The Interplay between Disciplinary Conventions and Research Methodologies in the Introduction Section of Research Articles: Citation Functions and Evaluations

Introduction

Citation practices are an indispensable characteristic of academic writing and have always been of paramount importance in English for Academic Purposes (EAP) (Hyland, 1999; Samraj, 2013; Swales, 2004, 2014). Much research has been done to look into these practices. The effect of writers’ linguistic background and expertise on citation practices has been documented in Dontcheva-Navratilova (2016), Lillis, Hewings, Vladimirou, and Curry (2010), among others. Moreover, differences in citation practices as a result of disciplinary variations have also received adequate attention (Charles, 2006; Harwood, 2009; Swales, 2014; Thompson, 2001, 2005; Thompson & Tribble, 2001), where it was shown that citations are realized differently in hard sciences and soft sciences, for instance. Citation practices have also been investigated in different writing tasks (Lee, Hitchcock, & Casal, 2018; Petric´, 2007), with the result that these practices are likely influenced by the type of task assigned. Some variations in citation practices have also been linked to intercultural differences (Hu & Wang, 2014; Shooshtari & Jalilifar, 2010).

However, one factor which might have an impact on citation practices, but has remained under-researched, is the research methodology that authors adopt for their articles (Swygart-Hobaugh, 2004). Swygart-Hobaugh (2004) points out that methodological underpinnings might be responsible for different citation patterns in research traditions. She examined how social scientists make reference to qualitative and quantitative sources when writing articles in line with either of the two research methodologies. She concluded that “quantitative articles were more likely to cite journal articles than monographs, while qualitative articles were more likely to cite monographs than journals” (p. 191). Zhang (2007) also found similar results regarding citations used in quantitative and qualitative debates, stating that qualitative researchers refer to a greater percentage of monographic works, but quantitative academics exhibit a higher proportion of citations from journal articles. Much recently, as a complementary work to the current study, Arizavi and Choubsaz (2021) investigated the disciplinary and research methodological effects on the formal characteristics of citation practices. The findings point to significant effects as a result of both variables, while the effects of the latter were more noticeable. These studies, however, have not reflected on variations in citation practices in terms of functions and authorial evaluative stances toward the cited works, especially when considered at disciplinary and cross-research methodological levels.

It is worthwhile mentioning that research methodologies are informed by the ideological paradigms the researchers happen to adhere to (Morgan, 2007; Riazi, 2016). It is the paradigms that characterize and guide research methodologies at practical levels, called research methods, which encompass data collection procedure, data analysis, and interpretation, frames of reference, and situating one’s study within a wider scope of a discourse community (Hyland, 2001; Morgan, 2007; Riazi, 2016; Swales, 2004; Ghiara, 2019). Therefore, it can be extrapolated that many academic writing features, including citation practices, might vary across research methodologies and disciplines because of the inherent paradigmatic characterizations of the methodologies. This consideration distinguishes the current study from other similar ones since no study, to date, has reported on the possible variations in citation practices as a consequence of the interplay between disciplinary conventions and the choice of research methodologies in research articles (RAs). Although citation functions and the authorial evaluative stance toward citations have been studied separately by Coffin (2009), Jalilifar (2012), and Thompson (2001, 2005), this study addresses these aspects at the cross-disciplinary level (Theoretical Linguistics and Applied Linguistics), across research methodologies (quantitative, qualitative, and mixed-methods), and a combination of these two variables. RA introductions are chosen as the corpus of the study since they are recognized as the hotspot of authors’ interactive engagement with the cited sources (Jalilifar, 2012; Loan & Pramoolsook, 2015; Yeh, 2010), where they refer to the cited studies as (counter-)evidence, or (un)support to their own claims. Taking the aforementioned variables together, the following research questions are formulated to guide the study:

• Are there any significant differences in citation functions in theoretical and applied linguistics research article introduction sections across the research methodologies employed in the research articles based on the framework developed in this study?
• Are there any significant differences in the authors’ evaluative stances toward citations in theoretical and applied linguistics research article introduction sections across the research methodologies employed in the research articles based on Coffin’s (2009) framework?

Literature Review

The citation function shows an author’s purpose and attitude toward a cited work in a novel text. Dubois (1988) identified four citation functions: direct quotation, paraphrase, summary, and generalization. Thompson’s (2001, 2005) and Thompson and Tribble’s (2001) studies can be considered pioneering endeavors to examine the rhetorical functions of citation across disciplines. In fact, Thompson’s (2001) work followed Swales’s (1986) original dichotomy of citation types. Swales (1986) identified a citation as integral if it is part of the grammatical composition of a sentence and non-integral if it is enclosed in brackets and does not form a grammatical constituent in the sentence. Thompson (2001) introduced a number of broad functional categories based on this dichotomy; namely, “non-integral source, non-integral identifying, non-integral reference, integral verb-controlling, and integral naming” (as cited in Samraj, 2013, p. 303).

