Reporting Verbs in Chinese MA Theses in Linguistics vs. International Linguistics Journal Articles

Abstract

Reporting verbs are commonly employed in academic writing, yet second language learners often encounter uncertainties regarding their appropriate usage during the writing process. This corpus-based study investigated the use of reporting verbs in Chinese MA theses in linguistics and international linguistics journal articles. It focused on the semantic categories of reporting verbs, the source types and the subject types of reporting clauses, as well as the interrelationships among these elements. The results showed the following: (1) regarding the semantic categories of reporting verbs, authors of MA theses tended to use SHOW verbs, whereas authors of journal articles were likely to employ ARGUE verbs; (2) in terms of reporting sources, while both groups of authors most frequently used textual references, authors of MA theses showed a secondary preference for uncited generalizations, whereas authors of journal articles more often employed self-sourced reporting; (3) when using reporting verbs, both groups of authors employed non-human subjects most frequently, followed by human and it subjects; (4) the usage of reporting verbs by two groups of authors revealed close interrelationships among the semantic categories of reporting verbs as well as the source types and the subject types of reporting clauses. Implications for L2 learners’ academic writing and EAP teachers’ instruction were also discussed.

1. Introduction

Reporting verbs are one of the most important elements in scholarly writing and are primarily used to attribute the content to non-author sources (Thomas & Hawes, 1994; Ma & Qin, 2017; Unudom & Unudom, 2020). The term “reporting verbs” was first introduced by Swales (1990), who characterized them as verbs used in statements to introduce the findings or arguments of previous researchers. Later, Thompson and Ye (1991) briefly described them as “verbs used in citations” and pointed out that reporting verbs can be used not only to state the writers’ own opinions but also to express emotional attitudes towards others’ views. Reporting verbs, essential for crafting evidence-based arguments (Kwon et al., 2018), enable writers to effectively integrate sources with their own ideas. Moreover, they are also one of the grammatical means for writers to express their stances and authorial voice in academic writing (Hyland, 1999; Jafarigohar & Mohammadkhani, 2015; Yilmaz & Özdem Ertürk, 2017). When citing claims, the use of reporting verbs is closely associated with the credibility of such claims and is essential both in formulating the writers’ own claims and positioning these claims in prior scholarly research (Bloch, 2010). Despite extensive research on their rhetorical importance (e.g., Charles, 2006b; Bloch, 2010; Kwon et al., 2018), reporting verbs remain a persistent challenge for many L2 novice writers, especially for Chinese MA students who often struggle to interpret the pragmatic nuances behind different verb choices. Understanding how L2 novice writers use reporting verbs is therefore crucial for revealing how academic stance and evaluative meanings are developed in L2 academic writing.

Academic discourse represents a specialized genre, covering academic articles and theses. A prominent identifying mark of academic articles that draws researchers’ interest lies in the reported viewpoints or references to previous studies. Usually, these are introduced via reporting verbs (Thomas & Hawes, 1994; Zhou & Sun, 2022). In the realm of academic writing, writers are confronted with a set of essential tasks. Firstly, they are tasked with creating a coherent and trustworthy self-image and presenting their research in a credible way. Secondly, they also need to handle their interaction with the discourse community. To be specific, this process involves linking their ideas to prior studies and showing alignment with the community’s accepted knowledge (Hyland, 1999). Reporting verbs hold a pivotal position in academic writing. As a vital rhetorical means for forming citations (Lee et al., 2018), they are employed to convey information retrieved from other sources and position the current research within its relevant academic field. Particularly, authors should clearly state and appraise others’ viewpoints, which helps to illustrate how their current work either extends or modifies previous statements (Pramoolsook, 2021). Evidently, the choice of proper reporting verbs in academic writing is of immense importance. Nevertheless, numerous Chinese English learners face challenges in comprehending the pragmatic functions and implicit meanings of reporting verbs. Consequently, they find it difficult to accurately select reporting verbs to express their stances and create an academic research space in academic writing.

In the past, many scholars have examined reporting verbs in academic writing (e.g., Jafarigohar & Mohammadkhani, 2015; Yilmaz & Özdem Ertürk, 2017; Kwon et al., 2018; Lee et al., 2018; Zhou & Sun, 2022). Since the use of reporting verbs can also be influenced by the interaction of elements within reporting clauses, researchers should also take these elements into account when studying reporting verbs. However, only a few scholars have focused on the source types and the subject types of reporting clauses as well as the interrelationships between them in academic discourse when analyzing reporting verbs (Charles, 2006b; Kwon et al., 2018). Moreover, existing studies have rarely compared reporting verb use across corpora that differ both in writer expertise and genres (e.g., Hyland, 2002; Kwon et al., 2018). Addressing this gap is crucial for informing L2 academic writing pedagogy and helping authors of MA theses cultivate more effective reporting and stance-taking strategies. In view of this research gap, this study aims to compile two corpora: Chinese MA theses in linguistics (CMTL) and international linguistics journal articles (ILJA). These corpora will be used to study the use of semantic categories of reporting verbs, the source types and the subject types of reporting clauses, as well as the interrelationships among these elements. By comparing these two corpora, the study contributes empirical evidence that deepens our understanding of how reporting verbs function across different levels of writer expertise and genres and offers implications for improving L2 academic writing instruction.

2. Literature Review

2.1. Classification of Reporting Verbs

Many scholars have put forward different classification criteria for reporting verbs (e.g., Thompson & Ye, 1991; Thomas & Hawes, 1994; Francis et al., 1996; Charles, 2006a). However, some of them have their own limitations. Thompson and Ye (1991) initially proposed a classification model and analytical framework for reporting verbs. Analyzing reporting verbs mainly from the perspectives of denotation and evaluation, they put forward a classification model for reporting verbs. However, the subclassifications in this model overlapped with each other, and some reporting verbs were categorized into two subclassification groups simultaneously. Thus, although this model laid the foundation for the classification of reporting verbs, it is not clear enough. Subsequently, based on 129 reporting structures in 11 medical journal articles, Thomas and Hawes (1994) classified reporting verbs into three categories: discourse verbs, real-world verbs, and cognitive verbs. While this classification criterion contributed to the research on the classification of reporting verbs, it faced limitations such as the limited corpus size and disciplinary constraints because it was restricted to medical journal articles. As a result, this criterion cannot be directly applied to academic writing across all disciplines. Building on previous research, Francis et al. (1996) developed a new semantic classification model and divided reporting verbs into four categories according to their semantic meaning: ARGUE verbs, SHOW verbs, FIND verbs, and THINK verbs. Then, the explanation of each verb group was adapted by Charles (2006b). Based on Francis et al.’s (1996) model and Charles’ (2006b) refined explanations, Kwon et al. (2018) classified “note” under FIND verbs and incorporated the phrase “talk about” to ARGUE verbs to make the categorization more coherent. In this study, we utilize Charles’ (2006b) semantic classification framework for reporting verbs, along with his explanations of each category, and also integrate Kwon et al.’s (2018) revised classification, which specifies the particular reporting verbs within each category. Further details on how this framework is applied in the analysis are provided in Section 3.1.

