Explicitness of Attribution in Academic Discourse

Academic discourse is a kind of dialogic interaction between scholars and the interplay of ‘averral’ and ‘attribution’. Citation, as source using, is the means of attributing the borrowed propositions to a particular source. This study addresses the issue of classifying citations. By comparing the classification scheme of integral citations with that of non-integral citations, the authors argue for the necessity of a form-based scheme. A new typology of non-integral citations is proposed according to their formal features. The sub-types of citation (e.g., chorus-citation, solo-citation) are characterized along the continuum of attribution explicitness, ranging from low to high.


INTRoDUCTIoN
Citation is the important formal aspect of source using, as well as the essential element of argumentation in academic writing. Recent years witnessed increasing interest in citation analysis in the hope of uncovering the rhetorical practice underlying the variation of citation. Effective analysis of citations depends on effective classification and description of the forms and functions of citations. Despite Swales ' (1990) dichotomy of non-integral and integral citations, views on the sub-categorization of either are quite divergent, especially for non-integral citations. This study focuses on non-integral citations. Following Thompson & Tribble's (2001) function-based typology, we propose a formbased typology as an alternative. Based on the assumption that the specific communicative functions of language are usually associated with particular syntactic forms, this study aims to explicate the surface features associated with the functions of non-integral citations. We propose the dimensions of attribution explicitness to explain the rationale and motivation of the forms and functions of citations, despite the interview-induced claim that no clear rationales always emerge for either form or function of authors' citing behaviors (Harwood, 2008(Harwood, , 2009. Citations have been categorized in various ways, either formally (Charles, 2006;Swales, 1990) or functionally (Hyland, 2004;Harwood & Petrič, 2012;Thompson & Tribble, 2001). The widely accepted dichotomy of integral and non-integral citation, proposed by Swales (1990) is form-based (whether the names of cited author play any syntactic role in the sentence), so is the classification of integral citations (Thompson & Tribble, 2001;Swales 2014). Swales's (2014) five-fold subcategorization (Author as subject, Author as agent, Author as adjunct, Author in NP, Author-other) is similar to but more fine-grained than Thompson & Tribble (2001). Verb-controlling citation, in Thompson & Tribble' term, means roughly the same as Author as subject. In the same vein, the last category of both schemes are synonymous.
Naming citation subsumes the three intermediate sub-types in Swales's scheme: Author as agent, Author as adjunct, Author in NP. The following examples are taken from Swales (2014: p.124).
1. It was hypothesized by Myers (1966) that the freshwater fishes of the West Indies dispersed from Central America. 2. According to Myers (1966), freshwater fishes of the West Indies likely dispersed from Central America. 3. Myers' 1966 hypothesis proposed that freshwater fishes....
All three sentences are the citational variations of the same content. In (1), the cited author "Myers" is the agent of the passive structure "be hypothesized". In (2), the prepositional phrase "According to Myers (1966)" is the adjunct of the sentence. In (3), the author's name (in possessive form, Myers') is part of a noun phrase.
While Thompson & Tribble's (2001) sub-categorization of integral citation is form-based, their scheme of non-integral citations is function-based. The functions include Source (attributing a proposition to another author), identification (identifying an agent within the sentence it refers to), reference (referring the reader to another text), and origin (indicating the originator of a concept or a product).
We argue that form-based typology is more valid and reliable, because the function is not as transparent as forms, and is difficult to identify. Harwood (2008) considers it a subjective and rather chancy business to attempt to read off citation function from purely textual evidence. According to Cronin (1984), the purposes of and motives for many citations remain elusive and evasive. Bhatia (2004) emphasize the "privateness" involved in the choice of authors to cite and the way in which they are cited.
For the reasons given above, the function-based typology is less operationalizable. As an alternative, a form-based typology of non-integral is in order. According to the tenets of cognitive linguistic theories, the specific communicative functions of language are usually associated with particular syntactic forms. This study aims to explicate the surface features associated with the functions of non-integral citations.