Later studies, however, offered functional taxonomies of citations for different disciplines and carried out comparative analyses (Harwood, 2009; Hu & Wang, 2014; Petrić, 2007). Petrić (2007) compared citation functions in two sets of low and high-rated theses and reported that citations in low-rated theses sought more descriptiveness rather than ‘evaluativeness’ of previous works, while citations in high-rated theses were more evaluative and less descriptive. This typology is nevertheless considered meagerly flexible (Samraj, 2013). In an interview-based study, Harwood (2009) offered details of citation functions in computer scientists’ and sociologists’ writings. He asked informants to assign functions to the citations they had in their texts and comment on them. Harwood concluded with evidence for inter- and intra-disciplinary variations, adding that the type of the paper (i.e., theoretical or empirical), the readership, and the publication venue lead to differences in citation practices. Results from studies employing these classifications indicated that apart from disciplinary discrepancies, authors’ proficiency levels and linguistic backgrounds are also major factors in the variations of citation practices (Chang, 2013; Hu & Wang, 2014; Jarkovská & Kučírková, 2020; Jomaa & Bidin, 2019; Loan & Pramoolsook, 2015; Shooshtari & Jalilifar, 2010).

Some studies (e.g., Harwood & Petrić, 2012; Jalilifar, 2012; Jomaa & Bidin, 2019; Lee et al., 2018; Mansourizadeh & Ahmad, 2011; Samraj, 2013) expressed dissatisfaction with previous categorical analyses of citation, and instead called for more attention to context-dependent rhetorical functions of citations in their localized contexts. Samraj (2013) notes that generalizing citation functions from one context to other seemingly similar contexts would definitely end up in an inadequate representation of the latter contexts. She explored the functions of citations in the discussion sections of master’s theses and RAs from biology using Thompson’s (2001, 2005) taxonomy and then expanded the model to include other context-sensitive functions in her study. Unlike previous studies, that assign only a single rhetorical function to citation in the discussion section (i.e. commenting on results), Samraj’s results showed that citations are used for a range of different rhetorical functions all over this section in both master’s theses and journal articles.

Analyzing the authorial evaluative stance toward citations is perhaps less researched (Coffin, 2009; Hyland, 1999; Jalilifar, 2012; Lee et al., 2018; Petric´, 2007; Shooshtari, Jalilifar, & Shahri, 2017; Thompson & Ye, 1991). Thompson and Ye (1991) examined how writers use reporting verbs in evaluating citations in their papers. The positive and negative connotations of the reporting verbs were studied, and it was found that negative evaluations were presented with greater meticulousness than positive evaluations. Hyland (1999) analyzed citation realizations in a large corpus based on Swales (1986) and examined the authorial stances toward the cited sources by looking into the manners that citations were incorporated in the articles, i.e., cited sources as quotations, summaries, and generalizations from several other studies. He concludes that disciplinary variations in citations manifest themselves both in forms and functions, where in some disciplines, authors use more quotations, while in other disciplines, summaries and generalizations are more common. Petric´ (2007) and Jalilifar (2012) found task-related (i.e., theses written by novices vs. articles written by experts), as well as disciplinary variations in the evaluative devices indicating stance toward citations. Collectively, the findings of these two studies reveal that article writers are more cautious when taking critical stances toward the cited contributions, and this is more evident in social sciences and humanities than in hard sciences.

Coffin looked into the authorial stance toward cited studies and discovered that non-expert authors chiefly assumed “a non-committal stance toward cited sources rather than taking a strong positive or negative position” (2009, p. 2). On the other hand, expert writers took a more committal stance but they exercised utmost caution in passing their negative evaluative judgments on the cited sources. Lee et al. (2018) also found L2 student writers largely embraced a non-committal stance by “acknowledging or distancing themselves from cited materials, suggesting that L2 students are inclined to show deference to the perceived authority of published sources” (2018, p. 9). Shooshtari et al. (2017) used Coffin’s framework to analyze a corpus of applied linguistics articles published in national and international journals to see if authors’ linguistic and cultural backgrounds affect their citation practices. They found cross-linguistic and cross-cultural differences in the way authors take a stance toward citations. They believe non-native authors usually take negative stances toward citations to show their criticality as they know this would be positively evaluated by the gatekeepers of the journals they would contribute to. To put the discussion of stance-taking in a nutshell, it is well-documented that writers of varied levels of expertise take a stance toward citations differently because of their unequal language proficiency, various linguistic background, and disciplinary obligations.

Having reviewed the studies on citation practices, it is evident that the effect of the interplay between disciplinary and research methodological variations on citation practices has gone unnoticed. Thus, this study intends to inform novice academic writers and other EAP practitioners in the field of linguistics about how citation practices in RAs might be influenced by the disciplines they affiliate with and the research methodologies they adopt for their research.

Theoretical Framework

Typology of Research Methodology

Research methodologies are informed by particular paradigmatic principles. For instance, research methods such as correlational, survey, or experimental methods can be subsumed under a quantitative research methodology, which affiliates with the positivist paradigm, while conversation analysis or case studies are subordinated under a qualitative research methodology, which associates with the constructivist paradigm (Ghiara, 2019; Morgan, 2007). In general, RA authors draw on three main research traditions. Operational definitions of the three research methodologies are as follows:

Quantitative research “covers all those studies which are informed by (post)positivism as their underlying worldview or paradigm and aim at explaining social and educational phenomena objectively” (Riazi, 2017, p. 259). Being theory-driven, quantitative research aims at generalizations from sample findings to a target population. It is descriptive, explanatory, and at times predictive in nature, and all of these are achieved through hypothesis testing, quantification, and statistical analysis (Richards, Ross, & Seedhouse, 2012). Qualitative research includes an array of methods linked to interpretivism or the social constructionism paradigm. Riazi (2017) posits qualitative research is built on:

Multiple subjective realities, recognizing the researcher’s and participants’ value systems in research, studying the social phenomena in their natural setting with no manipulation and control over the setting, describing the object of the study from the perspective of participants (emic perspective), using thick description, and using qualitative data and analysis to make inferences about the social phenomena. (p. 256)

Mixed-methods research is an approach combining quantitative and qualitative data and analyses in a single study to address research problems. It differs from the multi-method approach where methods from one research paradigm may be melded (Riazi, 2016). The reason for mixing methods in this methodology is to attain “triangulation, complementarity, initiation, development, or expansion” (Riazi, 2017, p. 193).

Functional Typology of Citations

Thompson’s (2001, 2005) typology of citation functions is chosen for the first round of functional analysis of the corpus. As stated in Section 2, it adds a number of functional sub-types to Swales’s (1986, 1990) tightly formal model. This model is advocated by Jalilifar (2012), who believes that the classification is “comprehensive and it takes account of all the citation types” (p. 29). Thompson divides citations into integral and non-integral citations, with three sub-categories under the former and four more under the latter (2001, pp. 95-96).

INTEGRAL CITATIONS

1. Verb Controlling: the citation acts as the agent that controls a verb, in active or passive voice.
2. Naming: the citation is a noun phrase or a part of a noun phrase.
3. Non-citation: there is a reference to another writer but the name is given without a year reference.

NON-INTEGRAL CITATION

1. Source: non-integral source citations attribute a proposition to another author.
2. Identification: this identifies an agent within the sentence it refers to. The information within the parentheses identifies the author of the study referred to.
3. Reference: this type of citation is usually signaled by the inclusion of the directive “see”.
4. Origin: Origin citations indicate the originator of a concept or a product, in this case, the creator of the Wordsmith Tools program.

However, it is possible to criticize Thompson’s (2001, 2005) model on the grounds that despite Thompson’s (2005) claim, it still leans more toward the formal classifications than the localized rhetorical functions of citations. This criticism has already been lodged by Samraj (2013), who mulls over the fact that the context where citations occur is a determining factor in assigning their functions. For this reason, an inductive discourse analysis (Barton, 2002) method was used to look for context-dependent functions of citations in the RA introduction sections which might not be identified by Thompson’s (2001, 2005) model. According to this method, texts are continually read and analyzed until patterns of discourse, and citation functions in our case, emerge based on recurrent features. Then, the emergent list of functions was brushed up as more texts were analyzed and functions became satisfactorily distinct from one another. Finally, the corpus was again examined with the settled classification. Samraj (2013) employed this method of discourse analysis and came up with a list of eight functions to classify citations in the discussion sections of articles; namely, Comparison of results, Interpretation of results, Explanation of results, Evaluation of study, Evaluation of field, Research recommendations, Applied recommendations, Background.

Authorial Evaluation of Citations

Authors may take a committal or non-committal stance toward others’ works in order to establish their own stance or standpoint (Hyland, 1999). One framework that has dealt with this strand of research is that of Thompson and Ye (1991), who analyzed and proposed categories for stances taken in the verb controlling the type of integral citations. According to this classification, the reporting verbs can be characterized as a) factive, b) counter-factive, and c) non-factive (Thompson & Ye, 1991). However, this framework has little to offer for non-integral citations and other citation forms than verb controlling clauses. Moreover, the terms factive, counter-factive, and non-factive do not reflect the authors’ stance since they are more associated with the truth-value of propositions rather than the authors’ positions.

Thus, in the quest for a comprehensive evaluative model, Coffin’s (2009) writer stance framework is used in this study. As Hu and Wang (2014) point out, this framework underscores interactive commitment “in which the writer is engaging retrospectively with previous authors and communicating prospectively to an audience” (p. 17). Drawing on the attribute subsystem of the Appraisal Theory (Martin & White, 2005), Coffin (2009) concentrates on the linguistic choices which writers make “to engage with and negotiate voices and viewpoints” (p. 169). This framework consists of three dimensions (i.e., writer stance, textual integration, and nature of source); however, as the focus of the third research question in this study is on the writer stance, only the first dimension is taken into account. Coffin (2009) defines the writer’s stance as a position taken toward “the words, observations, viewpoints, and theories that comprise the referenced source” (p. 170). Coffin’s (2009) framework consists of four stance types:

• Acknowledge conveys a neutral position where no evaluative comment is given;
• Distance indicates no accountability for the reliability of the information is offered;
• Endorse is meant to support or agree with a cited source;
• Contest reflects a direct refusal of the cited proposition.

Method

Corpus Selection

A multi-faceted research design is adopted in this study. A sample of 180 RAs from two close-knit disciplines of Theoretical Linguistics (90) and Applied Linguistics (90) was selected from six journals. Thirty RAs were purposively selected from each journal, 10 articles complying with each of the research methodologies in concern. It is important to note that theoretical RAs are by definition not empirical, while applied RAs are empirical and aim at resolving (Creswell, 2011) language-related real-world issues. In order to avoid the intervening effect of time on citation practices and other generic features (Berkenkotter, 2008), and to maintain the recency of research methodologies, much caution was exercised to restrict the selection of articles to those published in the last decade.