2.2. Relevant Studies of Reporting Verbs

Strong citing skills enable writers to precisely situate their research within a specific academic field, highlight their own academic propositions, and clearly express their evaluations of the reported propositions (L. X. Yang, 2013; Ma & Qin, 2017). To cultivate this ability, writers must have an in-depth understanding of the diverse characteristics of reporting verbs in academic writing (Eckstein et al., 2022). Among the existing studies on reporting verbs, some have focused on reporting verbs in all citation structures (Pickard, 1995; Hyland & Jiang, 2019). However, other studies have solely concentrated on reporting verbs within the “v + that” structure, which is a standard form in academic writing (Charles, 2004; Hyland & Tse, 2005; Parkinson, 2013a, 2013b). While most research on reporting verbs adopts a comparative approach, the study by Kwon et al. (2018) specifically focused on L2 writers. They explored first-year L2 writers’ use of reporting verbs in academic writing, examining semantic categories, rhetorical functions of reporting verbs, and the relationship between them to inform pedagogical improvements. From a broad perspective, other related studies can be primarily categorized into several groups: research on reporting verbs across disciplines, under different mother tongue backgrounds and different academic levels.

First, we review research on the use of reporting verbs across different disciplines. Hyland (1999) investigated the use of reporting verbs across eight disciplines. His findings revealed that writers in the humanities and social sciences were significantly more likely to employ discourse reporting verbs than writers in the natural sciences and engineering. These differences were attributed to contextual factors and deeply embedded epistemological conventions rather than simple agreement. For instance, writers in the humanities and social sciences often use these verbs to explicitly construct a discursive framework for their arguments, whereas writers in the natural sciences and engineering tend to employ research verbs to report cumulative experimental activities. In addition, Charles (2006a) analyzed the phraseological patterns in clauses formed by reporting verbs in politics/international relations and materials science. The findings revealed that the most frequent reporting patterns in both disciplines were integral citations with human subjects. Later, Charles (2006b) studied disciplinary variation and how stance related to the grammatical and semantic patterns of reporting verbs. He discovered that, in comparison with the natural sciences, social sciences generally permit a higher level of writer presence or visibility. More recently, Eckstein et al. (2022) investigated the distribution of reporting verbs across six disciplines in 270 academic research background sections. Their study demonstrated significant disciplinary variations in both lexical diversity and frequency distribution of reporting verbs. There may be significant differences in the patterns of using reporting verbs across different disciplines, yet such disparities are sometimes not observed within the sub-fields of the same discipline. For instance, Suntara and Usaha (2013) did not find any notable differences in the reporting verbs used in articles on linguistics and applied linguistics.

Next, we examine research on the use of reporting verbs by researchers with different native languages. Liu and Zhou (2014) conducted a comparative analysis of stance-building strategies within reporting clauses employed by English and Chinese writers in the field of applied linguistics. Their findings reveal that ESL (English as a Second Language) writers tend to utilize implicit affirmation to express their own viewpoints and strong-attitude verbs when referring to others’ perspectives. In contrast, native English speakers often opt for neutral verbs accompanied by intensifying expressions. Later, Yeganeh and Boghayeri (2015) compared the frequency and functions of reporting verbs in international journal articles written by English writers and national journal articles written by Persian writers. Their analysis showed that the two groups differed from each other when using reporting verbs, though the most frequently occurring pattern was the same. Further, Liardét and Black (2019) analyzed reporting verb usage by EAL (English as an additional language) and native learners within the framework of appraisal theory, contrasting their patterns with expert writers’ practices. Similarly, Marti et al. (2019) established four corpora to investigate the use of reporting verbs in academic discourse in terms of nativeness and expertise. Their findings demonstrated that native and non-native expert writers had similar reporting practices, while non-native novice writers differed significantly from others.

Finally, we turn to research on the use of reporting verbs by researchers at different academic levels. Bloch (2010) analyzed the application of concordancing to develop materials for teaching about the functions of reporting verbs between learners and published authors. In particular, the study identified the rhetorical criteria governing the choice of reporting verbs, such as the strength of commitment and the distinction between the writer’s and the cited author’s voice. Furthermore, authentic examples were presented to illustrate rhetorical subtleties, such as the use of state or claim to set up negative evaluations, which contrasts with the more neutral usage often found in student writing. In a related study, Parkinson (2013a) conducted a comparative analysis of reporting verbs in “that” complement clauses, finding that while ESL students demonstrated a fundamental awareness of academic conventions, their range of reporting verbs was narrow and they frequently relied on non-academic sources in their citations. Building on this finding, Friginal (2013) conducted research on the use of reporting verbs across student and professional corpora. Interestingly, the results demonstrated that the total frequency distribution of reporting verbs was nearly the same in the two corpora, while students showed a tendency to overuse certain reporting verbs. Further, Swales (2014) studied academic writing in biology by undergraduates and graduates at different grade levels. He found that all groups of students tended to use non-factive reporting verbs, and factives were less commonly used overall, though genre and discipline led to some variations.

As shown above, though many scholars have examined reporting verbs in academic discourse from diverse perspectives, only Charles (2006b) and Kwon et al. (2018) have examined how different factors within reported clauses influence the use of reporting verbs. However, Charles (2006b) focused solely on self-sourced reporting while overlooking other-sourced reporting. Additionally, Kwon et al. (2018) limited their analysis to students’ academic articles without comparing students’ usage with that of expert writers. In summary, reporting verbs are predominantly utilized within reporting clauses, and this represents a complex area of study influenced by various variables (Marti et al., 2019). Therefore, it is essential to scrutinize the impact of the interplay among multiple elements within reporting clauses on the usage of reporting verbs by L2 novice writers. However, very little is known about how these challenges manifest when comparing L2 novice writers, such as MA students, with experienced scholars publishing in international journals. Such a comparison is essential because it helps reveal the extent to which L2 novice writers align with expert rhetorical norms, and where they diverge in ways that may limit their participation in academic discourse. Furthermore, both self-sourced and other-sourced reporting verbs necessitate systematic investigation. Consequently, there remain gaps in research within this relevant field.

