstracts published in a journal. The findings revealed that most abstracts used Move 3 (Method, 46 abstracts), Move 4 (Product, 44 abstracts), and Move 2 (Purpose, 40 abstracts). In their corpus, authors most frequently applied Step 1 within Move 2 (Purpose), Step 1 within Move 4 (Product),

and Step 3 within Move 3 (Procedure). These patterns suggest that step-level realizations provide critical insight into how rhetorical functions are operationalized within moves, thereby motivating the present study’s focus on step-level variation across both disciplinary and linguistic contexts.

However, step-level research remains comparatively limited in scope. Most studies focus on a single language or a single disciplinary domain, which restricts their ability to account for interaction effects between language and discipline. Consequently, it is still unclear whether observed step distributions are driven primarily by disciplinary epistemologies, linguistic conventions, or their combined influence. Moreover, few studies have integrated step-level analysis with probabilistic or inferential methods capable of testing such interactions explicitly.

Summary and Research Gap

To synthesize the existing literature and clearly position the current study, Table 1 provides a summary of key previous research, highlighting the corpora, foci, findings, and limitations.

In sum, existing research has established that RA abstracts exhibit both cross-disciplinary and cross-linguistic variation in rhetorical organization. Move-level analyses have identified relatively stable macro-structures, while step-level studies have revealed finer-grained functional variation. However, as shown in Table 1, previous research has either examined move-level variation across disciplines or languages or step-level realizations within a single language or discipline, leaving the interaction between language and discipline at the step level largely unexplored. In other words, cross-linguistic studies rarely operate at the step level, and step-level studies seldom adopt a balanced cross-disciplinary design.

To address this gap, the present study adopted a genre-analytic framework capable of capturing both stable macro-structural patterns and finer-grained rhetorical variation. Specifically, it employed a move-step model that allows systematic comparison across languages and disciplines while remaining sensitive to the epistemological assumptions embedded in abstract writing. The following section outlines the theoretical framework underpinning the analysis and explains how the selected model operationalizes RMs and steps for cross-linguistic and cross-disciplinary comparison.

Theoretical Framework

Move analysis provides a functional approach to genre investigation by examining how texts are organized into recurrent rhetorical units that fulfill specific communicative purposes (Swales, 2004, 2019). Since its initial application to RA introductions (Swales, 1981), move analysis has been extended to a range of academic genres, including abstracts, leading to the development of several influential analytic models (Bhatia, 1993; Santos, 1996). These models converge in treating abstracts as functionally structured summaries, but differ in their treatment of purpose statements, methodological detail, and evaluative conclusions, differences that have direct implications for cross-disciplinary and cross-linguistic comparison.

Hyland’s Move-Step Model as an Analytic Lens

This study adopted Hyland’s (2000) five-move model, Introduction, Purpose, Method, Product, and Conclusion, as its primary analytic framework. Unlike earlier models (e.g., Swales’ CARS model), Hyland’s framework was specifically designed to capture the functional organization of abstracts across disciplines. Its emphasis on communicative purpose rather than surface linguistic form makes it particularly suitable for comparative genre analysis.

In this framework, each move represents a distinct rhetorical function, which may be realized through one or more steps. Steps are treated as sub-move units that operationalize how a move’s communicative goal is achieved in discourse. For example, the Method move may be realized through steps specifying participants, instruments, procedures, or data sources. This step-level granularity enables a more precise examination of how rhetorical functions are distributed within abstracts and how they vary across languages and disciplines.

Interpreting Disciplinary Epistemology through Step Bundles

Rather than treating moves and steps as purely formal features, this study interpreted recurrent step combinations as indicators of underlying disciplinary epistemologies. For instance, frequent clustering of Participant and Evaluation steps may signal a discipline’s emphasis on empirical validation and human-subject relevance, whereas repeated Instrument–Product pairings may reflect a problem–solution orientation typical of engineering research. At the same time, the framework acknowledges its limitations: move-step analysis cannot capture all dimensions of rhetorical meaning, such as stance, evaluative intensity, or interpersonal positioning, which fall outside the scope of the present study.

By employing Hyland’s model as an analytic lens rather than a prescriptive template, the study aimed to balance methodological rigor with theoretical sensitivity, allowing for systematic comparison while recognizing the contextual constraints of abstract writing.

