Citation use enables a dialogic culture in which writers synthesize multiple studies and position their sides in an academic community. Proper citations foreground the writers’ voices and improve the persuasiveness of an argument (Amnuai, 2018; Brown, Johnson, Smyth, & Cardy, 2014; Hyland, 1999; Mansourizadeh & Ahmad, 2011). Citation quality and quantity and their skillful incorporation into published papers is crucial for contributing to scholarly expertise and a position in the academic community. Competent citation use indicates a writer’s familiarity with the field and builds a professional identity in a tacit manner (Cargill & O’Connor, 2013).
Despite the importance of citing sources in academic writing; many writers, especially L2 writers, find it difficult to cite properly. Novice writers are often accused of patchwriting (Howard, 1995), knowledge-displaying (Petrić, 2007) or voice-repeating, rather than conversing (Samraj, 2013). For example, they may use ‘according to …’ repeatedly to parrot the claims of others and underuse non-integral citations (Thompson & Tribble, 2001, p. 100); they cannot choose appropriate content to cite (Xu, 2012; 2016), and depend heavily on the attribution function (Petrić, 2007). Many L2 writers cite previous studies to ‘display’ their subject knowledge rather than provide elaboration and clarification, in effect emphasizing knowledge-telling over knowledge-transforming (Lee, Hitchcock, & Casal, 2018, p. 2).
To facilitate L2 writers’ citation use, we argue that teaching them to understand common citation forms in various disciplines is far from enough; what to cite and why to cite also need to be highlighted. ‘What to cite’ refers to the content being cited, be it theory, definition, or findings in sources. ‘Why to cite’ emphasizes the purpose of using sources for giving attribution to the author(s), comparing sources, or providing examples. Drawing on Xu’s (2012) classification of content and Mansourizadeh and Ahmad’s (2011) functional categorization, we examine citation use in four disciplines (hard/soft; life/non-life), so as to inform writers and EAP practitioners about the epistemic conventions of information use and about recontextualizing cited information in one’s own writing.
Citation form, content, and rhetorical functions
To explore how writers interweave their persuasive statements and opinions using multiple sources, this study will define in-text citations as attributions to sources of information, such as data, methods, and ideas (Hu & Wang, 2014, p. 14). Although an in-text citation should follow journal requirements in format style regarding brackets, squared brackets, or superscripts (Hyland & Jiang, 2019), these style issues are not the focus of the present study. Our purpose is to investigate how a piece of citation fits in the new context in terms of its syntactic structure, description of information, and its rhetorical functions. Whereas the use of reporting verbs emphasizes the writer’s stance towards the authors being cited, the citation of relevant topics demonstrates a general understanding of certain fields. Following terms widely adopted in the existing literature, the present study names the syntactic structure of citation use as citation form (e.g., Cao & Hu, 2014; Hyland, 1999, 2002; Thompson & Tribble, 2001), the description of information as citation content (Xu, 2012; 2016), and the rhetorical functions as citation function (e.g., Harwood, 2008; Mansourizadeh & Ahmad, 2011; Petrić, 2007; Samraj, 2013; Thompson & Tribble, 2001).
In terms of citation form, a widely used citation framework is the form-based dichotomy of integral citation and non-integral citation, as proposed by Swales (1990). Integral citation, as its name suggests, constitutes the part of the sentence citing the author(s) by name (naming) or by indicating their actions (verb controlling) (Thompson & Tribble, 2001). Example 1 shows an integral citation with the reporting verb ‘found’, emphasizing a factual and non-disputable position towards the cited studies and cited author(s). Example 2 mentions the authors’ names in integral form, highlighting the cited authors alone. In contrast, non-integral citation ‘plays no explicit grammatical role in the sentence’ (Swales, 1990, p. 141) since it puts the cited author(s) outside of the sentence, usually in brackets, superscripts or footnotes/endnotes (see Example 3). Non-integral citations tend to be information-focused while integral ones are author-focused (e.g., Mansourizadeh & Ahmad, 2011; Samraj, 2013; Thompson, 2005).