2.3. Overview of the Present Study

Based on Charles’ (2006b) subject framework as well as Kwon et al.’s (2018) semantic and source frameworks, the present study explored the semantic categories of reporting verbs, the source types and the subject types of reporting clauses in Chinese MA theses regarding linguistics and international linguistics journal articles. It also further analyzed the interrelationships among these elements to provide a more comprehensive understanding of reporting verbs. Specifically, the study centered on answering the following research questions (RQs):

RQ 1: What are the similarities and differences in the distribution and frequency of the semantic categories of reporting verbs, the source type and the subject types of reporting clauses between Chinese MA theses in linguistics and international linguistics journal articles?

RQ 2: What are the interrelationships among the semantic groups of reporting verbs, the source types and the subject types of reporting clauses in Chinese MA theses regarding linguistics and international linguistics journal articles?

3. Methods

3.1. Analytical Framework

Kwon et al.’s (2018) semantic and source frameworks and Charles’ (2006b) subject framework jointly constitute the analytical framework of this study (see

Table 1. Analytical Frameworks of this Study.

Categories		Examples
Semantic types	ARGUE verb	At last, it is often said that … (CMTL 6)
	SHOW verb	The study by Révész et al. (2023) showed that … (ILJA 23)
	FIND verb	… we found that both low and high-anxiety learners … (CMTL 7)
	THINK verb	While it might be reasonable to assume that learners of English … (ILJA 31)
Source types	Self as source	I think that includes your choice of words … (ILJA 33)
	Textual reference	… indicating that the scale is suitable … (Zhou, 2020). (CMTL 7)
	Uncited generalization	Research suggests that … (CMTL 17)
Subject types	Human	The researchers found that … (CMTL 20)
	Non-human	Findings from Fones’ (2019) study indicated that … (ILJA 29)
	It	Thus, it was suggested that … (ILJA 35)

). These frameworks are highly influential in research on reporting verbs and provide a well-established foundation for examining the semantic categories of reporting verbs as well as the source types and the subject types of reporting clauses.

With respect to the semantic classification of reporting verbs, the framework developed by Charles (2006b) and later refined by Kwon et al. (2018) was adopted in this study. Compared with earlier semantic classification schemes such as those proposed by Thompson and Ye (1991) and by Thomas and Hawes (1994), which have been criticized for overlapping subclasses or limited disciplinary scope, the semantic framework proposed by Charles (2006b) and later refined by Kwon et al. (2018) offers clearly defined and mutually exclusive categories. This clarity makes it particularly suitable for corpus-based analyses and cross-genre comparisons, and it was therefore selected for the present study. Based on this semantic framework, reporting verbs are divided into four groups, as shown in

Table 2. Classification of Semantic Categories.

Category	Examples
ARGUE verbs	Argue, suggest, assert, predict, write, explain, conclude, mention, admit, observe, accept, claim, imply, complain, point out, say, add, hypothesize, insist, maintain, propose, remark, reply, speculate, stress, contend, state, report, postulate, acknowledge, posit, talk about
SHOW verbs	Show, illustrate, indicate, demonstrate, confirm, mean, reveal
FIND verbs	Find, realize, observe, discover, establish, infer, recognize, identify, note
THINK verbs	Think, hold, assume, feel, hope, know

. All verbs in the corpora were coded strictly in accordance with these established verb lists and semantic categories.

The four semantic categories are defined as follows:

• ARGUE verbs relate to the act of writing and various other communication modalities. They denote the specific function that the communication serves (e.g., argue, suggest, propose).
• SHOW verbs relate to the depiction of a person’s expression or a particular situation (e.g., show, demonstrate, reveal).
• FIND verbs relate to the process of acquiring knowledge or arriving at a judgment about something (e.g., find, discover, observe).
• THINK verbs relate to the process of mental activity (e.g., think, assume, feel).

As for the subject types, the grammatical subject types can be distinguished into a noun group with human reference, a noun group with non-human reference and an introductory it followed by passive voice (Charles, 2006b). Pan and Lv (2025) argued that when writers employ human subjects, they highlight their own involvement and assume direct accountability for the claims made. They further noted that opting for non-human subjects allows writers to partially distance themselves, transferring responsibility to impersonal entities. Finally, they explained that utilizing it subjects enables writers to conceal their presence, assigning responsibility to the abstract it. Additionally, it subjects imply that authors are capable of eliminating personal biases in their writing and establishing a sense of authorial distance between themselves and the reporting clauses (Hyland & Jiang, 2018).

According to Kwon et al. (2018), reporting sources can be classified into three types: textual reference, self as source, and uncited generalization. Textual reference refers to the act of attributing information to specific textual sources; self as source refers to the expression of the writer’s own actions or opinions; and uncited generalization refers to the act of presenting general information or positions from other sources without providing corroboration.

3.2. Corpus

A linguistic corpus is a comprehensive, systematically organized, and linguistically valid collection of language data presented in electronic format (Omarova et al., 2025). In this study, two corpora were established: Chinese MA Theses in Linguistics (CMTL) and International Linguistics Journal Articles (ILJA). To ensure the validity and reliability of the results, the following criteria were adopted when selecting the research materials.

Firstly, the texts in CMTL were taken from “China Master’s Theses Full-text Database” in CNKI. When retrieving MA theses in linguistics by Chinese postgraduates from the CNKI database between 2020 and 2024, this study employed a stratified random sampling method to select articles as evenly as possible, thereby ensuring methodological rigor. Specifically, the theses in linguistics were first stratified by publication year, forming five annual cohorts (2020–2024). From each cohort, four theses were randomly selected, resulting in a total sample of 20 theses. Regarding author institutions, the authors came from comprehensive universities, normal universities, science and engineering universities, etc. (see Appendix A). This sampling framework balanced temporal distribution and institutional diversity while minimizing selection bias. Kennedy (1998) believed that a corpus with 100,000 to 500,000 words can reveal some language phenomena. In total, 386,129 words of Chinese MA theses in linguistics were collected (see Table 2).