The field of linguistics was chosen because, like other fields of social sciences, a plentitude of studies can be found complying with the tripartite research methodologies. Distinctively, it is dissimilar to fields of study that lend themselves only to one particular research tradition, say purely quantitative or purely qualitative (Creswell, 2011; Morgan, 2007; Riazi, 2016). This field of inquiry is divided into theoretical and applied disciplines as separate disciplines. Theoretical Linguistics aims at examining the structure of languages including phonetics, phonology, morphology, syntax, semantics, pragmatics, language change, first language acquisition, language disorders, and comparative linguistics. Applied Linguistics explores language teaching in the educational setting, second language acquisition, and sociolinguistics.

The first step in corpus selection was selecting the right journals and RAs. As stated earlier, six journals were selected for the purpose of the study based on the description given in the aims and scope of each journal about being theory-oriented or applied in focus. Three journals (Journal of Linguistics, Journal of Pragmatics, and Lingua) were categorized as theoretically-oriented, primarily aiming at disseminating knowledge about language (i.e., its systems and mechanisms). Applied Linguistics journals were easier to choose since most of the top-ranking journals in the field of linguistics pertain to this category. Annual Review of Applied Linguistics, Applied Linguistics, and International Journal of Applied Linguistics were chosen to represent applied linguistics. Unlike theoretical linguistic journals, applied linguistic journals spearhead a problem-solving campaign. Determining the type of research methodology of the articles came either from the article writer or writers’ explicit verbatim in the articles or the current study authors’ inference based on the definitions and assumptions of each methodology.

Procedures

Having downloaded the articles, the researcher performed the cleaning process by excluding other sections of the articles. To minimize the effect of section length, article introductions containing between at least 500 and 2000 words at most, with an adequate number of citations, were chosen. The average length of an introduction section in the theoretical linguistics articles was 941 words, while it was 863 words in the applied linguistics introduction sections. Moreover, the average length of the introduction sections was calculated for each research tradition (879, 924, and 824 words in quantitative, qualitative, and mixed-methods, respectively). Based on arbitration and to avoid extreme outlier cases, articles including 10 citations at least were chosen; however, the upper limit was not determined. The article authors’ native language, singularity or multiplicity of the authorship, and geographical distributions of the articles were not treated as intervening factors.

Articles were coded as quantitative, qualitative, and mixed methods, all in-text citations in the main body of the articles were highlighted, and functions to every single instance of citation were assigned. Each citation was also given a stance value, i.e., Acknowledge, Distance, Endorse, and Contest. Citations in epigraphs and footnotes were not counted. For counting purposes, multiple non-integral citations enclosed within a pair of brackets were considered as separate instances although they might entertain one function. Like Samraj (2013), citation functions were identified by considering the sentence containing the citations and the surrounding context, i.e., neighboring sentences, in case the functions could not be easily inferred. There were instances where a citation performed multiple functions. The main function was only taken into account. For instance, “there is another trend, pursued by the authors of this article, which focuses on the evaluation of educators, students, and learning per se in the domain of learning (Oxford et al., 1998; Jin & Cortazzi, 2008; Nikitina & Furuoka, 2008a,b; Wan et al., 2011)” (Applied Linguistics, Journal of Applied Linguistics, Mixed-methods, Article 6), the non-integral citations apparently provide support for a claim the authors have put forth concerning the new trend established in their study, but the same time they provide an overview of the studies that have already identified the appearance of this new trend in the literature. However, since they are situated in indicating a Niche Move (Swales, 2004) of the introduction, they are perceived as support rather than overview, based on the citation function taxonomy developed in this study. Yet another challenge was similar repeated citations within consecutive clauses or sentences. Unlike Samraj’s (2013) analysis, which gave these instances one single function, in this study they were taken to be independent instances.

The corpus was manually searched for recognized citation indicators like a date in brackets, Latinate references (like ibid.), authors’ names, personal subjective pronouns, terms like ‘the researchers’ or ‘the authors’, organizations, and institution names. These were all screened out against the bibliographies. Following Hyland’s (1999) procedure, self-citations, proper nouns indicating schools of thought (unless they refer to a particular person, for instance, ‘Chomskyan proponents like White points out …’), or the like that did not incorporate former content into the article were omitted.

Coming to the reliability, ten percent of the corpus was independently analyzed by the researcher and a colleague of his, who was explicitly informed of the objectives, frameworks, and coding schemes. For the aspect of the study that Coffin’s (2009) framework was employed, a deductive qualitative method was adopted because each citation instance was assigned an evaluative value already present in the literature. Perfect inter-coder reliability (.83) was obtained applying Cohen’s Kappa Coefficient. Moreover, an inductive qualitative method was used to come up with emerging patterns and categories of functions and simultaneously analyze the data based on the developed taxonomy. For this phase, the researcher and his colleague went through recursive rounds of discussion to add, remove, subsume, coalesce, and label the emerging categories, as well as to assign functions to the citations. In this method, inter-coder agreement (a softer form of inter-coder reliability) was more preferable (Campbell, Quincy, Osserman, & Pedersen, 2013), the result of which was 81% in the first round and 88% in the second round.