Kieffer and Lesaux (2012a) found a consistent pattern across all groups of learners that … [AL0103, p. 874] (verb-controlling integral citation)
Our findings of the prevalence … are also similar to a recent U.S. study by Tsiouris et al. (2012) … [Edu0402, p. 976] (naming integral citation)
Oxygen reduction/evolution reactions (ORR/OER) are critical reactions…  [Phy0103, p. 1] (non-integral citation)
In terms of citation content, a top-down categorization is preferred due to its comprehensive coverage (Xu, 2012). Based on her examination of three part-genres in fifteen applied linguistics research articles, Xu (2012) proposed six types of commonly cited content. One is topic/content, where author(s) cite the current research direction, interest or specific concepts, similar to what Coffin (2009, p. 164) calls ‘ideas’ (see Example 4). Another is finding, which refers to the discoveries or results made by previous scholars. The citation of finding is often preceded by a reporting verb synonymous with ‘to find’ as shown in Example 5. Opinion/explanation involves the citation of attitude and stance towards previous works, similar to ‘knowledge claims’ in Hu and Wang’s (2014, p. 14) terms, or the analysis of findings in published studies. Example 6 uses ‘powerful’ to confirm the cited authors’ opinions about synthetic CREs. Definition/term gives the source(s) a technical term or terminology (also in Hu & Wang, 2014). Method refers to the designs, data, or tools adopted in existing literature. Theory/model concerns the citation of principles, theories, or frameworks.
Fluorescent low-molecular-weight gelators (LMWGs), … are drawing great attention due to their potential in… [Phy0201, p. 1] (topic/content)
Wang et al. (2006) found that after controlling for phonological awareness and oral receptive vocabulary… [AL0103, p. 873] (finding)
Synthetic CREs are powerful tools for uncovering interactions between TFs (Gertz et al. 2009; Sharon et al. 2012; …). [Bio0501, p. 778] (opinion/explanation)
CarbonMide, a carbon fibre reinforced Polyamide twelve powders from EOS, has a tensile strength… [Phy0301, p. 1019] (definition/term)
we adapted a previously developed fluorescence resonance energy transfer (FRET)-based real-time assay of DNA strand exchange14. [Bio0401, p. 1] (method)
Kearney’s (2002b) four-factor model was not supported by all three indices of fit. [Edu0302, p. 183] (theory/model)
This citation content framework is comprehensive and useful for instructing L2 writers. If opinion/explanation citations predominate in the Results and Discussion section as found in Xu (2012), the instructor can encourage L2 writers to evaluate different opinions and/or take sides in explaining their results. Therefore, this framework can be effective in teaching L2 writers to integrate proper content into their citation practice.
In terms of citation function, a corpus-based approach has been broadly adopted (Mansourizadeh & Ahmad, 2011; Petrić, 2007; Samraj, 2013; Thompson & Tribble, 2001) because it has the capacity to process large amounts of data, thus enabling the detection of generalized patterns. As the identification of rhetorical functions is derived from contextual cues, some researchers refer to function together with content, such as comparison of one’s work with that of other others (Petrić, 2007) or evaluation of the field or evaluation of study (Samraj, 2013). However, comparison can be made between opinions or models and not only between one’s work and that of other others. Both evaluation of study and evaluation of field belong to the function of evaluation. Evaluation, a rhetorical function, shows the purpose or intention of the writer(s) while study and field refer to content being cited, or what the writer(s) cite. The reason why the present study separates citation content from citation function is to differentiate ‘what to cite’ from ‘why we cite’.
As the study’s aim is to examine citation use in two part-genres - the Introduction and the Results and Discussion - the categorization of citation functions in Discussion sections (see Samraj, 2013) may not be applicable to the cases in Introduction sections. The citation function of interpretation of results (Samraj, 2013, p. 305) mainly refers to the explanation of findings in the Discussion part since this function seldom occurs in the Introduction section. Therefore, the functional categories in the present study summarize the major rhetorical functions that citations perform in both part-genres.
Disciplinary variation and generic variation in citation use
As each discipline follows specific conventions of knowledge construction, its citation use will vary to some extent. Integral citations account for 65% of the total citations in philosophy whereas non-integral citations are used more than 60% of the time in other disciplines (molecular biology, magnetic physics, electronic engineering, mechanical engineering, sociology, marketing and applied linguistics), according to Hyland (1999, p. 347; 2002, p. 123). Philosophy is not the only discipline preferring integral citations. Charles (2006) demonstrates the predominance of integral citation with human subjects in politics and materials science (over forty per 100,000 words). Thompson and Tribble (2001, p. 94) also show that integral ones comprise over 60% of the total citations in agricultural economics. Nevertheless, non-integral citations dominate in many hard science fields (engineering in Mansourizadeh & Ahmad, 2011; general medicine in Hu & Wang, 2014; psychology in Cao & Hu, 2014), as well as in some soft science fields (education in Cao & Hu, 2014). Whether there is a hard/soft distinction in citation type needs further examination.