Methodology

Research Design

This study employed a corpus-based, cross-linguistic, and cross-disciplinary genre-analytic design to examine rhetorical move-step realizations in RA abstracts. The design integrated qualitative rhetorical coding with quantitative inferential statistics to investigate systematic variation in abstract structure as a function of language (English vs. Persian), discipline (CEE, Psychiatry, AL), and their interaction. This mixed analytic approach allows both fine-grained functional interpretation and statistically grounded comparison across groups.

Corpus Compilation

The corpus comprised 300 RA abstracts, including 150 English and 150 Persian texts. For each language, 50 abstracts were sampled from each of the three disciplinary domains (CEE, Psychiatry, AL), yielding a fully balanced design (Language × Discipline). This balance ensured that no single language or field disproportionately influenced the observed rhetorical patterns.

The selection of disciplines was theoretically motivated. Following Hyland’s (2000) characterization of disciplinary knowledge as existing along a continuum rather than a rigid hard-soft dichotomy, the three domains were chosen to represent distinct epistemological orientations: CEE (engineering-oriented, problem-solution driven), Psychiatry (medical, empirically evaluative), and AL (social science-oriented and interpretive). This configuration enabled systematic investigation of disciplinary effects while avoiding oversimplified binary classifications.

All abstracts were drawn from peer-reviewed journals indexed in Scopus or Web of Science and published between 2022 and 2024, ensuring temporal consistency and alignment with current abstract-writing conventions.

Journal Selection and Quality Control

To ensure cross-linguistic and cross-disciplinary comparability, journals were selected based on explicit and verifiable criteria rather than evaluative labels such as “high-quality” or “high-credit.” Specifically, inclusion required:

1. Indexing tier: Journals had to be indexed in Scopus or Web of Science at the time of data collection. Persian-language journals were additionally required to be indexed in ISC and Scopus to ensure international visibility.
2. Disciplinary scope: Clear alignment with one of the three target disciplines, as indicated in the journal’s stated aims and scope.
3. Abstract policy: Mandatory inclusion of an abstract for all empirical RAs.
4. Author guidelines: Availability of explicit abstract-writing guidelines, including scope and length expectations.
5. Publication window: Articles published between January 1, 2022, and December 31, 2024, to minimize diachronic variation.
6. Language: Abstracts written exclusively in English or Persian.

These criteria were applied uniformly across languages to minimize systemic bias arising from journal prestige, editorial practices, or abstract conventions.

Corpus Selection Procedure

Abstract selection followed a PRISMA-style screening process to enhance transparency and replicability. During the identification stage, 366 abstracts were retrieved from six journals. During screening, 66 records were excluded because they were non-empirical articles, conference abstracts, review papers, or fell outside the specified publication window or word-length range. The final corpus consisted of 300 abstracts (150 English, 150 Persian), evenly distributed across the three disciplines. Detailed corpus characteristics, including journal titles, indexing status, and article counts, are reported in Table 2.

Abstract Format Considerations

The corpus consisted predominantly of unstructured abstracts. However, in the Psychiatry sub-corpus, structured abstracts (e.g., Objective, Methods, Results, Conclusion) were common. To avoid systematic bias, section headings were not treated as coding units. Instead, RMs and steps were identified solely on the basis of semantic content, regardless of typographic structure.

Thus, a sentence appearing under a heading such as Conclusion was not automatically coded as a Conclusion move unless it fulfilled the functional criteria defined in the analytical framework. This procedure ensured functional equivalence across structured and unstructured abstracts. The potential facilitative effect of structured formats on the explicit realization of certain moves, particularly evaluative Conclusion steps, is addressed in the Discussion and Limitations sections.

Analytical Framework

Rhetorical analysis in the present study was conducted using Hyland’s (2000) five-move model, comprising Introduction, Purpose, Method, Product, and Conclusion. This framework was selected because it was explicitly developed for RA abstracts and is designed to capture their functional organization across disciplinary contexts. Compared with earlier models, Hyland’s framework offers greater analytical coverage and flexibility: Swales’ (1990) CARS model does not explicitly accommodate a distinct Purpose move, Bhatia’s (1993) model does not account for niche-establishing or gap-related functions, and Santos’s (1996) model was derived from a relatively small, single-discipline corpus. By contrast, Hyland’s model was constructed on the basis of a large, multi-disciplinary dataset (800 abstracts across eight fields), making it suitable for cross-disciplinary and cross-linguistic comparison. As noted by Suntara and Usaha (2013), this model has demonstrated strong explanatory power in capturing abstract rhetorical structure.