Results and Discussion

A total of 5201 citations were found in the corpus, of which 818 (15.7%) were integral and 4383 (84.3%) were non-integral. There were 2385 (45.9%) citations in the Theoretical Linguistics RA introductions, with an average of 26.8 per introduction, while 2816 (54.1%) citations were found in the Applied Linguistics RA introductions, with an average of 32.1 per introduction. Across research methodologies, 1653 (31.8%) citations were found in the quantitative RA introductions, with an average of 27.5 per introduction, 1903 (36.6%) in the qualitative corpus, with an average of 31.7 per introduction, and 1645 (31.6%) citations in the mixed-methods article introductions, with an average of 29 per introduction.

Inductive functional analysis

It is unfair to deny Thompson’s (2001, 2005) efforts of providing a functional taxonomy for citations in academic research. However, it seems that Swales’s (1986, 1990) original formal classification had such an encroaching impact on Thompson’s functional categories that has resulted in dysfunctioning categories that can hardly account for the functions of certain instances of citation. In fact, Thompson’s functional categories are in the service of formal categories and cannot be regarded autonomously functional. Some functions are either ignored or subsumed under a general category. Another shortcoming of Thompson’s framework is that labels are imposed on categories despite the fact that an integral citation might have the same function as a non-integral citation; there is a pejorative rationale why the category Verb controlling should be different from Source or Attribution functionally. Given all these shortcomings in previous models, in this study, six categories are proposed to represent the rhetorical functions of citations in their specific context, regardless of their linguistic realization and position in the clauses.

1. Prominence: this includes citations that indicate the centrality, importance of, and need for the issue in concern. This function corresponds to Swales’s (2004) communicative Moves 1 (Establishing a territory) and Move 2 (Indicating a niche) to some extent. It can be seen as a strategy adding more credibility to the urgent need for the study, which otherwise would be considered less imperative.

Example 1: “There has been a call for more research into the moderating effects that individual factors have on how learners respond to and use the written CF they receive (Ellis, 2008)” (Applied Linguistics, Journal of Applied Linguistics, Quantitative, Article 2).

1. Definition: in this function, citations serve to lay out disciplinary terms and concepts that are essential to orientate the readership and limit the scope of the study. Sometimes, authors need to provide background information through situating their research in already established terms and definitions.

Example 2: “The term ‘initiator’ goes back to Quirk et al. (1985: 444), where it is a label for a functional class of adverbs and interjections that commonly start utterances in spoken discourse” (Theoretical Linguistics, Lingua, Mixed-methods, Article 6).

1. Overview: citations in this category provide an outline of the research trends, history, and state-of-the-art studies in relation to the issue they desire to unpack. This is often realized as a list of studies that have followed a similar line of inquiry.

Example 3: “One important trend in this discourse investigates the role of metaphor in enhancing learning (Low, 1988; Lazar, 1996; Deignan et al., 1997; Boers, 2000; Littlemore & Low, 2006)” (Applied Linguistics, Journal of Applied Linguistics, Mixed-methods, Article 5).

1. Review: this refers to citations that indicate particular key studies that help shape the foundation of the article. References and attributions to aims, methodologies, approaches, and major findings are other manifestations of this category.

Example 4: “Alcón-Soler (2015) also dealt with pragmatic changes during study abroad. That study explored to what extent pragmatic instruction and length of study abroad influence learners’ ability to mitigate requests in email communication” (Applied Linguistics, Annual Review of Applied Linguistics, Quantitative, Article 1).

1. View: citations in this category are authors’ views and claims on the issue, no matter in favor or against it. They simply function to explicate different aspects of the issue more intelligibly.

Example 5: “Independently of Kuno (1973), Teramura (1982: 106–108) discusses another group of verbs of inherently directed motion in Japanese that mark their object with either accusative case -o or the source marker -kara from” (Theoretical Linguistics, Journal of Linguistics, Quantitative, Article 1).

1. Support/rejection: these citations act as back-up to the claims of the author’s article. In other words, the cited source(s) lend support to the author’s claim or reject claims and findings of previous studies in favor of the claim in concern.

Example 6: “Besides students, learning advisors also benefit from conducting ICs because they can better understand learners’ writing difficulties through close examinations of written work (Huijser, Kimmins, & Galligan, 2008)” (Applied Linguistics, International Journal of Applied Linguistics, Qualitative, Article 3).

Table 1 demonstrates variations in the citation functions based on the proposed taxonomy at the disciplinary level. There are significant differences between the two disciplines in all the functions (critical value: 3.841 and significance level: .05). The highest percentage in both disciplines belongs to the function of Overview with 1165 (48.9%) instances in the Theoretical Linguistics and 822 (29.2%) instances in the Applied Linguistics corpus. Comparing other functions, 129 (5.4%) citations signify Prominence in the theoretical linguistics corpus, whereas 263 (9.3%) citations indicate this function in Applied Linguistics article introductions. This difference can be ascribed to the authors’ tendency to rely more on their own rhetoric to voice the grave need for carrying out their study rather than using multiple citations to indicate so. Concerning Definition, Theoretical Linguistics authors utilized this function in 295 (13.4%) instances, which is more than their counterpart in 130 (4.6%) instances. One explanation can be the epistemic nature of theoretical fields which impose an extra cognitive burden on the readers; therefore, to lower this cognitive overstrain, theoretical linguistics authors depend on definitions. Regarding Review, Applied Linguistics authors used this function in 648 (23%) instances, while Theoretical Linguistics authors used it in half that number, i.e., 307 (12.8%) instances. This is likely because Applied Linguistics authors are inclined to draw on previous research to establish a background for their study (Arizavi & Choubsaz, 2021; Hryniuk, 2017). Looking at View and Support/Rejection functions, Applied Linguistics authors employed these two functions more than Theoretical Linguistics authors, insinuating their proclivity to take a judgmental stance on others’ works. This has already been demonstrated by Shooshtari et al.’s (2017) study as a strategy to win the acceptance of journal gatekeepers.