Research articles are regarded as one of the most fundamental forms for presenting new knowledge within the academic realm (Berkenkotter & Huckin, 1995). The control corpus contains 45 articles from international linguistics journals. When selecting them, several steps were carried out. Initially, from the list of 2023 SSCI-indexed journals in linguistics, four journals with the highest impact factors were selected. They are Journal of Second Language Writing, System, the Modern Language Journal and Computer Assisted Language Learning. Academic articles published in authoritative journals have all undergone a rigorous peer-review and editorial review process (Connor, 2004). According to Carter (1990), authors who publish in such venues typically possess the academic and genre-specific knowledge required to participate as experts in their discourse communities. At the same time, it should be acknowledged that the polished quality of published texts may also partially reflect journal editorial and peer-review practices, which collectively help ensure adherence to the linguistic and rhetorical norms of expert academic writing. Then, to further guarantee standardization and nativeness, the authors of the selected articles have to meet one of the following two criteria: (1) As the criteria proposed by Rowley-Jolivet and Carter-Thomas (2014), the authors need to be associated with an institution in an English-speaking nation. (2) According to the criteria by Marti et al. (2019), the authors are required to be native to or educated in an English-speaking nation. For each journal, the study employed a stratified random sampling method to select 2–3 articles that met the criteria at each time point, with each year from 2020 to 2024 considered as a separate time point. About ten articles were selected randomly from each journal to avoid the influence of the stylistic features of academic articles in a certain journal (see Appendix B). In general, the total word count extracted from each journal is similar (see Appendix C). A total of 386,003 words from international linguistics journal articles were collected (see

Table 3. Composition of the Two Corpora.

Corpus	CMTL	ILJA
Number of texts	20	45
Total corpus size (tokens)	386,129	386,003

).

3.3. Research Procedures

First, we converted the extracted texts into a text document. Unnecessary elements such as charts, formulas, acknowledgments, and references were removed manually. Second, the indexing function in AntConc 4.0.4 was used to retrieve sentences containing reporting verbs. We conducted Keyword-in-Context (KWIC) searches for “v + that” clauses. This study focused on the single “v + that” structure, retaining the word “that” in all retrieved sentences to conduct a rigorous, in-depth comparison of how both groups of authors use this structure. Given that some occurrences exhibit multiple parts of speech (for instance, “claim” can function as both a verb and a noun), we also manually ensured that each occurrence of the word was used as a verb, verified through contextual analysis (Uba, 2020). The semantic distribution of different verb groups in the two corpora was analyzed according to the criterion proposed by Kwon et al. (2018). In practice, reporting verbs are realized through reporting clauses. Thus, the subject types and the source types of reporting clauses were also examined. Finally, the study explored the interrelationships between the source types and the subject types of reporting clauses.

To minimize subjectivity and ensure coding consistency, a transparent and operational coding protocol was established prior to the analysis. While grounded in the fundamental conceptual definitions provided by Charles (2006b) and Kwon et al. (2018), we further operationalized the criteria for coding source types to ensure analytical precision. Specifically, distinct lexical and syntactic indicators were used to distinguish the three categories. First, regarding Textual Reference, the primary criterion was whether the item denoted empirical evidence external to the author’s cognition. Accordingly, items such as “results”, “findings”, and “data” were classified as textual references. Similarly, “tables” and “figures” were treated as independent textual entities presenting data and were therefore coded into this category. Second, the category of Self as Source was restricted to instances where the writer explicitly functions as the cognitive source. This included clauses whose grammatical subjects directly encode the writer’s reasoning, such as “we argue that” or “we propose that”. Furthermore, nouns such as “analysis”, “model” (i.e., those developed or applied in this study), and “approach” were coded as self as source specifically when they represented the author’s analytical or interpretive acts. Finally, regarding Uncited Generalization, this category was restricted to reporting clauses that introduce shared disciplinary knowledge or generalized claims without explicit external attribution.

Based on the criteria established above, the retrieved “v + that” clauses were processed using UAM Corpus Tools 3.3x to systematically code the group categories. The coding process was primarily conducted by the first author, adhering rigorously to the pre-established operational protocol. To mitigate subjectivity and address ambiguous cases, a systematic expert consultation mechanism was implemented. Specifically, instances that fell on the boundaries of the defined categories were flagged during the initial coding. These flagged cases were then subjected to critical review with the corresponding author. During these consultations, the linguistic context of each ambiguous item was scrutinized against the coding criteria until a consensus was reached. To further ensure the reliability and stability of the coding, the first author re-coded a random sample of approximately 15% of the data after the initial coding was completed. The percentage of agreement between the two rounds of coding was calculated at 96%, indicating a high level of internal consistency in the application of the coding system. Finally, raw frequencies of the reporting verbs were calculated and transformed into standardized forms (per 100,000 words) to facilitate a feasible comparison between the two groups of authors.

4. Results and Discussion

4.1. Distribution and Frequency of Reporting Verbs and Clause Elements

4.1.1. Semantic Categories

Table 4. Distribution of Semantic Types.

	CMTL		ILJA
	Frequency	Percentage	Frequency	Percentage
ARGUE verbs	644 (166.82) 1	32.3	606 (157.00)	42.1
SHOW verbs	790 (204.56)	39.6	472 (122.28)	32.8
FIND verbs	400 (103.60)	20.1	299 (77.46)	20.8
THINK verbs	161 (41.70)	8.1	61 (15.80)	4.2
Total	1995 (516.77)	100	1438 (372.54)	100

¹ The number in (.) denotes the normalized frequencies.

shows that reporting verbs occur with a higher frequency in CMTL than in ILJA. In CMTL, the total frequency is 516.77 per 100,000 words, while in ILJA it is 372.54 per 100,000 words. According to Charles (2006b) and Kwon et al.’s (2018) classification criterion, reporting verbs can be semantically divided into four groups: ARGUE verbs, SHOW verbs, FIND verbs, and THINK verbs. The semantic frequency distribution of the four verb groups in the two corpora is shown in Table 4. In CMTL, SHOW verbs are the most frequent reporting verbs (204.56), followed by ARGUE verbs (166.82), and then FIND verbs (103.60). THINK verbs occur least often (41.70). In ILJA, ARGUE verbs occur most frequently (157.00), followed by SHOW verbs (122.28), FIND verbs (77.46), and THINK verbs (15.80).