Citation content choice is a significant challenge for L2 writers when they have to decide what to cite to support their arguments (Xu, 2016). She found that experts and PhD students in applied linguistics cite more opinion/explanation and findings than topic/content, while undergraduates and masters cite more content of previous studies than other functions. Her robust work provides a comprehensive picture in applied linguistics, but the distribution of citation content in other disciplines has not yet been examined.
The rhetorical functions of citations differ across disciplines and part-genres. Thompson and Tribble (2001) compared eight agricultural botany theses and found that source and identification are frequently used in Introduction sections as well as in Results and Discussion sections, while origin and reference are quite common in methods sections. The highly rated theses in Petrić (2007) demonstrate an overall preference to attribution (>70%) throughout all sections, but more evaluation in literature review sections than other sections, and more application in Analysis and Conclusion parts than others. Student theses are one data source embodying the part-genre variation, but expert writers’ papers exhibit a similar pattern. Mansourizadeh and Ahmad (2011) examined expert papers in the chemical engineering field and found that while identification, establishing links and attribution are predominantly used in papers’ Introduction sections, support, comparison and identification are frequent in Results and Discussion sections. In a study of Discussion sections of biology papers, Samraj (2013, p. 304) found that expert writers adopted a comparison of results and interpretation of results more often than not to highlight their own findings. The above-mentioned variations will confuse L2 writers in their citation practices because these variations could be related to conventions of disciplines and the rhetorical purposes of writing. Therefore, it is necessary to examine both the syntactic and functional use of citations in part-genres, together with the disciplines.
As some part-genres have fairly dense citation use compared to other academic papers’ sections, choosing typical sections in research articles is both necessary and practical. As Hyland and Jiang (2019, p. 65) state, ‘[Citations] are heavily concentrated in the early sections of the article’ to facilitate readers’ quick discovery of relevance and creativity of an argument. Not only does the Introduction section accommodate considerable citation use, but the Results and Discussion section works towards ‘constructing and reinforcing the principal lines of argument pursued’ (Liu & Buckingham, 2018, p. 97); thus, citation use in this part-genre also merits due attention. Hence, the present study will examine two part-genres in academic papers: the Introduction section and the Results and Discussion section.
To offer L2 writers a systemic view of citation use in experts’ writing, we adopt a corpus-based method to examine a tripartite pattern, focusing on the three aspects: Form-content-function in research articles across four disciplines (condensed matter physics, biochemistry and molecular biology, special education, applied linguistics). Specifically, we aim to answer three questions:
How does citation use in the tripartite model (citation form, citation content, and citation function) in the Introduction sections vary across these four disciplines?
How does citation use (citation form, citation content, and citation function) in the Results and Discussion sections vary across the four disciplines?
What are the most frequent co-occurring patterns of citation context and citation function?
To answer these questions, the second author collected sixty empirical research articles from the four disparate academic disciplines and compiled a corpus of EJAC-2018 (English Journal Articles’ Corpus-2018), with a total number of 363,948 tokens. As our purpose was to examine citation features, irrelevant information such as author name, and all visual materials such as graphs, data tables, excerpts from sources, personal interviews, appendices, acknowledgements, and references were excluded. The tokens of each part-genre were calculated in the four disciplines to enable genre-based comparison (see
Descriptive statistics for EJAC-2018
|Disciplines||No. of articles||Tokens||Tokens in introduction||Tokens in R & D|
Our study’s criteria for selecting the five top journals in each discipline involved two considerations. One was their leading roles in the specialized discipline. We referred to their five-year impact factors (2014-2018), ranging from 1.9 (education) to 30 (biology). All journals were confirmed to be authoritative by subject experts we consulted personally. Another consideration was the impact of the manuscript submission guides of these journals on citation form. Although these guides explicitly require a particular format (numerical endnotes, superscript numbers or author-date in the parenthesis), they do not mandate citation form specifically. Admittedly, because ‘numerical conventions predispose the writer to choose non-integral citations’ (Swales, 1990, p. 47), the requirement of a specified format may affect writers’ choice between integral and non integral forms. As there is no significant relationship between numerical conventions and citation forms (Charles, 2006), our study discounted the impact of submission guides.