Table 1. Citation Functions in Disciplines based on the Developed Taxonomy

*df: 1

According to Table 2, which illustrates variations across research methodologies, citations indicating Prominence are the least favored in the quantitative article introductions, with only 60 (3.6%) instances, while citations performing this function are proportionally used more in the qualitative 164 (8.7%) and mixed-methods 168 (10.2%) article introductions. These results show that authors of quantitative articles usually prefer indicating the need for conducting their studies, using sentences and clauses of their own instead of using others’ ideas. Definitions, however, are used more in quantitative articles (200 (12.1%) of the total number of citations in this category) than in qualitative 186 (9.8%) and mixed-methods 39 (2.4%) article introductions. This can be explained by the need to define more terms and concepts in quantitative articles to operationalize the variables in concern. Overview and Review have the largest share of the total number of citations across the three research methodologies. Citations acting as Overview form almost half of the total number of citations in the mixed-methods article introductions, while in quantitative and qualitative articles this function shows smaller shares in comparison. The greater tendency to exploit this function in the mixed-methods article introductions can be attributed to the need for referring to studies from two different research traditions to establish the history and research trends on the subject (Hu & Wang, 2014; Jalilifar, 2012). Regarding citations as Review, quantitative articles had used 357 (21.6%) instances in comparison with the other two research methodologies that showed roughly equal distribution. Qualitative article introductions incorporated 393 (20.6) citations functioning as View, which are nearly twice as many as the number of citations functioning the same in the other two research traditions. The higher percentage of this function in the qualitative article introductions can be associated with the narrative nature that authors employ to report on others’ contributions, by stating their purposes, scrutinizing their methodologies, and reflecting on their findings. On the other hand, this is not quite common in quantitative and mixed-methods articles, as authors suffice to pinpoint certain aspects of previous studies and not dwell on them for long (Arizavi & Choubsaz, 2021; Harwood, 2009). Last but not least, citations in Support or Rejection featured more in quantitative articles 378 (22.8%) because authors, as stated earlier, tend to pinpoint some aspects in favor of the issue they have brought up or against others’ claims to highlight their own standpoints.

Subjecting the raw data to inferential statistics, Table 2 shows significant differences in all the functions across research methodologies (critical value: 5.991and level of significance: .05). These findings enable us to understand that the taxonomy proposed in this study could explicitly spell out the functions of citations across research methodologies in the corpus.

Table 2. Citation Functions across Research Methodologies based on the Developed Taxonomy

*df: 2

Cross-tabulating disciplines and research methodologies, as seen in Table 3, an equal number of instances of Prominence in both theoretical and applied linguistics article introductions, 30 instances in each discipline, can be seen within quantitative articles. However, there are twice as many instances of Definition in the theoretical article introductions than in the applied linguistics ones, which can be explained in light of the epistemic nature of theoretical articles for which definitions of terms are highly important to orientate the readers to the venue the authors desire to lead them. The Overview takes the highest number in both disciplines, with 246 (32.6%) and 229 (25.5%) instances in theoretical and applied linguistics article introductions. The most obvious difference in frequencies between the two disciplines is the function of View, which is about three times larger in theoretical articles than in Applied Linguistics articles. The greater number of cases of Review and View instances in Applied Linguistics article introductions is presumably due to the authors’ tendency to draw on a variety of studies to guide their readers to the claims they advance, unlike the theoretical article authors, who rely on Definitions and Overviews to do so. At last, the marginal discrepancy between the two disciplines in Support/Rejection can be linked to the nature of stance-taking in the quantitative RA introductions. More explicitly, applied linguistics authors are generally inclined to back up their own claims either by seeking support or detachment from cited contributions.

In the qualitative RAs, the situation is more complicated. The functions of Prominence and Definition are in opposite directions, i.e., more instances of Prominence are used in Applied Linguistics articles, while more instances of Definition are used in Theoretical Linguistics articles. The same explanations as the ones given for quantitative citation functions can be brought up here, too. The larger number of citations functioning as Overview in the theoretical article introductions than in the Applied Linguistics articles suggests that in theoretical articles authors simply provide a list of studies without delving into individual studies to create a new space for the study at hand. This finding contrasts with the Function of Review, which is used more in the Applied Linguistics articles, as Table 3 indicates. Just like the function of Review, the functions of View and Support/Rejection had larger shares in the Applied Linguistics article introductions. All these differences boil down to the way authors deal with cited works, i.e., whether they analyze cited works in detail or just mention them in the form of a list of recognized contributions to the field.