Compared with authors of international journal articles, authors of Chinese MA theses use more reporting verbs in their academic writings, which parallels the findings in Marti et al. (2019). One possible reason for this difference is the distinct rhetorical expectations associated with the two genres. MA theses function as an apprentice academic genre in which the literature review occupies a central position, and authors of MA theses are encouraged to demonstrate familiarity with previous studies through explicit and frequent reporting. Reporting verbs therefore serve as an important resource for establishing research territory and showing engagement with existing scholarship. Another possible explanation relates to the authors’ developmental stages. Authors of journal articles draw on a wider range of evidential and attribution strategies that extend beyond reporting verbs. Because they can integrate prior work in more varied and sometimes less explicit ways, their overall frequency of reporting verbs may be lower even though their engagement with the literature remains strong.

SHOW verbs are related to denoting a situation (Charles, 2006b). They are the most frequent reporting verbs in CMTL (204.56) and the second most frequent ones in ILJA (157.00). Since a great number of factive verbs belong to SHOW verbs, we can observe that authors in CMTL focus more on presenting facts in their academic writing. This finding also indicates that they place significant emphasis on assertive information and put forward more compelling claims, which are supported by more abundant factual evidence and exhibit stronger persuasive effects (Pan & Lv, 2025). SHOW verbs typically refer to research activities rather than textual discourse, serving as linguistic resources for presenting empirical observations, methodological outcomes, or analytical results (Charles, 2006b), as in excerpt (1). Their relatively high frequency in CMTL suggests that authors of MA theses rely more on concrete factual statements in argumentation when using the “v + that” structure, whereas authors of journal articles prioritize the interpretation of results and the construction of discursive arguments when employing this structure.

(1)	The results of one-way ANOVA indicates [sic] that … (CMTL 15)

ARGUE verbs are associated with writing and other communication modalities, signifying the function that communication is fulfilling (Charles, 2006b). They illustrate the text-based and discursive features of academic writing (Marti et al., 2019). ARGUE verbs are the most frequent reporting verbs in ILJA (157.00), but they are the second most frequent reporting verbs in CMTL (166.82). This frequency distribution implies that authors in ILJA are more inclined to put forward tentative claims rather than factual and confirmatory ones (Parkinson, 2013b; Swales, 1990) when using the “v + that” structure. With ARGUE verbs, authors can demonstrate their in-depth thinking about an issue by citing controversial ideas of different scholars and summarizing them, which shows the authors’ critical thinking. Additionally, the functions of ARGUE verbs reflect a clash of different viewpoints and the proposal of new ideas, as in excerpt (2). This verb group occurs with a higher frequency in ILJA than in CMTL. This disparity is mainly because scholars publishing in international journals are expected to construct rigorous arguments and promote knowledge transformation, rather than merely displaying knowledge. ARGUE verbs also help authors of journal articles establish their research value and status.

(2)	Coffin and Donohue (2014) argue that … Arnold and Ducate (2006) pointed out that … (ILJA 3)

FIND verbs are linked to coming to know something and the majority of FIND verbs are used in the extended description or discussion of cited writers’ work (Charles, 2006b). Different from ARGUE verbs, which focus on presenting the writers’ viewpoints, FIND verbs tend to present the results and findings of research (Charles, 2006b). In both corpora, the frequency of FIND verbs is lower than that of ARGUE verbs and SHOW verbs. From the analysis of CMTL and ILJA, we observe that FIND verbs effectively capture the scientific nature of the research process and the objectivity of the results. This verb group also highlights that the reported content is an objective outcome derived from the research methodology, thereby reinforcing the empirical basis of the cited work, as in excerpt 3. Thus, FIND verbs enhance the credibility of cited content in academic writing. Furthermore, they are frequently employed in passive constructions, as seen in common bundles like “it could be found that” and “it should be noted that” (excerpt 4).

(3)	After eight weeks of teaching experiments, it was found that the argumentative writing group … (CMTL 18)
(4)	It should be noted that … (ILJA 34)

THINK verbs denote cognitive processes such as believing, knowing, or understanding (Charles, 2006b). On this basis, THINK verbs are also considered as verbs that “initiate negotiated claims” (Marti et al., 2019). In both corpora, THINK verbs have the lowest frequency of occurrence, accounting for only 8.1% in CMTL (41.70) and 4.2% in ILJA (15.80) of the total occurrence of reporting verbs. The low frequency of this verb group in both corpora is likely due to its subjective nature of expression. Academic research emphasizes facts, data and proven theories, so there is a need to reduce the influence of personal subjectivity. However, THINK verbs often carry the writers’ personal thoughts and speculations, which do not conform to the objectivity and impartiality of academic research findings. It may also lead readers to doubt the scientific nature of the reported content, as in excerpts (5) and (6). Therefore, both groups of authors seldom use THINK verbs to report the main ideas of the texts.

(5)	He thought that … (CMTL 1)
(6)	It was assumed that … (ILJA 3)

4.1.2. Source Types

The choice of source, along with the selection of reporting verbs, helps in building the writers’ stance (Thompson & Ye, 1991; Hunston et al., 1993; Hunston, 1995). According to the results in

Table 5. Distribution of Source Types.

	Textual Reference		Self as Source		Uncited Generalization
	CMTL	ILJA	CMTL	ILJA	CMTL	ILJA
ARGUE verbs	590 (152.80) 1	555 (143.78)	26 (6.73)	34 (8.81)	28 (7.25)	17 (4.40)
SHOW verbs	759 (196.57)	449 (116.32)	18 (4.66)	20 (5.18)	13 (3.37)	3 (0.78)
FIND verbs	380 (98.41)	270 (69.95)	11 (2.85)	28 (7.25)	9 (2.33)	1 (0.26)
THINK verbs	109 (28.23)	43 (11.14)	12 (3.11)	15 (3.89)	40 (10.36)	3 (0.78)
Total	1838 (476.00)	1317 (341.20)	67 (17.35)	97 (25.13)	90 (23.31)	24 (6.22)

¹ The number in (.) denotes the normalized frequencies.