From each of the five top journals in each discipline, three empirical articles published online in 2014-2018 were chosen to represent disciplinary research articles (hereafter termed RAs). The study’s choice of representative disciplines was based on two considerations. One was that the disciplines of ‘hard/soft’ and ‘life/non-life’ (Biglan, 1973, p. 202) would influence citation form because according to recent research, integral citations were found more commonly in philosophy, a typical soft science subject, than in physics, a hard science (Hyland, 1999). Another consideration was whether the articles are ‘applied/pure’ studies (Biglan, 1973). Our scope was narrowed to four applied second-level disciplines, i.e., biochemistry and molecular biology, condensed matter physics, applied linguistics, and special education, as shown in
The criteria of first-level discipline selection (adapted from Biglan, 1973)
After the corpus of sixty articles was compiled, MaxQDA 12.0 was used to annotate citations in each paper’s Introduction and in each paper’s Results and Discussion sections. In the process of identifying citations in these two part-genres, each bracket or superscript number was counted as one citation (Mansourizadeh & Ahmad, 2011; Samraj, 2013; Xu, 2012). The material in brackets was intended to give additional information which was not considered to be citations in the present study. Although brackets are ubiquitous in soft sciences, superscript numbers are required by hard science journals to evidence citations. The difference in hard science journals’ use of superscripts was accounted for in our quantitative analysis.
Before annotation began, a coding system was clarified and agreed upon by the two researchers. The citation form and content were relatively easy to identify due to syntactic and semantic features, but citation function could be singular and multiple, adding to coding complexity. Similar to Mansourizadeh and Ahmad (2011) and Samraj (2013), we assigned one function to one citation place regardless of single or multiple citations. We noted the primary function (Aijmer, 2013) for each case to match with citation content later. To achieve this goal, the surrounding contexts were often referred back to in order to ascertain particular functions.
After examining the functional taxonomies of various disciplines, such as agricultural botany and agricultural economics (Thompson and Tribble, 2001), gender studies (Petrić, 2007), physics (Moravcsik & Murugesan, 1975), and engineering (Mansourizadeh & Ahmad, 2011), we arrived at a modified version of citation functions.
The categories of citation form, content, and function in the present study
|Citation form||Integral (verb controlling)||Citation function||Attribution|
|Integral (naming)||Link / synthesis|
|Topic / content||Further reference / exemplification|
|Opinion / explanation||Application|
|Definition / term||Other|
|Theory / model|
The modification of functional taxonomy was based on these considerations: first, link, and comparison were independent functions. Unlike other researchers who distinguish between them by agents (see Petrić, 2007; Mansourizadeh & Ahmad, 2011), we differentiated them according to rhetorical purpose, designating a link to show the establishment of a research network by citing more than two studies, and comparison to emphasize similarities or differences detected between two studies cited. Secondly, identification (see Thompson & Tribble, 2001) was included because our pilot study showed that identification is often used in hard science academic papers to reveal an agent in a sentence by using a citation. Thirdly, citations with unidentifiable or obscure uses were coded as other.
Two functions in Petrić’s (2007) taxonomy were not adopted: statement of use and evaluation. The statement of use function is frequently used in theses to show writers’ decisions or intentions (Petrić, 2007) while seldom being used in research articles (Xu, 2012). The function of evaluation is performed by the use of evaluative language such as ‘I think’ (Petrić, 2007, p. 245), which is not a focus of the present study. Instead of a separate category, the potential evaluative power of citations was grouped as link/synthesis or comparison because integrating previous studies into the writers’ line of argument involves subtle evaluative techniques.
For reliability, the second author coded all data, and the first author chose 10% of the texts on which to do independent annotation. The cross-rater reliability of citation form, citation content, and citation function between the two researchers was 100%, 100% and 98%, respectively.
Citation use in Introduction sections across four disciplines
Placed at the beginning to establish importance, a paper’s Introduction section is assumed to provide the basis for claiming the prominence of one’s research and for defining the writer’s position within an academic community (Swales, 1990).