INTERPLAy BETwEEN AVERRAL AND ATTRIBUTIoN
Research means standing on the shoulders of giants. Academic writing, in essence, is a kind of dialogic interaction between scholars, specifically the writer and other researchers. In the minds of academic writers, the in-depth rhetorical structure that underlies an effective argumentative writing is the template of "They say, I say", so good writers "do more than make well-supported claims ("I say"); they also map those claims relative to the claims of others ("they say")" (Graff & Birkenstein 2010: xix). Charles (2006a: 494), using Sinclair's (1986) notions of 'averral' and 'attribution', claims that "when an attribution is made, a proposition is credited to a source other than the writer and responsibility is assigned to that person or entity." The attribution part of a text is the propositions that are other-sourced, i.e. the source is attributed outside the thesis writer (e.g. some authors; the Secretary-General). It is further subdivided into integral citation, non-integral citation, and general reference. (For further details see Charles 2006b). General reference refers to "a category of research reports in which there is no speciðc reference point clearly identiðable." (Charles 2006b: 315) In a sense, general reference is kind of attribution that is less explicitly signaled. Consider the example: 4. It is well known that uniformity along the length of the superconducting wire is one of the key factors... (taken from Charles 2006b: 315) Averral, as the opposite of attribution, is self-sourced, and is also divided into three types: emphasized averral, hidden averral, averral without attribution (Hunston 2000, Charles 2006a): • Emphasized averral: The writer stresses that they are the source of the proposition by overtly attributing it to themselves and thus they take responsibility for its reliability (e.g. I think; this author suggests). • Hidden averral: The writer attributes a proposition to some unspecified source (e.g. people in general). • Averral without attribution: The writer avers a proposition without attributing it to a source; thus the reader is left to infer that the writer is responsible for the proposition (e.g. it can be seen).
It seems that all unattributed propositions in a text are averral and "the writer is assumed to take responsibility for their veracity" (Charles 2006: 494). We would add that the writer has the choices of making himself visible or not.
In our understanding, the boundary between averral and attribution is not clear-cut. The overlapped area between them is the hidden averral (or "general attribution" using Charles's terms) and hidden attribution (general reference).
overt averral--hidden averral: general attribution--hidden attribution:general reference--Overt attribution: Citation (integral; Non-integral) The overlapping categories of reporting are quite similar: there is no clear indication of whether the report is self-sourced or other-sourced; the report sounds to be other-sourced, but the source is not given, for the absence of citational reference. Charles (2006b) summarized the features of hidden reports that are other-sourced. Markers in clauses with human subjects are summary references to several researchers (e.g., writers who think that...); those in clauses with non-human subjects are often general references to schools of thought (e.g., the motivational approach); markers in it clauses are usually adverbs (e.g., often).
However, the above characterization seems inadequate when applied to identify specific instances. For instance, the following sentence patterns, taken from Charles (2006a), were deemed to be of two types of hidden reports: self-sourced and other sourced. They are so similar that it is almost impossible to tell them apart: • It is hoped that… • It has long been known that … • One can argue that … • Many have argued that … • The competitive model predicts that … In our understanding, the two types of reports should be treated as the same and fall into the same category, and are termed "hidden reports".
Some of the hidden reports are self-sourced, others are generally attributed. Further complication is that general attribution is occasionally followed with a parenthetical reference to indicate its source. In other words, general attribution is flexible about indicating the sources or not. Therefore, it is difficult to draw a line between them, as Charles (2006b: 315) notes, "it is unclear whether the report refers to a speciðc reference point or not." It seems that sometimes the writer chooses to blur this distinction and to introduce indeterminacy into the text." overt averral --Hidden reports: with(out) citation--Overt attribution: Citation (integral; Non-integral) An explicit source use (citation) consists of two components: Proposition presentation plus documentation. The former belongs to the main body of text while the latter is parenthetical, it serves as an in-text shorthand that links to the end-of-text bibliography entry where full-fledged information is provided. In a non-integral citation, the two parts are separately given; Integral citation is the blending of the two. The original author plays a part in both components. Specifically, according to the author-date documentation system, when we move the author's name(s) from the inside of parenthetical reference () to the outside, a non-integral citation is transformed into integral. In other words, we are assuming that the form of integral citations is derived from that of non-integral citations, and the latter is the "deep structure" of the former, to use the terms of Chomskyan linguistics. For example, the skeletal integral citation "Zhangsan argues that …. (####)" can be considered as the transformed rewriting of the non-integral one "Some researcher argues that …. (Zhangsan ####)". Or "There is an argument that … (Zhangsan ####)" Our argument that non-integral citations are more basic and normal also accords with the statistics of cross-discipline comparison that the frequency of non-integral citations outnumbers that of integral (Hyland 1999, with philosophy being the only exception). Charles's (2006b: 316) result is the opposite, where integral citations boast a higher proportion. I think the reason is related to the data. Charles's data is confined to the portion of citation with reporting that-clauses. It fails to take into consideration non-reporting non-integral citations.
Compared with non-integral citation, citing integrally is a kind of rhetorical manipulation. The immediate effect of moving the author's name is to get it highlighted (Weissberg & Buker, 1990), and the writer's rhetorical purpose could be "acknowledging the prominence of the cited author" and resort to the authority to lend support your claims, among others.
Among the three sub-types of integral format, the cited author's status is elevated to a different degree, where verb-controlling is more highlighted than naming.
As mentioned above, in the integral citation, the cited author is emphasized when his name is moved to and appears in, proposition presentation part of the text. Meanwhile, the sentence internal name also has a "side-effect". In our understanding, the cited author's name, alongside with other element (e.g. reporting verbs), serves to signal other-sourcedness, as well as locate the source. In contrast, there is no name-signal in non-integral citations. Therefore, the boundary between cited ideas (attribution) and writer's own claims (averral) is not clear, unless otherwise indicated. The most effective way of marking attribution is using a reporting clause. We suggest that there are also other marking options that are less discussed, and that different citing formats reflect different degree of explicitness of attribution marking.
In the following section, we are going to explain the different options of attribution marking in non-integral citation format.