, textual reference is the most prevalent source type (476.00), followed by uncited generalization (23.31), while self as source is the least utilized in CMTL (17.35). In ILJA, textual reference is likewise the most frequent (341.20), but self as source rises to the second position (25.13), with uncited generalization being the least used (6.22). Interestingly, this distributional sequence in ILJA formally aligns with the results reported by Kwon et al. (2018) in their student corpus. However, this superficial alignment stems primarily from differences in classification frameworks. Kwon et al. (2018), drawing on Charles (2006b), classified non-human subjects such as “Table 6” and “Figure 7” as self as source, thereby inflating the frequency of this category. In contrast, the present study adopted a more rigorous classification approach, categorizing such references as textual references. Consequently, cross-study comparisons must be approached with caution regarding their methodological foundations. Significantly, under the unified framework of this study, the frequency of self-sourced reporting by authors in ILJA remains higher than that of authors in CMTL. This finding suggests that authors of journal articles are more inclined to establish and highlight their academic authority and authorial voice, reflecting a critical rhetorical shift from novice to expert competence.

Textual reference is the most prevalent reporting type in both CMTL (474.46) and ILJA (341.20). This reporting type attributes information to its corresponding textual origins and identifies what we were originally expected to find (Kwon et al., 2018). As a reporting strategy, textual reference helps authors construct objectivity and align their arguments with established scholarly positions. It emphasizes that the cited part comes from a source other than the author of the article. Consequently, the source other than the author of the article should be responsible for the authenticity of the quoted statements. As a result of this reliance on external sources, by providing external validation, textual reference is a crucial means for authors to support their own arguments and position themselves within the academic community. Specifically, textual reference in CMTL aligns most closely with SHOW verbs (196.57), whereas in ILJA, it predominantly collocates with ARGUE verbs (143.78). This divergence reflects distinct rhetorical strategies. The prevalence of SHOW verbs in textual reference in CMTL suggests that authors in CMTL prioritize evidential reliability, treating cited sources as established facts to validate their knowledge base, as shown in excerpt (7). In contrast, the preference for ARGUE verbs in textual reference indicates that authors in ILJA view reporting as discursive claims, employing non-factive verbs to engage in dialogic interaction with previous scholarship rather than merely reporting results, as in excerpt (8).

(7)	Previous research has indicated that the English language consists of a large portion of multi-word units … (Skehan, 1998; Boers, Eyckmans, Kappel, Stengers, Demecheleer, 2006).
(8)	This is also the view of Taylor (2014): Despite arguing that beginner L2 learners need … (ILJA 37)

In contrast, “self as source” is the second most frequent reporting source in ILJA (17.35), while third in CMTL (25.13). It means referring to oneself as the source of action or information (Kwon et al., 2018). Additionally, it indicates that the content of the quoted parts comes from authors, emphasizing their subjectivity and presenting them as researchers. From a rhetorical perspective, the self as source is a strategy for demonstrating authorial identity and epistemic authority. Its higher frequency in ILJA signals that expert authors are more willing to take ownership of their claims and position themselves as contributors to disciplinary knowledge. Compared with authors of MA theses, authors of journal articles employ the self as a source not merely to describe their actions, but to construct a confident and authoritative scholarly voice, as shown in excerpt (9). Additionally, ARGUE verbs are most closely associated with the self as source in both corpora. This verb group collocates with 43.8% of self-referrals in CMTL and 35.1% in ILJA, as in excerpt (10).

(9)	We argue that these results suggest the need for increased pedagogical attention to these less commonly taught PM types … (ILJA 32)
(10)	Therefore, it can be concluded that the writing tasks that occupy more attention resources … (CMTL 15)

Uncited generalization refers to assertions drawn upon general knowledge, interpretations, or beliefs without providing a specific source (Kwon et al., 2018). As a reporting strategy, uncited generalization draws upon presumed shared disciplinary knowledge and signals the writer’s expectation that readers will recognize and accept the claim without attribution. It is employed not only to cite the viewpoints of other scholars, but also to introduce widely accepted perspectives or facts. In both corpora, this type of reporting is the least frequently occurring reporting form, with a normed frequency of 23.31 in CMTL and 6.22 in ILJA. The reason why this reporting source appears the least frequently is that it does not clearly indicate the source of the viewpoints and results. Moreover, the content is mostly the subjective judgment of writers. Although the content cannot be directly judged as wrong, it is difficult to trace the source of the viewpoints and evaluate their reliability. This is also not conducive to writers establishing their own academic stance and does not conform to the rigorous and scientific nature of academic writing, as in excerpts (11) and (12). In addition to these functional considerations, the infrequent use of uncited generalizations may also be influenced by the linguistic forms in which such statements typically occur. Many uncited generalizations appear in relatively fixed structures that are more restricted in their distribution in academic prose (e.g., fixed passive phrases). Their low structural variability and limited functional range may further contribute to their overall low frequency in both corpora.

The low frequency of uncited generalization in both corpora suggests that both groups of authors are cautious about presenting shared disciplinary claims without external justification. The even lower use in ILJA further indicates that authors of journal articles adhere more strongly to these norms, preferring to ground arguments in a traceable source, rather than relying on implicit disciplinary consensus.

(11)	It is widely acknowledged that females perform much better than males on whole in the language learning. [sic] (CMTL 2)
(12)	It is widely accepted that teachers’ beliefs and practices are situated in their own unique teaching contexts … (ILJA 33)

4.1.3. Subject Types

The grammatical subjects of clauses could be divided into three types: human, non-human and it subjects (Charles, 2006b), which bear responsibility for conveying stance towards the proposition expressed within the clauses (Man & Huat, 2019). Employing non-human nouns as subjects enables the writers to transfer responsibility for the proposition in the that-clause to the entity functioning as subject (Charles, 2006b). Additionally, utilizing non-human nouns as the subjects of reporting clauses appears to suggest that the propositions are grounded in external evidence, as opposed to being based on the writers’ personal or subjective opinions (Pan & Lv, 2025). This approach can enhance the objectivity and credibility of the propositions. In both corpora, reporting verbs occur in non-human subject clauses most frequently (see

Table 6. Distribution of Subject Types.