Within mixed-methods, theoretical articles outnumbered applied linguistics articles only in two functions of Definition and Overview, while in other functions Applied Linguistics article introductions contained more instances. Generally speaking, the authors’ overuse of Definition and Overview on the theoretical side can be justified by acknowledging the role of these two functions in the quasi-quantitative and qualitative nature of these studies, where authors have to present their ideas by incorporating two perspectives into one single research report. This finding parallels Riazi’s (2017) definition of mixed-methods research methodology.

Table 3. Citation Functions in Disciplines across Research Methodologies based on the Developed Taxonomy

The Chi-square test (Table 4) shows within qualitative and mixed-methods research methodology, there are significant differences between the disciplines with regard to citation functions (critical value: 3.841 and level of significance: .05). However, within quantitative research methodology, the two disciplines show significant differences in five functions but no significant differences in two functions, i.e., Prominence and Overview. Moreover, comparing the citation functions in theoretical articles across research methodology, significant differences were found in all functions at a critical value of 5.991 and a level of significant .05. Lastly, there are significant differences in citation functions in applied linguistics corpus across research methodologies, except for Review, where there are no significant differences.

Table 4. Chi-square: Comparisons of Citation Functions within and between Research Methodologies based on the Developed Taxonomy

*df for within comparisons: 1   *df for between comparisons: 2

Authorial Evaluation of Citations

As Table 5 demonstrates, 2455 (47.2%) citations were used to Acknowledge the cited materials, which means the authors have taken a neutral stance. This finding is in agreement with Lee et al. (2018) and Shooshtari et al.’s (2017) results, who view citation as a way for authors to articulate an expert opinion that provides an explanation for the standpoints they advocate. The Applied Linguistics authors seem to favor this stance type more than Theoretical Linguistics authors. The second most common stance type is Endorse, with a total of 1610 (31%) instances. In this category, the difference in raw frequencies between the two disciplines is not substantial enough to indicate any disciplinary variation (a difference of only 3.6%). Distance and Contest stance types were not used as frequently as the previous two types, as they both form 1136 (21.8%) instances of the total number. Regarding Distance, Theoretical Linguistics authors employed it in 527 (22.1%) citations, while Applied Linguistics authors used it in 180 (6.4%) instances. The Contest was the least typical, comprising 120 (5%) instances in Theoretical Linguistics and 309 (11%) in the Applied Linguistics sub-corpora, which means theoretical linguistics authors take more judicious stances toward others’ findings.

Table 5. Authorial Evaluation of Citations in Disciplines

*df: 1

The Chi-square test revealed a significant difference between the two disciplines in all stance-taking types (critical value: 3.841 and level of significance: .05). The findings of this analysis confirm Coffin (2009), Jalilifar (2012), and Lee et al.’s (2018) results, who found disciplinary variations in the way authors take a stance toward the cited works. This means article authors’ stance-taking manners are influenced by the disciplines they adhere to. In addition, there was an infinitesimal connection between the citation functions of Support or Rejection and author stances of Endorse or Contest. In fact, the citation functions of Support and Rejection are used in relation to the authors’ standpoints, whereas Endorse and Contest are the authors’ evaluations of the cited materials: citation functions act proactively, while authorial evaluative stances act retroactively.

Table 6 presents a cross-research methodological comparison of the authorial evaluative stance toward citations. Acknowledge is the most common type of evaluation in quantitative 876 (53%) and qualitative 1047 (55%) article introductions, but it is used only in 532 (32.4%) instances in the mixed-methods corpus. As for Distance, in which authors exploit hedging devices to lower their commitment to a claim, authors in the three research traditions seem to have used this evaluation type approximately the same. Endorse is used most frequently in the mixed-methods article introductions, 809 (49.2%), which is twice as large as the number of instances in the other sub-corpora. The lowest percentage belongs to Contest. In quantitative articles, it is used in 166 (10.1%) instances and it is used almost equally in the mixed-methods corpus in 149 (9%) instances, whereas this evaluation type was the lowest in qualitative article introductions 114 (6%).

The greatest variations, as the table indicates, are in Acknowledge and Endorse. This can be explained by reverting to the presence of an authorial voice in the academic discourse in the three research traditions. That is, acknowledging and endorsing the cited works to establish backgrounds for one’s claim would leave a marginal space for later critical review of the articles. In contrast, if authors tend to distance themselves from the sources they cite or contest the findings of others, the chances of criticism will be higher. These findings cannot be verified or disconfirm other previous findings on the grounds that there has been no study in the literature, focusing on this aspect of citation with regard to research methodological underpinnings.

Table 6. Authorial Evaluation of Citations across Research Methodologies

*df: 2

Inferentially, there are significant differences in the evaluative stances taken across research methodologies in all stance types (critical value: 5.991 and level of significance: 3.841). Although there are no comparable results from other studies to confirm or disconfirm the present findings, it is safe to claim that the research methodologies adopted in RAs most likely affect the stance-taking behavior of authors.