Within this framework, each move represents a distinct communicative function and may be realized through one or more steps. In the present study, moves and steps were operationalized following Hyland’s original definitions. Detailed step descriptions, coding criteria, and borderline examples (e.g., Definition vs. Generalization; Introduction vs. Purpose) are provided in Appendix A to ensure transparency and replicability.

Unit of Analysis and Treatment of Dual Moves

In the present study, the unit of analysis was the clause or clause complex conveying a single dominant rhetorical function. Sentence boundaries were not treated as decisive, as a single sentence may realize multiple rhetorical functions, while a single function may extend across multiple clauses. This clause-based approach is consistent with prior genre-analytic research and allows more precise identification of rhetorical functions in compressed abstract discourse.

In cases where a single textual segment simultaneously fulfilled two distinct rhetorical functions, dual moves (DMs) were assigned (e.g., stating the objective while providing information about designs). To ensure analytical consistency and prevent inflation of move counts, DMs were treated according to the following protocol:

1. For move presence/absence analysis: An abstract containing a DM was considered to possess each of its constituent moves. For instance, an abstract with a Purpose + Method DM was counted as having both Purpose and Method moves present. This approach prioritizes rhetorical function over syntactic packaging.
2. For total move instance counts: In calculating total move occurrences, each DM was counted as one move instance to avoid double-counting. This conservative approach ensures that overall move density calculations are not artificially inflated.
3. For inferential statistical tests: DMs were included as follows: In Fisher's exact tests for move presence, constituent moves within DMs contributed to the presence counts as described in point 1. Moreover, DMs themselves were analyzed as a separate categorical variable to examine rhetorical compression patterns.

Coding Procedure

Coding followed a multi-stage, replicable workflow:

1. Training and calibration: The primary coder studied Hyland’s move-step taxonomy and conducted pilot coding on a subset of abstracts to establish decision rules.
2. Primary coding: All abstracts were manually coded for moves and steps at the clause level. For analytical consistency, each move and step was assigned a standardized alphanumeric code during annotation. Moves were labeled using initial letters (e.g., Introduction = I, Purpose = P, Method = M, Product = PR, Conclusion = C), and steps were coded using move-specific extensions. Clauses realizing two rhetorical functions simultaneously were coded as DM. The complete coding scheme, including all move- and step-level symbols and illustrative examples, is provided in Appendix A.
3. Reliability check: Two coders independently coded 30% of the corpus.
4. Verification and adjudication: Ambiguous cases were revisited, discussed with expert coders, and resolved through consensus.
5. Final coding lock: The dataset was finalized prior to inferential analyses.

Reliability

To establish inter-coder reliability, two expert coders independently reviewed 30% of the corpus (15% English and 15% Persian abstracts within each discipline). Agreement was calculated using Cohen’s Kappa (κ). The results indicated substantial agreement for both moves (κ = .81, 95% CI [.76, .86]) and steps (κ = .76, 95% CI [.70, .82]). Discrepancies were resolved through discussion prior to final analysis.

Data Analysis

Following coding, frequencies and percentages of moves and steps were calculated. Inferential analyses were conducted to evaluate cross-linguistic and cross-disciplinary differences as follows.

1. Research Question One: Presence–absence contrasts for key moves (Introduction, Method, Conclusion) were tested using Fisher’s exact tests, with odds ratios (ORs) and 95% confidence intervals (CIs).
2. Research Question 2: Step-level variation was examined using chi-square tests of independence conducted on contingency tables crossing step realization with Language and Discipline. Cramér’s V was reported to indicate effect size.
3. Research Question 3: Abstract move-pattern preferences were analyzed by calculating predicted probabilities from observed pattern frequencies within each Language × Discipline group, with 95% CIs estimated using standard approximation methods.

Together, these analyses enabled systematic testing of rhetorical variation across languages, disciplines, and analytic levels.