TyPoLoGy oF NoN-INTEGRAL CITATIoNS
As we know, non-integral citations are distinguished from integral ones in terms of the formal features, that is, whether the names of cited author play any syntactic role in the sentence. One can even identify a non-integral citation by looking for the sequence of author-date in brackets (if the documentation style is APA). In this sense, bracket is a formal marker that differentiates integral and non-integral citations. This identification procedure can be automatized and run on computer software.
Since the function-based typology is less operationalizable, as an alternative, a form-based typology of non-integral is in order. The issue of our concern is how to identify the sub-types of nonintegral citations, and whether there are any formal markers that could differentiate the sub-types.

Chorus Citations: Citing Collectively
In academic writing, when a writer is using sources, he may cite one work, two or more works at a time. Citing three or more works in a single parenthetical reference would produce a long reference with a lengthy list of names in parentheses. Some exceptional parenthetical reference is so long that it keeps running two or more lines on end. Despite the risk that "[a] long parenthetical reference … may prove intrusive and disconcerting to your reader" (Gibaldi 2009 p.229), many writers still choose to cite more than one work in a single parenthetical reference.
The options concerning the number of citees seem to be too commonplace to merit our attention. The writer's citing choice seems to be arbitrary. However, we would argue that there is important distinction between citing one work or more than one work at a time, and the choice among them is motivated.
For the sake of expository convenience, we would coin the new terms-"solo citation" and "chorus citation" to refer to the practice of citing one work at a time individually and citing more than one work collectively, respectively, by analogy with the programs of a sing-a-song performance.
The analogy between singing and citing becomes obvious if we think of the relative prominence of the person involved in the solo and chorus. The solo singer, as the only performer on the stage, is much prominent than a counterpart in a chorus, as is the case for citees in a solo citation and chorus citation.
In chorus citations, as shown in the example (4), readers' can't identify whose point is who? What is each citee's specific point? In fact, it is not the writer's purpose to distinguish each author's point, but to assimilate the bundle of studies, by mixing them up. This kind of citation is also called "Ceremonial references" (Adatto and Cole, 1981) used to signal the orientation and belongingness of a publication, or "Orienting Reference Lists" (Hargens, 2000) to legitimate the citing research and lead the readers into the research area.
While arguing for the legitimacy and belongingness of a research, the researcher is introducing himself into the academic community of previous researchers. Imagine a situation when the writer is entering a room with the presence of some well-known researchers, he may be too busy to shake hands with them one by one. Instead, he chooses to give "nods all round to previous researchers" (Swales 2014: 135). The chorus citations writer seems to be too eager and hasty to enter the room to get each face recognized. Therefore, we suppose a chorus citation marks attribution in a least explicit way.
As we mentioned above, chorus citations emphasize the similarity among a list of relevant works, and pay less attention to the distinction between them. At other times when there is the necessity to differentiate the cited works, another form of citation is used for this purpose.
The works are not cited collectively, but individually, i.e. in solo citation format. Compared with the nodding all round way of chorus citation, solo citation is a kind of "greeting by shakinghands" which is done one by one. (It is impossible for someone to shake hands with more than one person simultaneously, isn't it?) The citations in the sample sentence share a common formal feature: in passing, the writer point out the focus or feature of each cited work one by one. The syntactic expression involved is very short, typically a noun phrase, they are called "short solo citations", in contrast with "extensive solo citations" which will be addressed later (cf. page 16).
Both citing formats, chorus citation and solo citation, are used in literature introduction, and either form could be rewritten as the other. For example, the sentence (5) would read: 7. A number of studies have suggested the importance of metadiscourse (Schiffrin, 1980;Crismore, 1989;Crismore and Farnsworth, 1990…).
Similarly, (4) could be rewritten as short solo citation, by drawing the keywords from the title of the cited works, which would be:
It is interesting to note that the rewriting looks more awkward than the original, because of the extreme closeness and overlap between the last two studies. The original writer must have successfully taken this into consideration in his writing. However, a novice writer would fail to notice the subtle difference between solo citation and chorus citation and make a choice randomly. Therefore, an effective writer's choice of citing format for reviewing a particular group of works, either solo or chorus, is motivated, rather than arbitrary.
It is assumed that the default form of citation is a solo one. In the following discussion, the word "solo" is omitted unless otherwise indicated.

Run-in Short Quotation
In academic writing, there are occasions when it is necessary to quote others word by word. The purposes of direct quotation would range from "summoning the voice of an authority, preserving the source author's original wording, or emphasizing the accuracy of your borrowing from the source" (Slade & Perrin, 2008: p.59).
The quotations, in terms of their locative relation to the main text, can be run into or set off from the text. The distinction is based on the length of cited words, with the extended quotations (e.g. more than forty words in APA, MLA) set off from and short ones run into text. It seems that both short and extended quotation perform the same functions despite the length variation. However, according to our observation, the format differences of quotations correspond to different stylistic effect.
An extended quotation is put in the set-off block to highlight its other-sourcedness. Although there is no explicit rule that regulates the specific ratio of extended quotations in an academic writing; the suggestion is that direct quotations should be kept as short as possible. This means that extended quotations should be avoided unless its use is well justified. There could be strategic control or manipulation of the length of cited words.
Indeed, we find that the run-in text quotations can be further divided into two patterns: 9. Metadiscourse is the author's linguistic and rhetorical manifestation in the text in order to "bracket the discourse organisation and the expressive implications of what is being said" (Schiffrin, 1980: 231).
10. George Orwell, explains that "Dickens's criticism of society is almost exclusively moral." The quotation (the words enclosed in quotation marks) in (8) is not a grammatically complete sentence, but a sentence fragment. If a full sentence is run into text, it would give rise to a quotation in integral form, introduced in the pattern of either as verb controlling (e.g. Zhangsan argues that "…", as in (9)) or naming (According to Zhangsan, "…."). Generally speaking, a sentence fragment is not a full-blown proposition, so a quotation like (8) is called a "non-propositional quote", for the sake of differentiating it from other propositional quotes like (9). When a run-in short quotation is reduced as short as one or two words, it looks almost the same as a scare quote, but they are functionally different, since scare quotes are mainly used to express irony or sarcasm (Fairclough 1992: p.104).
The two patterns of citation (propositional quotes and non-propositional ones) are complementary to each other. When a non-propositional quote is incorporated in the text, they arrive at the combination of averral and attribution (signaled by the quotation marks). The status of cited author is suppressed and not allowed to appear on the stage, with his voice overheard from background, where the main text is analogous to the stage, and parenthetical reference to the off-stage background. In the "non-propositional quote", the quotation marks have an additional function, that is, signaling other-sourcedness, while in the case of propositional citations, which are always in the integral format (as in example 9), the function of signaling other-sourcedness, is shared by three markers: the naming of cited author, using reporting verb, as well as the quotation marks (c.f Section 2 on page 8, other-sourcedness). Therefore, a propositional quote is more explicitly marked than a non-propositional quote.
"Non-propositional quotes" sound less intrusive, so they are favored by many scholars. However, there is difficulty in incorporating sentence fragments, considering the risk of distorting the original meaning. The overall meaning of the output must be well-tuned, like having both authors (citing and cited) speak in the same tone. Syntactically, the quoted fragment should agree with the citing co-text. In this sense, non-propositional citation is a high-tech transplant of syntax and semantics. This writing tactic is hard to harness so it is less likely to be deployed by less proficient writers.
Another formal feature that distinguishes propositional quotes from the non-propositional ones is the location of sentence-final punctuation mark: the full-stop is located inside quotation marks for the former, but outside in the case of the non-propositional quotes.