	Human		Non-Human		it
	CMTL	ILJA	CMTL	ILJA	CMTL	ILJA
ARGUE verbs	329 (85.20) 1	346 (89.64)	214 (55.42)	244 (63.21)	79 (20.46)	16 (4.15)
SHOW verbs	29 (7.51)	39 (10.10)	747 (193.47)	432 (111.92)	3 (0.78)	1 (0.26)
FIND verbs	202 (52.31)	224 (58.03)	66 (17.09)	58 (15.03)	124 (32.11)	17 (4.40)
THINK verbs	99 (25.64)	50 (12.95)	18 (4.66)	6 (1.55)	7 (1.81)	5 (1.30)
Total	716 (185.43)	659 (170.73)	1065 (275.82)	740 (191.71)	214 (55.42)	39 (10.10)

¹ The number in (.) denotes the normalized frequencies.

). This prevalence is because linguistic research involves experimental research, models and data. As shown in excerpts (13) and (14), non-human nouns are usually used as subjects to describe the contents of elements like “the above data”, “model analysis”. Thus, a substantial number of reporting verbs that collocate with non-human subjects are employed in linguistic texts. This linguistic strategy enables authors to eliminate personal bias in their writing and establish a sense of authorial distance between themselves and the reported clauses (Hyland & Jiang, 2018).

(13)	The descriptive data indicate [sic] that … (CMTL 7)
(14)	The results of the mixed effects model analysis showed that … (ILJA 23)

Reporting clauses with human subjects show an emphasis on the researchers’ identity and they are the second most frequent reporting clauses in both corpora (see Table 6). They also emphasize textual qualities such as formulating an argument and making room for discussion in the social sciences (Charles, 2003). Human subjects often appear as “the author”, “I”, or “we”, allowing readers of the articles to easily trace the source of the viewpoints. Here we take “I” and “we” as examples, which are used to express the authors’ own views or describe the specific information presented in the articles’ charts. When “I” serves as the subject, readers can clearly know that the views or achievements being quoted are those of the authors, as in excerpt (15). When authors use “we” as the subject of a clause to quote a view, the implicit meaning is that there is more than one person holding the view, as in excerpt (16). The use of “we” in this context includes several situations: there may be more than one writer; it may refer to the writers and the readers; or it could refer to the disciplines as a whole (Kuo, 1999). Therefore, “we” implies the meaning of a team. Compared to using “I” as the subject, using “we” together with a reporting verb can create a closer sense and increase credibility.

(15)	However, after preliminary discussions and a review of their backgrounds, I concluded that … (ILJA 22)
(16)	However, we found that there was … (CMTL 7)

Reporting clauses with it subjects appear the least frequently in both corpora and they hide the originators of views (see Table 6). It followed by the passive voice suggests the presence of an agent, which can usually be inferred from the context by readers (Charles, 2006b). As can be seen in excerpt (17), when writers use it subjects, the stance they convey is objective and can obscure their attitude towards the quoted views. It subjects allow writers to detach their personal opinions and establish a sense of authorial separation from the reporting clauses (Hyland & Jiang, 2018). In fact, we can regard human subjects as a sign of expressing more subjective viewpoints, while non-human and it subjects can be grouped together and regarded as signs of introducing more objective viewpoints. Since the total number of non-human and it subjects in both corpora exceeds that of human subjects, both groups of authors tend to use more objective forms to present their viewpoints when using “v + that” structure.

(17)	It can be concluded that the processing of derived words is inextricably linked to a significant amount of language input. (CMTL 19)

4.2. Interrelationships Between Source Types and Subject Types

When considering both the source types and the subject types of reporting clauses, we find that both groups of authors exhibit different tendencies in their usage, as shown in

Table 7. Combination of Source Types and Subject Types.

	CMTL		ILJA
	Frequency	Percentage	Frequency	Percentage
T-Human 1	644 (166.78) 2	35.0	581 (150.52)	44.1
T-Nonhuman	1013 (262.35)	55.1	699 (181.09)	53.1
T-it	181 (46.88)	9.8	37 (9.59)	2.8
Textual reference	1838 (476.00)	100	1317 (341.20)	100
S-Human	15 (3.88)	22.4	64 (16.58)	66.0
S-Nonhuman	28 (7.25)	41.8	32 (8.29)	33.0
S-it	24 (6.22)	35.8	1 (0.26)	1.0
Self as source	67 (17.35)	100	97 (25.13)	100
U-Human	57 (14.76)	63.3	14 (3.63)	58.3
U-Nonhuman	24 (6.22)	26.7	9 (2.33)	37.5
U-it	9 (2.33)	10.0	1 (0.26)	4.2
Uncited generalization	90 (23.31)	100%	24 (6.22)	100%
Total	1995 (516.66)		1438 (372.54)

¹ “T” represents textual reference, “S” represents self as source and “U” represents uncited generalization. ² The number in (.) denotes the normalized frequencies.

. When it comes to textual reference, authors in both CMTL and ILJA exhibit a tendency to employ non-human subjects most frequently. Human subjects rank second, and it subjects are used the least. In self as source, the most frequently occurring subject type in both corpora is non-human subjects. In CMTL, the second most frequent one is it subjects, and the least frequent is human subjects. In ILJA, the second most frequent is human subjects, while the least frequent one is it subjects. In uncited generalizations, both groups of authors most frequently use human subjects, followed by non-human subjects, with it subjects being the least used.

In textual reference, both groups of authors prefer to use non-human subjects. Textual reference implies attributing information to other texts, and the authenticity of the introduced viewpoints should be ensured by sources other than the authors themselves. When using non-human subject clauses to introduce textual reference, abstract nouns such as “the table”, “the results” and “previous research” are frequently employed as the subjects, as in excerpt (18). Additionally, other scholars’ names are also used as the subjects of the clauses to introduce viewpoints in textual reference to give credit to other scholars in both corpora, as in excerpt (19).

(18)	Zhang and Nicol (2022) explored the lexical alignment of bilinguals using image naming tasks, and the results show that … (CMTL 18)
(19)	Foster (1999, 2001) argues that … (CMTL 20)

In self as source, both groups of authors most commonly use non-human subjects to quote their own views or express their own viewpoints in an objective way, as in excerpt (20). The human subject “we” is employed to present authors’ viewpoints by both groups of authors, while “I” is mostly used by authors of international journal articles to express their own arguments. This difference reflects the tendency of authors in CMTL to adhere to genre conventions favoring objectivity and a collective voice, in contrast to the greater rhetorical flexibility and choice exhibited by authors in ILJA, which includes the use of explicit self-referrals. Moreover, authors of MA theses usually complete their theses under the guidance of their supervisors. Since the work is considered collaborative achievements, authors of Chinese MA theses avoid using “I” as the subject of reporting clauses. As can be seen in excerpt (21), when it serves as the subject to present self-referrals, it often functions merely as a dummy subject without actual semantic reference.