Results

Move Presence and Absence Across Disciplines and Languages

The study’s first research question sought to probe into the RMs used in the English and Persian abstracts of the RAs in the three diverse areas of CEE, Psychiatry, and AL published in peer-reviewed journals indexed in Scopus or Web of Science between 2022 and 2024. Utilizing Hyland’s (2000) five-move framework, the frequency and distribution of RMs were first examined across disciplines and languages (Table 3).

As Table 3 displays, all of the RMs presented in Hyland’s (2000) framework were found across the six corpora. It also indicates that the Product, Purpose, and Method moves were the most frequent ones (i.e., 89.33%, 82.66%, and 81.66%, respectively) in the English and Persian abstracts of the three areas of CEE, Psychiatry, and AL. Only 57% of the papers contained the move of Conclusion in their abstracts. That was the least recurrent RM in the whole corpus. The Introduction move was present in 63.66% of the corpus, indicating that the Introduction move was not consistently realized across abstracts.

Disciplinary Contrasts in Move Presence

From a cross-disciplinary perspective, a marked difference among the three disciplines was observed in using the Conclusion move, which appeared differently in the RA abstracts from quantitative and qualitative aspects. This move was compulsory and more common in the Psychiatry abstracts, compared to those of CEE and AL. While almost all Psychiatry abstracts (98%) ended with the fifth move (Conclusion), only 24% of the CEE abstracts contained this move, although the CEE journals have similar word limit policies and explicitly require authors to state the purpose of their research, principal results, and major conclusions (see their Author Guidelines sections). Psychiatry abstracts showed near-categorical realization of the Conclusion move, coinciding with the widespread use of structured abstract formats in this discipline. Abstracts of AL were somewhere between these two, manifesting 54% of the occurrence of Move 5. To statistically test these contrasts, Fisher’s exact tests were conducted, yielding the following outcomes shown in Table 4.

Table 4 indicates that in AL, the Conclusion move appeared in 44/50 English abstracts (88%) but only 10/50 Persian abstracts (20%). A Fisher's exact test confirmed a highly significant language effect in AL (p = 4.59×10⁻¹²). The odds of an AL abstract in English containing a Conclusion move were about 29 times the odds for an AL abstract in Persian (OR = 29.33, 95% CI [9.77, 88.03]). Comparing disciplines (combined languages) revealed that Psychiatry abstracts almost always included a Conclusion move (98/100) while CEE abstracts rarely did so (24/100); this difference was extreme (p < .0001; OR ≈ 155.17, 95% CI [35.56, 677.09]). These contrasts quantify and confirm the patterns described above, supporting the first hypothesis: Psychiatry, within this corpus, favors Conclusion moves, whereas CEE and Persian abstracts are far less likely to include them.

Another set of contrasts pertained to the Method move, existing in almost all the Psychiatry abstracts (97%). However, several abstracts in the other two disciplines were observed without using this move. To assess such disciplinary variation in the use of this move, the researcher compared its presence across CEE (85/100), Psychiatry (97/100), and AL (63/100) abstracts. Table 5 displays the results.

As shown in Table 5, pairwise Fisher tests confirmed that Psychiatry employed Move 3 significantly more than both CEE (p = 0.0048; OR = 3.52, 95% CI [1.44, 9.17]) and AL (p < 0.0001; OR = 15.35, 95% CI [5.28, 50.11]). CEE also exceeded AL in Method-move use (p < 0.0001; OR = 4.43, 95% CI [2.32, 8.70]). These contrasts reinforce the pattern that Psychiatry abstracts most consistently elaborate methodological details, AL abstracts least often do so, and CEE abstracts occupy an intermediate position.

A further cross-disciplinary contrast was attributed to the use of DMs. While it was common to include more than one move in a single segment in the AL abstracts (40%), rarely did the CEE and Psychiatry abstracts embed moves in one segment (17% and 5%, respectively). The most frequent embedded moves (i.e., 20) occurred between the moves of Introduction and Purpose or Purpose and Method in AL. To statistically test these contrasts, Fisher’s exact tests were conducted, suggesting the following results (Table 6).