The citation features in Introduction sections across four disciplines (ratio: per 1,000 words)
|Biology (ratio)||Physics (ratio)||Education (ratio)||Applied Linguistics (ratio)||Total|
|Content / Topic||11(1.3)||34(3.7)||74(2.8)||136(5.2)||255(3.6)|
|Opinion / explanation||26(3.2)||17(1.8)||64(2.4)||92(3.5)||199(2.8)|
|Definition / term||28(3.4)||30(3.3)||43(1.6)||52(2.0)||153(2.2)|
|Theory / model||5(0.6)||6(0.7)||19(0.7)||10(0.4)||40(0.6)|
|Link / Synthesis||79(9.6)||62(6.8)||130(4.9)||107(4.1)||378(5.4)|
|Further reference / exemplification||17(2.1)||30(3.3)||90(3.4)||87(3.3)||224(3.2)|
Cross-disciplinary differences were then examined by an online log-likelihood calculator (https://iresearch.unipus.cn/info/method/keyangongju_2.html), yielding log-likelihood values (hereafter LL) and significance levels (hereafter p). Experienced writers tend to favour non-integral citations in these four disciplines although writers in AL demonstrate fewer non-integral citations than writers in other disciplines (LL(AL vs Edu) = 26.44, p = .000; LL(AL vs Phy) = 52.03, p = .000; LL(AL vs Bio) = 77.38, p = .000). The dominance of non-integral citations echoes what has been found in research articles (Cao & Hu, 2014; Hyland, 1999; Mansourizadeh & Ahmad, 2011) and doctoral theses (Thompson & Tribble, 2001). It can be inferred that expert writers prefer foregrounding their ideas and creativity or emphasizing a natural flow of their scientific story by non-integral citations. Here, the importance of the cited authors is emphasized in implicit ways, and the higher frequency of verb-controlling integral citations than of naming integral ones demonstrates the originators’ less prominent role.
When it comes to citation content,
Citing content/topic also shows a difference in discipline. It is employed significantly less in biology (1.3 per 1,000 words) than in the other three disciplines (LL(Bio vs AL)= 20.89, p = .000; LL(Bio vs Edu)= 4.90, p = .027; LL(Bio vs Phy)= 8.62, p = .003), yet its use in applied linguistics (5.2 per 1,000 words) seems significantly higher than in education (LL = 18.20, p = .000) and biology.
Among seven citation functions, attribution is dominant in all four disciplines (> 7 per 1,000 words), followed by links/synthesis (> 4 per 1,000 words), while application seldom occurs (less than 1 per 1,000 words). There is one major disciplinary difference in the distribution of links/synthesis in that the frequency of synthesizing more than three distinct sources in biology and physics is significantly higher than that in the other two disciplines (LL(Bio vs Edu)= 22.68, p = .000; LL(Bio vs AL)= 34.00, p = .000; LL(Phy vs Edu)= 4.27, p = .039; LL(Phy vs AL)= 9.58, p = .002).
Citation use in Results and Discussion sections across four disciplines
Even though citation use in the Results and Discussion section is numerically much less prevalent than in the Introduction section (6.8 > 21.1 per 1,000 words), citations in the Results and Discussion by expert writers display distinct rhetorical strategies in this part-genre (see
The citation features in Results and Discussion sections across four disciplines (ratio: per 1,000 words)
|Biology (ratio)||Physics (ratio)||Education (ratio)||Applied Linguistics (ratio)||Total|
|Opinion / explanation||36(0.6)||27(0.9)||39 (1)||45(1)||147(0.8)|
|Definition / term||54(0.9)||3(0.1)||8(0.2)||9(0.2)||74(0.4)|
|Content / topic||24(0.4)||3(0.1)||20(0.5)||22(0.5)||69(0.4)|
|Theory / model||12(0.2)||3(0.1)||27(0.7)||9(0.2)||51(0.3)|
|Link / synthesis||78(1.3)||42(1.4)||47(1.2)||40(0.9)||207(1.2)|
|Further reference / exemplification||24(0.4)||12(0.4)||31(0.8)||36(0.8)||103(0.6)|
Among the four disciplines the study measured, average citation density in the Results and Discussion section (av. 8 per 1,000 words) is the greatest in biology, supporting its reputation as a ‘heavily citing’ tradition (Hyland & Jiang, 2019, p. 70; Samraj, 2013). In academic biology papers, we found that citation forms were concentrated in the non-integral form and were not as varied in form as compared to papers from the other three disciplines. By contrast, in education and AL academia, the naming integral form seems to be preferred. We found it was employed in more than 1.2 per 1,000 words, demonstrating how an acknowledgement of originators aligns with the commonly held belief. This is the citing preference in the soft sciences (Cao & Hu, 2014).