Extensive Solo Citations
Solo citations, if it has no special markers, is divided into short and extensive citations. Short solo citations is functionally complementary to chorus citations (cf. p. 12 of this paper), an opposite of extensive citations formally. Swales (1986) made the distinction between "short" and "extensive" citations at sentence level citation that are at a single sentence level and those encompass more than one sentence (cited in Thompson & Tribble, 2001: 92). For our purpose, we redefine "short" and "extensive" distinction by setting the criteria at the phrase level: a citation involving only one phrase (typically NP) is a short one, and other are extensive.

Extensive Citations: Reporting vs. Non-Reporting
In fact, extensive citation is very complex. It is the hybridity of indirect quotation (paraphrase), hidden averral, explicit and implicit other-sourced attribution. Non-integral extensive citation is less explicit in the marking of attribution, because the absence of cited author's name (non-integral) means the failure of signaling other-sourcedness, and the length of text (extensive) may hinder the calibration of propositions. Within the extensive stretch of text, the citer's point and citee's point are mixed-up, so readers can hardly identify which is which. The indeterminate boundary between averral and attribution may lead to ambiguity. For example: Citation is central ... because it can provide justification for arguments (Gilbert, 1976) (cited in Thompson & Tribble, 2001). This citation has two propositions. There could be different interpretations: both or either proposition is attributed to the cited author. Furthermore, the omission between the two propositions is mysterious.
It's interesting to note that the same work (Gilbert 1976) is cited in another paper. Citation is central to the social context of persuasion as it can provide justification for arguments and demonstrate the novelty of one's position (Gilbert, 1976;Berkenkotter andHuckin 1995) (cited in Hyland 1999:342).
This citation is more complicated and gives rise to a greater ambiguity, which is difficult to resolve without referring to the original text of the cited works.
The repetitive use of obscure extensive citations is not acceptable in academic discourse. Swales (2014: 135) uses the term "parenthetical plonking" to refer to the style of overusing non-integral non-reporting extensive citations.
In non-integral citing forms, there are also other means to signal other-sourcedness, without mentioning the cited author's name. We suppose that reporting verbs (with its sentence patterns) can be used to signal other-sourcedness, besides opening an 'evaluative space' for writer comment (Thompson and Ye 1991: 369). Charles (2006aCharles ( , 2006b analysed the V that Patterns of reporting verbs, identified three types of clause according to the grammatical subject of the reporting clause: a noun group with human reference; a noun group with non-human reference; introductory it followed by passive voice. The following three exemplify non-integral citation form of each of the clause type (adapted from 2006a):