As shown in Table 6, DM usage differed significantly across disciplines. Fisher’s exact tests revealed that AL abstracts were significantly more likely to contain embedded rhetorical functions than those in CEE (OR = 3.25, 95% CI [1.69, 6.28], p = .0005). An even stronger contrast was observed between AL and Psychiatry, where AL abstracts were over twelve times more likely to employ DMs (OR = 12.67, 95% CI [4.73, 33.89], p < .000001). In contrast, DM usage was relatively rare in Psychiatry abstracts (5%), indicating a strong preference for functionally discrete clause-level realizations. These results provide inferential confirmation that rhetorical compression through DMs is discipline-sensitive rather than incidental.

Cross-Linguistic Contrasts in Move Presence

From a cross-linguistic perspective, the English abstracts were longer than their Persian counterparts, containing more moves in general (i.e., English abstracts: 585 moves; Persian abstracts: 476 moves). More particularly, the English abstracts comprised the Conclusion section more than the Persian ones. The most considerable difference between these two languages was found in the English and Persian AL abstracts concerning this section. While most of the English AL abstracts (88%) contained this mandatory move, just 20% of the Persian AL abstracts presented the Conclusion move. Additionally, only three English CEE abstracts lacked the Product move, while 11 Persian abstracts in this major did not employ this move. Another distinction between the abstracts in the two languages was discovered regarding the occurrence of the Introduction move. In comparison to Persian, the English abstracts used the Introduction move, mainly focusing on prominence and gap, more frequently. To evaluate cross-linguistic differences in the use of Move 1 (Introduction), the researcher compared its presence across all abstracts in English (111/150) and Persian (80/150) (see Table 7).

Based on Table 7, a Fisher's exact test revealed a statistically significant difference (p = 0.00012). The odds that an English abstract included the Introduction move were approximately twice those for a Persian abstract (OR = 2.38, 95% CI [1.48, 3.82]). This result confirms that English abstracts more consistently foreground backgrounding and contextualization than their Persian counterparts, which included Move 1 less often across all three disciplines.

Step-Level Distributions Across Disciplines and Languages

The next research question of the study focused on the frequent steps of RMs of the English and Persian RA abstracts from high-quality journals of the three distinct domains of CEE, Psychiatry, and AL from 2022 to 2024. Adopting Hyland’s (2000) framework for the RM organization of abstracts, the researcher estimated each step’s frequency and percentage in the abstracts of the three fields in the two languages. (Table 8).

Overall, as Table 8 demonstrates, 1760 steps were observed in 300 RA abstracts across the three majors of CEE, Psychiatry, and AL in English and Persian with a mean of 5.86. It can be claimed that every abstract, of any field in any language, was, on average, comprised of about six steps. While the Purpose step was the most frequent (16.76%), followed by the Product and Instrument steps (16.3% and 14.6%, respectively), the Generalization, Evaluations, and Definition steps were the least commonly used ones by the authors (0.96%, 1.7%, and 3.06%, respectively).

Disciplinary Variation in Step Distributions

Psychiatry abstracts employed the most steps (692), followed by AL (583) and CEE (485). This indicates that while the average Psychiatry abstract used 6.92 steps, the CEE one used 4.85 steps. Similar to the previous comparisons, AL fell between the two majors with a mean of 5.83. While the most frequent step pattern in the CEE abstracts was Prominence-Gap-Purpose-Instrument-Procedure-Product, the Prominence-Gap-Purpose-Participant-Instrument-Procedure-Product-Conclusion-Implication pattern was the most common one in the abstracts of Psychiatry. In the AL abstracts, the Prominence-Gap-Purpose-Participant-Instrument-Product-Implication pattern was the most favorable one. It can be seen that their first three common steps were almost the same (Prominence, Gap, Purpose).