When it comes to citation content, findings are cited much more often than other content types. In the Results and Discussion section, citing findings from other sources would enable comparison and contrast, or further explanation, while citing findings in the Introduction section mainly sketches research development. Apart from findings, biology writers prefer citing methods while Education and AL writers tend to cite opinion/explanation (1 per 1,000 words). The experimental nature of biology, where innovation of methods moves the discipline forward, would lead to greater use of citing methods. In biology, the frequency of definition/term citing is also significantly higher than in the other three disciplines surveyed: (LL(Bio vs Edu) = 16.30, p = .000; LL(Bio vs Phy) = 15.99, p = .000; LL(Bio vs AL) = 18.73, p = .000); and education articles cite theories/models much more than the other disciplines (LL(Edu vs Bio) = 18.78, p = .000; LL(Edu vs AL) = 17.79, p = .000; LL(Edu vs Phy) = 15.16, p = .000).
In Results and Discussion sections, experienced writers vary their citations’ functions, not relying on one function, though attribution dominates in all four disciplines (range 1.8 to 2.9 per 1,000 words). There are differences; however, in the second-choice rankings between Introduction and Results and Discussion sections. Unlike Introduction sections where link/synthesis ranks second, comparison is the second most frequently used function in Results and Discussion sections (more than 1.5 per 1,000 words), indicating a necessity to compare one’s findings with those of other experts, as published in the existing literature.
The co-occurring patterns of citation content and citation function
To understand how the features of a citation impacts context, we examined co-occurring patterns of citation content and function in two part-genres. Cut-off frequencies were set at 3 per 1,000 words in Introduction sections and 1 per 1,000 in Results and Discussion sections because the number of citations in the latter sections is three-fold fewer than those in the former. This significant difference in citation number between the two sections is not uncommon (Mansourizadeh & Ahmad, 2011), but limits the generalization of reoccurring patterns.
Co-occurring patterns of citation content and function in two part-genres
|Introduction||Pattern 1: attributing the findings, definition/term, opinion/explanation to other sources|
|Pattern 2: linking/synthesizing the findings, opinion/explanation, content/topic from various sources|
|Pattern 3: further referencing the findings to other sources|
|R & D||Pattern 1: attributing the findings to other sources|
|Pattern 2: comparing the findings from other sources|
|Pattern 3: linking/synthesizing the findings from various sources|
After examining citation form, content, and function in the Introduction and the Results and Discussion sections, we found that there are epistemic influences on disciplinary conventions among the four disciplines although they are not as sharply differentiated as if they were from two distinct worlds. Traditional soft sciences do not overwhelmingly use the integral form to flag authorship. Soft sciences display a non-integral-dominant writing style, which probably relates to what Hyland and Jiang (2019, p. 74) called ‘scientism’ in the social science disciplines.
The predominance of non-integral form in the four disciplines under our examination has its merits since using superscripts and/or placing author information in brackets does not interrupt information flow but, rather, will allow for foregrounded ideas and propositions (Hewings, Lillis, & Vladimirou, 2010). Such a message-salient strategy can contribute to academic objectivity (e.g., Cao & Hu, 2014; Hyland, 2002) by minimizing the human effect on statements. Following this convention helps establish the image of a knowledgeable ‘insider’ in the academic community (Ivanic, 1998, p. 68), that of an authority who is joining the academic conversation.
Despite the overall tendency to rely on the non-integral form, soft science writers embrace more verb-controlling integral forms in Introduction sections and more naming integral forms in Results and Discussion sections than their hard science counterparts. In their Introduction sections, writers in education and AL fields are likely to show their stance towards propositions via reporting verbs because these disciplines are characterized by a horizontal knowledge structure, where ‘non-comparable principles of description based on different, often opposed, assumptions’ are ubiquitous (Bernstein, 1999, p. 173). In Results and Discussion sections however, the same writers establish their claims on acknowledging voices of other sources to highlight a ‘knower visibility’ (Cao & Hu, 2014, p. 28). The practice of paying tribute to prestigious authors suggests a hierarchical knower structure in soft science disciplines (Maton, 2007).