As shown in Table 7, within the quantitative research methodology, Acknowledge and Endorse exhibited no major variations between the theoretical and applied linguistics RA introductions. However, proportionally Theoretical Linguistics authors used more Distance 164 (21.7%) than Applied Linguistics authors 85 (9.5%), while the reverse was true regarding Contest, where Applied Linguistics authors objected to the findings or methodologies of the cited sources 105 (11.7%) more than their counterparts in the theoretical articles 61 (8.1%). One explanation for the differences in Distance and Contest in the disciplines in this research tradition can be associated with the extent to which the readership of each discipline would buy the authors’ claims and accept the stance taken. Within the qualitative tradition, Acknowledge and Distance are used more often in Theoretical Linguistics article introductions than in Applied Linguistics ones. However, Endorse is used more often by Applied Linguistics authors than by Theoretical Linguistics authors. Moreover, there is a difference between the disciplines (theoretical linguistics 36 instances (3.5%) and applied linguistics 78 instances (9%)) concerning the Contest evaluative stances. Finally, within the mixed-methods research tradition, Applied Linguistics authors used 406 (38.7%) instances of Acknowledge, which are more than the counterpart 126 (21.1%) instances in the Theoretical Linguistics corpus. Concerning the Distance stance, there are 110 (18.5%) instances in theoretical article introductions, which is more than the 45 (4.3%) instances in the Applied Linguistics article introductions. The most common stance in the mixed-methods articles is Endorse stances; there are 337 (56.5%) instances of this type in theoretical article introductions, while there are 472 (45%) instances in the Applied Linguistics texts. This major discrepancy indicates that, unlike theoretical linguistics authors, applied linguistics authors overwhelmingly take a favorable stance toward cited claims and findings. As for Contest stances, applied linguistics authors, unlike theoretical linguistics authors, appear to take an unfavorable critical stance toward cited sources in 126 (12%) instances, while in the theoretical articles this stance type is the least favored, 23 (3.9%) instances.

Table 7. Authorial Evaluation of Citations in Disciplines across Research Methodologies

According to Table 8, there are significant differences between the two disciplines within each research methodology at a critical value of 3.841 and the level of significance of .05, except for the Acknowledge cell. Additionally, regarding the stance types variations in each discipline across research methodologies, there are significant differences observed in all three-way comparisons (critical value: 5.991 and level of significance: .05). These findings support the claim that article authors take varied evaluative stances toward the citations, depending on what research methodology they employ and the discipline they adhere to (Shooshtari & Jalilifar, 2010).

Take 8. Chi-square: Comparisons of Authorial Evaluation of Citations within and between Research Methodologies

*Degree of freedom for within comparisons: 1   *Degree of freedom for between comparisons: 2

Overall, the findings in this study would enable the readers to discern the potential associations between authors’ compliance with disciplinary conventions and the research methodologies they employ in their studies. In this equation, disciplinary conventions seem to entertain less effect than the research traditions chosen in the studies. This means citation practices are influenced more by norms imposed by research methodologies, which themselves are situated in paradigms with assorted epistemological bedrocks. However, it must be acknowledged that there are other factors that impact citation practices. For example, the section of the RA where the citations are used and the journal guidelines can place deviate authors from their ordinary citation practices.

Conclusion

This study analyzed citations in theoretical and applied linguistics RA introductions across research methodologies at three levels of comparison, namely, disciplinary, cross-research methodological, and combinatory disciplinary cross-research methodologies. To this end, an inductive discourse analysis method (Barton, 2002) was adopted to identify and label context-dependent functions. Having identified the functional categories, the corpus was reanalyzed. The findings indicate explicit variations at all the three levels. As the functions were found to be context-sensitive, the variations could be interpreted in light of paradigmatic, epistemic, and methodological bearings.

The second phase of the study probed into the authorial evaluative stance taken toward cited sources. The majority of authors in both disciplines took neutral and/or supportive stances. It seems that authors in both disciplines are influenced by the rhetorical conventions and audience expectations of the disciplinary community they write for, and this has resulted in some minor variations. At the research methodological layer, it was found that authors who choose to distance themselves from or contest others’ cited works are profoundly influenced by the research methodologies they employ. This indicates an apparent causative link between stance-taking and research methodologies. The third layer of analysis considered the variations in disciplines within each research methodology. The findings revealed that while some stance types are more common in Applied Linguistics, other stance types are typical of the Theoretical Articles.

This study can provide university instructors with research-based findings on citation practices which enable them to expose students to the delicacies of each discipline concerning referencing and how to enrich one’s text through proper use of citations. Novice writers who struggle to create a research space in the fierce competition of scholarly publication can also improve their writing prowess, if they know how to coalesce disciplinary and research methodology-related conventions and preferences. Finally, materials developers would benefit from the outcomes of this study not only through raising awareness regarding proper citation practices in different disciplines or research methodologies but also by designing and devising hands-on tasks and exercises to aid writer students to improve their practices.

The study is not without shortcomings, as it employed a relatively small corpus from the field of linguistics; therefore, future research can replicate the study with a larger corpus from the field of linguistics or other disciplines. Comparing authors’ citation practices in soft and hard sciences can be a lively line of inquiry to tease out the effects of epistemic differences. Researchers can also detect citation variations in native and non-native speakers’ texts in light of research-methodological impacts (Hryniuk, 2017). More importantly, differences associated with cross-research methodological and disciplinary effects can be examined in academic discourse, when generic features such as authorial stance (Coffin, 2009), reader engagement (Hyland, 1999; Jalilifar, 2012), and evidentiality (Dehkordi & Allami, 2012) are brought to limelight.

Declaration of Conflicting Interests

The author declares that he has no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Funding

This study was not funded by any institute.