Cross-Linguistic Variation in Step Distributions

English abstracts generally took more steps than the Persian ones (i.e., 955 and 805 steps, respectively). The most regular step sequence for the abstracts in English was the Prominence-Gap-Purpose-Participant-Instrument-Procedure-Implication, but in Persian, the Prominence-Purpose-Participant-Instrument-Product step pattern was used the most. Four major divergences are observable between the two languages. First, the frequency of the Gap step varied noticeably from English to Persian. While this step comprised 9.21% of the total steps in the English abstracts, only 3.85% of the Persian steps were for presenting the Gap. Second, the Persian authors used fewer Conclusion-related steps (10.42%) than their English counterparts (15.69%). More specifically, implementing the Evaluation step by English and Persian researchers was different. The English abstracts presented the evaluation of their findings in the last move more than their Persian counterparts (2.4% and 0.86%, respectively). Third, unlike the English abstracts, some Persian abstracts began with definitions. According to Table 8, the Definition step was more favored by the Persian authors in comparison to their English counterparts (4.47% vs. 1.88%, respectively). The Persian authors tended to define new emerging things or concepts, which are mainly derived from other languages such as English or Arabic, in the abstract section to ease comprehension and avoid any confusion. Finally, no research questions or hypotheses were directly mentioned in the English abstracts; however, some Persian abstracts did this within the Purpose step.

To analyze the step-level data, a series of chi-square tests of independence were conducted for each step to examine the language and discipline interaction, yielding the following results in Table 9.

Move-Pattern Preferences and Probabilities

The study’s final research question attempted to identify the most conventional RM sequences of the RA abstracts from distinguished journals within the three distinct fields of CEE, Psychiatry, and AL in English and Persian from 2022 to 2024. To address this question, the researcher extracted and examined the patterns of abstracts one by one, following Hyland’s (2000) framework for RM patterns (Table 10).

Analysis of the abstracts suggested the existence of 11 move patterns, four of which most frequently appeared in the corpora. Only a few abstracts (10 out of 300, 3.33%) did not follow any particular move patterns and were considered out-of-order abstracts. As illustrated in Table 10, the authors employed the pattern of five moves, Introduction-Purpose-Method-Product-Conclusion, in their abstracts as the most frequently used one (38.33%). The next three commonly used patterns were Introduction-Purpose-Method-Product (23.33%), Purpose-Method-Product-Conclusion (17%), and Purpose-Method-Product (12.33%).

Disciplinary Variation in Move Sequences

The majority of the Psychiatry abstracts followed the most standard patterns; around 96% used either the Introduction-Purpose-Method-Product-Conclusion (64%) or Purpose-Method-Product-Conclusion (32%) pattern. Moreover, there was a substantial difference between the CEE and Psychiatry abstracts concerning their RM patterns. While the most regular sequence in the former major was Introduction-Purpose-Method-Product (43%), no single abstract was found in the latter implementing this pattern. The AL abstracts fell somewhere between the two again, using this pattern as the second most common pattern (27%). However, the five-move order, with 38% frequency, was the most preferred move pattern among AL researchers. Overall, the Psychiatry abstracts had the most standard and uniform move patterns with only a few minor deviations.

Cross-Linguistic Variation in Move Sequences

Purpose-Method-Product was the most common pattern among the Persian abstracts of CEE, accounting for 34%. A perceptible difference between the abstract move patterns in English and Persian was that there were no non-canonical sequences in English, whereas 10 (6.66%) Persian abstracts deviated from canonical move sequencing, including partial or atypical move realizations. For example, some Persian abstracts used only one or two moves, some had only an introduction instead of an abstract, and others used moves in the non-canonical positions. A recurrent deviation involved the realization of the Purpose move prior to the Introduction move. Some Persian abstracts were organized in more than one paragraph. Another difference was seen in the length of Move 4, as Persian Product moves were generally shorter than their English counterparts. There was more discrepancy among the Persian abstracts, particularly in the field of AL, compared to the English abstracts regarding their RM patterns.

To quantify move pattern differences, predicted probabilities were computed directly from observed proportions, with 95% CIs based on normal approximation. Table 11 shows the predicted probabilities of common move patterns by language and discipline.

As illustrated in Table 11, across all disciplines, the four predominant move patterns showed clear and interpretable differences by language. Psychiatry displayed the highest degree of pattern standardization: English abstracts in this corpus favored the canonical I-P-M-PR-C pattern (p = .70, 95% CI [.57, .83]), with Persian abstracts showing a similarly strong preference (p = .58, 95% CI [.44, .72]). In CEE, English abstracts showed more dispersion across patterns, with I-P-M-PR dominating (p = .56) and I-P-M-PR-C occurring less frequently (p = .20). Persian CEE abstracts showed a more balanced distribution, with I-P-M-PR (p = .30) and P-M-PR (p = .34) both common.