(20)	The regression method of “enter” shows that the combination model of TG and SSAR significantly predicted the syntactic complexity of oral output … (CMTL 18)
(21)	It should be noted that the three instructors assigned the corpus activities as graded in-class and homework assignments. (ILJA 5)

In uncited generalization, both groups of authors use this type of reporting to describe knowledge that is commonly known, and it does not clearly indicate the source. When using uncited generalization, both groups of authors often use a noun group with human reference as the subjects. Although grammatically human, these subjects function as generalized collective references, which are conventionally used to appeal to a shared disciplinary community rather than specific individuals. Common collocations include “the researchers found that”, “some scholars argued that”. While human subjects indicate that the views are owed to a source, the authors do not mention the exact ones, as in excerpt (22). In such structures, we cannot trace the source and evaluate the accuracy and authority of the views. Therefore, when authors need to explain a certain phenomenon or the meaning of a noun, it is advisable for them to use other source types other than uncited generalization, which are more credible, though uncited generalization is not definitely false.

(22)	The researchers found that … (CMTL 20)

5. Conclusions

Proficiency in academic writing is essential for both L2 learners and expert authors, as it is a core skill in scholarly communication (Swales, 2004; Hyland, 2016). The primary objective of this study was to reveal potential disparities in the use of reporting verbs between Chinese L2 learners and international expert authors in academic writing. Regarding overall frequency, authors of MA theses employed a higher density of reporting verbs, a tendency that parallels findings by Marti et al. (2019). Semantically, authors of MA theses favored SHOW verbs to establish factual validity, whereas authors of journal articles prioritized ARGUE verbs to construct discursive claims, a divergence aligning with Friginal (2013). In terms of the source types and subject types, textual reference and non-human subjects predominate in both groups, reflecting shared disciplinary norms. However, distinct strategies emerged in the interrelationships. For instance, authors of MA theses primarily linked textual references with SHOW verbs and obscured their own agency behind non-human entities. Conversely, authors of journal articles connect textual references with ARGUE verbs and actively employ human subjects to project an explicit authorial voice. These rhetorical choices highlight how academic genre conventions and stages of socialization shape scholarly discourse.

In this study, the cross-cultural disparities in the usage of reporting verbs carry significant pedagogical implications for EAP teachers and L2 learners. For EAP teachers, it is of vital importance to guide and improve the use of academic reporting among L2 writers. Specifically, even though there’s a wealth of corpus-based academic research, its actual application in teaching and learning has not been given enough attention. When delving into corpora, the examination of whether certain linguistic patterns are overused or underused yields valuable data. Teachers are in a position to craft data-driven writing instruction by utilizing genuine textual materials. For example, they can encourage students to conduct a comparative analysis of their own usage of reporting verbs and that of experts. This analytical process can reveal specific discrepancies in the student’s use of linguistic features compared to expert writing. Based on these discrepancies, teachers can pinpoint what students have grasped and what they still lack in the use of reporting verbs in academic writing. Then, teachers can offer highly targeted instructional support to students. Additionally, EAP teachers could offer targeted academic writing courses to provide writing guidance for students, increasing students’ understanding and familiarity with different reporting verbs, and helping them better use reporting verbs for self-referrals or reporting other scholars’ viewpoints in academic writing. They could also further support learners by providing word lists of reporting verbs to enhance lexical precision (Huang, 2018). Proper use of reporting can assist L2 learners in expressing academic viewpoints and constructing academic positions. Teachers should guide L2 learners to realize that academic writing is not a one-way talk but a conversation within the academic community (Pan & Lv, 2025), where reporting serves as a critical tool for engaging with prior research and constructing scholarly arguments. By instructing students on the proper usage of reporting verbs, EAP teachers can empower students with the linguistic and rhetorical skills necessary to navigate academic discourse effectively.

For L2 learners, the findings of this study indicate that they have generally mastered the use of reporting verbs in academic writing but still face some issues at a specific level. L2 learners need to avoid overuse of reporting verbs, as they tend to rely heavily on the same ones, resulting in limited lexical variety. This overreliance on a narrow range of verbs might be due to rote learning and the use of fixed patterns. M. Yang and Lin (2009) discussed the concept of cultural citation practices. They explained that students with a rote-learning cultural mindset may not be familiar with common academic citation practices. As a result, their writing may involve plagiarism. In view of this issue, L2 learners should replace reporting verbs flexibly according to the context. For example, they can try to vary their expressions instead of just using the most common reporting verbs to make their expressions more diverse. Additionally, they should also reduce the use of uncited generalizations, which are reports without clear sources. Since the specific source of the uncited generalization cannot be traced back, it is difficult to objectively judge the validity of the statements. Moreover, L2 learners tend to place too much trust in the authority of other researchers and ignore the significance of their own viewpoints. Compared with expert writers, L2 writers are less likely to use “I” or “the author” as the subject to present their own academic ideas or make self-sourced reporting. Therefore, they should develop an awareness of self-sourced reporting in academic writing and use such reporting more often, which helps them better build their academic status. Understanding linguistic features allows MA students to create explicit, cohesive, and impressive theses (Manan & Noor, 2014). By grasping when and how to report in their writing, students could better show their proficiency in combining other scholars’ opinions with their own viewpoints.

However, there are also some limitations in this study. Firstly, the current two corpora comprise approximately 760,000 words in total. While this sample size is adequate for preliminary analysis, future studies should aim for at least one million words to ensure more robust findings. Secondly, the reporting verbs analyzed in this study were limited to those identified by Charles (2006b) and Kwon et al. (2018), which do not cover all possible verbs. Future studies could include a wider range of reporting verbs for more comprehensive results. Thirdly, the study focused exclusively on linguistics. In future research, the investigation of reporting verbs across other disciplines could be explored. Additionally, this study only examined reporting formed by “v + that” clauses. Although this structure is common in academic writing, it represents only one of many reporting strategies. Thus, the findings should be interpreted within this structural scope, and future studies could incorporate additional reporting forms for a more comprehensive analysis. In conclusion, it is hoped that the insights provided in this study can promote the development of L2 writers’ academic writing abilities and offer some implications for L2 writers in their academic writing and EAP teachers in their academic writing instruction.