Disciplinary and generic similarities can also be found in citation content and function. Attributing the findings to sources other than themselves tends to be a major citation pattern across all of the four disciplines and both part-genres. As the ‘simplest function’ (Petrić, 2007, p. 247), attribution provides the source of information and distinguishes findings or opinions of others from the writers’ own, akin to ‘self-defence’ by indicating such available support from within their academic set (Harwood, 2008, p. 8). A respectful posture of giving credit where credit is due squelches the potential for accusations of plagiarism (Abasi, Akbari, & Graves, 2006; Flowerdew & Li, 2007). As attribution only shows the sources and expands the dialogue, an overreliance on this function would blur the writers’ own attitudes and confuse readers, which is evident in low-rated writers (Petrić, 2007).
Although both sections largely cite findings, their rhetorical purposes are different. In the Introduction section, expert writers tend to cite others’ findings to link/synthesize a general research background and portray the topic’s current development, while in the Results and Discussion section, the writers are likely to cite to make comparison. Synthesizing findings from other sources reveals the writers’ proximity to the academic community, i.e., those with whom they identify and how they are identified by others (Hyland, 2012, p. 29). In contrast, comparing writers’ own findings with others’ can provide support for a writer’s argument, while contrasting them with inconsistent findings and demonstrates creativity and contribution (Basturkmen, 2009).
Several disciplinary differences in citation content and function are worth mentioning. Biology writers in their Introduction section cite content/topic less often than AL writers do, probably indicating their ‘specific’ (Becher & Trowler, 2006, p. 115) and communal interests. A preference to link/synthesis among biology writers who employed this function more than double as much as AL writers do is perhaps because the field of biology builds a hierarchical knowledge structure where knowledge is established on ‘highly general propositions and theories’ (Cao & Hu, 2014, p. 28) with fewer conflicts over topics or propositions. Another possible reason for the adoption of links/synthesis by biology writers could be their need to display abundant information within a limited space (Mansourizadeh & Ahmad, 2011, p. 158). As biology writers tend to show their creativity on method improvement, they cite method quite often in the Results and Discussion section, which is an unexpected discovery from our study.
For L2 writers and EAP practitioners, our study suggests the importance of using a form-content-function integration to facilitate proper citation use. In current academia, using the integral form seems unpopular or out of fashion, yet its role in showing writers’ stance and acknowledging originators is clearly defined in the soft sciences, as popular guidebooks indicate (e.g., Gayte, 2013; Swales & Feak, 2012; Wallwork, 2013). The emphasis on citation form alone in guidebooks however cannot reflect how various sources are interwoven into storytelling and argument establishment. If the co-occurring patterns of content and function have not yet been introduced, L2 writers will heavily rely on attribution (Petrić, 2007), and this will largely centre around content/topic (Lee et al., 2018). Lee et al. (2018) found that L2 students’ tendency to hold a neutral, non-committal stance stems from a limited language ability for critiquing sources. Yet we believe the neutral stance is partly because of their selection of citation content. Even if L2 writers would like to take a critical stance on information they cite, some content, such as broad topics or popular theories, are more difficult for students to evaluate. Describing relevant findings can move beyond simply acknowledging sources as it engages with source details.
The results of this study can be used as pedagogical material for L2 writers. One important activity in EAP classes is source use in different sections. Instructors can prepare typical citation samples for each section, such as link/synthesis in the Introduction and comparison in the Results and Discussion, and ask learners to identify each citation’s content and purpose. Through investigating rhetorical functions and specificity of content on their own, learners will develop skills in critical reading and will gradually pattern their own argument in similar ways to the models. We suggest that students’ greater awareness of the range of citation patterns and ability to use them purposefully can encourage L2 writers to incorporate more useful sources into an argument. Within a specific context, student practitioners need to closely analyse whether their citations truly serve a rhetorical purpose or are merely an extension of descriptive writing. Students are usually discouraged from doing the latter in favour of developing their persuasive writing rather than repeating the information they found.