Some authors have suggested that… (V, 3)
A recent study (N, 1996) has found that… It has been shown that …. (S, 1994) These clause patterns of reporting verb, has the function of signaling other-sourcedness, by attributing the proposition to an indefinite source other than the writer, of course, less explicit than mentioning the proper name of cited authors.
We would suggest other marker of other-sourcedness, besides reporting verbs V that pattern.
11. Thus the term is used to refer to non-propositional aspects of discourse which help to organise prose as a coherent text and convey a writer's personality, credibility, reader sensitivity and relationship to the message (Crismore et al., 1993).
The passive construction, describing an author act, done "by some researcher" with an agent implied, which is linked to the author in the brackets. This linking relation is called "identification" by Thompson and Tribble (2001: 95).
In academic writing, there are many expressions that are alternative to V that clause, and are used to refer to other's work. Especially, appositive nominal clauses, e.g. The finding that ….; There is strong evidence to support the argument that …, etc. These are also markers of attribution.
In additional to the sub-types of non-integral citations that are mentioned above, there is a further type. A citation which is prefixed with a directing word, such as "see…" or "e.g." is called directive citation. It is less commonly used in academic writings, but performs some special functions (see Hyland 2002 for more details). The directives are formally outstanding, so are considered as markers. Directive citations, together with Run-in short quotations (see sample 8 on page 8) which is marked with quotation marks, are the non-integral citations with markers.
The complete typology of non-integral citations is summarized in the following figure (Figure 1).
It seems that all schemes of categorization are in free variation, and that nobody is sure about which categorization is better or more feasible. Following Lakoff (1987), we argue that different scheme of categorization vary in terms of the degree of naturalness, and there are natural dimensions for citation analysis.

NATURAL DIMENSIoNS oF ATTRIBUTIoN EXPLICITNESS
It is believed that a reporting clause is the arena for denotation and evaluation in academic texts. In Thompson &Ye's (1991) model of Layers of reports, the reporting process consists of six stages in reverse chronological order: Writer act (writes, evaluates, read) and Author act (writes, thinks, researches). The model handles clearly "the choice of whether or not to introduce explicit evaluation in the reporting verb itself." (Thompson &Ye 1991: 378) We would add another function, attribution, to the reporting clause, and suggest the variation of attribution explicitness, similar to the explicitness of evaluation in the reporting verb (Thompson &Ye 1991).
In the academic text, the proposition presentation varies along three dimensions on whether attributing explicitly to other's work: • Signaling other-sourcedness: The choice of indicating whether a particular proposition is not the writer's own. • Locating the source: The choice of whether the specific source of cited ideas is given in the text, besides the information in brackets. • Calibrating the cited proposition content: The choice of identifying whose point is who, and specifying the citee's specific point.
Let's take a look of the citation forms, with the variations along the three dimensions plotted on the chart below (Table 1).
As noted in the diagram, the values of each citation form on three dimensions are diversified, ranging from all positive to all-around negative. We put them into three level of attribution explicitness according to the values on the three dimensions:

CoNCLUSIoN
In academic writing, the writer has the choice of attributing or not, and also the choice of attributing explicitly or not. We find that the distinction between averral and attribution is blurred. Compared with the previous research that is limited to the use of reporting verb (Thompson &Ye 1991; Charles (2006) -that reporting clause), this study takes a global view of the broad context of citation. In this study, citations are not merely the parenthetical part of an academic text. Citation as source using consists of two components: proposition presentation + documentation. Our analysis reveals the interaction between various forms of citation and the interplay between proposition and documentation.
Based on the formal features of citations, we propose a new typology of non-integral citations, and categorize them into sub-types like Short, Extensive, Built-in Quote and others. For each sub-type of citations, we analyze the explicitness of attribution, in terms of the three dimensions of attribution explicitness: signaling other-sourcedness, locating the source, and calibrating the cited proposition content. It turns out that the sub-types of citations form a continuum of attribution explicitness, ranging from low to high.
Citing properly means choosing the right citation form to fine-tune the explicitness of attribution. In a text of high explicitness of attribution, the cited authors may sound intrusive while the writer becomes invisible and fails to harness the development of the argument. This could possibly explain why integral citations are scarcely used in academic writing. On the other hand, less explicit citations are preferred because they help the writer to keep the ñow of the argument uninterrupted (Hewings et al., 2010). In the EAP classrooms, it is insufficient to teach learners the forms and functions of citations mechanically. More importantly, the learners need to be taught the rationale and motivation behind the choices of citation forms. Genre-based approach is recommended to cultivate learners' awareness of readership and the flexibility of lexicogrammatical choices that are context-dependent.