INTERACTIVE METADISCOURSE MARKERS IN EFL MAJORS’ SUMMARIES IN ENGLISH

EFL majors’ academic training includes the development of academic skills, primarily academic reading and writing, which are important since English majors have to read many books and papers during their studies and often have to write papers on the basis of what they have read. This implies the students’ ability to decide which parts of the text are important and will be mentioned in the paper they are writing, as well as the ability to synthesize the material in a reader-friendly manner in accordance with the Anglophone academic tradition, which heavily relies on the use of metadiscourse markers that guide the reader through the text. In order to investigate to what extent EFL majors use interactive metadiscourse markers (Hyland 2005, 2010), which concern the writer’s awareness of a participating audience and address ways of organizing discourse, a research study was conducted with 59 English majors in their fifth year (MA level), who read a paper published in an academic journal and were asked to write a 250-word summary. The material was analyzed with the AntConc freeware and the results are used as a basis for pedagogical recommendations that aim at improving students’ training in academic writing.


INTRODUCTION
Students of English language and literature (henceforth EFL majors) are trained both at the BA and MA levels to write shorter academic texts such as essays, summaries or state-of-the-art papers or longer academic texts such as seminar papers, which prepares them for the final act of writing an MA thesis. During their studies they read vast amounts of text and often write brief reports or longer seminar papers either as class assignments or as parts of their final exams. This process relies on deciding which parts of the source text are important and will be mentioned in the paper they are writing and being able to synthesize the source material in a reader-friendly manner in accordance with the Anglophone academic tradition. In other words, EFL majors often have to summarize source texts and deduce them to just a few sentences, which implies a writing task that relies on outside sources, authors and ideas. The author of the summary is, therefore, a filter for a large quantity of information, because he/she needs to decide what will be summarized and in what order, what will be emphasized, etc. and this is facilitated by the use of metadiscourse markers that guide the reader through the text. For that reason, this paper investigates to what extent EFL majors use interactive metadiscourse markers in summaries by analyzing a corpus of 59 texts written by MA students and offers some pedagogical recommendations that aim at improving students' training in academic writing, because the hypothesis is that students use metadiscourse markers, but their variation is low.

SUMMARIES AS A GENRE
Summaries are one of the most important texts in academic writing because reading to write (cf. Krashen, 1984) is an essential part of the work of any academic community. More precisely, every author of an academic text relies on previously written and published sources, which he/she first reads and then decides what parts of the source text are important for his/her own research. While writing their own papers, writers undergo "the process of synthesizing and organizing individual idea units into a summary or organized series of related general ideas" (Irwin, 1986: 5). Summaries put into concise form the essential information or ideas of a source text 1 and the process of summary writing therefore follows a clearly defined set of steps: the source text is closely read; one-sentence summaries are written for each paragraph or section; the author uses his/her own words while summarizing; all notes are then edited for clarity, cohesion and coherence; hence, the final result is a new text which is different from the source text in terms of form, though conveying the selected content and meaning of the source text.
Summaries can be seen as a genre because "members of a community usually have little difficulty in recognising similarities in the texts they use frequently and are able to draw on their repeated experiences with such texts to read, understand, and perhaps write them relatively easily" (Hyland, 2007: 149). For instance, Gagich and Zickel (2017: 85) list three characteristics of a good summary: neutrality (there should be no evaluative language, writers of the summary should not express their own opinion nor should they make their presence known with the use of pronouns "I" and "we"), brevity (summaries only highlight the most important information from the source text in an accurate way) and independence (summaries should make sense to someone who has not read the source text, so there should be no confusion about the main content and its organization). Furthemore, Gagich and Zickel (2017: 86-87) state that summaries should be about 10-15% of the source text in length and should have an internal organization of the introduction (introduces the author and the source text), body (presents main ideas and arguments from the source text) and conclusion (restates the most important points for readers to remember after reading the summary).
The task of the writer of a summary is, then, to bridge a gap between the source text and the reader of the summary by carefully reading, selecting and rephrasing the source text thus mediating in the process of information transfer. The summary has the task of bringing the source text closer to the secondary audience and the author of the summary has the task of organizing the summary in a way which will help the reader understand and follow its main points. This is best achieved with an appropriate and adequate use of metadiscourse markers, which will be discussed and elaborated in the following section.

METADISCOURSE MARKERS
Every text is essentially an act of communication between the writer and the reader and in this dialogue the writer has the power to "guide a receiver's perception of a text" (Hyland, 2005: 3). Since this exchange of information is never entirely objective, and it actually includes "the personalities, attitudes and assumptions of those who are communicating" (Hyland, 2005: 3), many aspects of the process of communication are expressed via metadiscourse. This view on the language in use was put forth by Zelig Harris in 1959 and further developed by Williams (1981), Vande Kopple (1985), Crismore (1989), Hyland (2005), Ädel (2006), Hyland (2010) and Mauranen (2010) over several decades. Although different authors have different views regarding the relationship between propositional material and metadiscourse, they all agree that metadiscourse is broadly defined as the way in which the writer tries to guide the reader through the text and help him/her understand the text and the author's perspective (Bogdanović, 2017: 99). This is achieved by a whole range of metadiscourse features, whose classifications and taxonomies depend on the perspective that various authors assume in their analyses. In this paper the basis of the analysis will be Hyland's (2005) model of metadiscourse, which is "a more theoretically robust and analytically reliable model of metadiscourse, based on a number of core principles and offering clear criteria for identifying and coding features" (Hyland, 2005: 37).
Hyland's model is functional, which implies the idea that metadiscourse is context-dependent. In other words, one and the same word or phrase could be taken as a metadiscourse marker in one context, but not in the other, because in the former case this word or phrase has the function of a metadiscourse marker, but in the latter it does not. This model divides metadiscourse markers into two broad categories: interactive and interactional. The first category "concerns the writer's awareness of a participating audience and the ways he or she seeks to accommodate its probable knowledge, interests, rhetorical expectations and processing abilities" (Hyland, 2005: 49). The use of interactive metadiscourse markers ensures that the writer signals how discourse is organized and how he/she constructs it with the reader's needs, knowledge and experience in mind. The second category, interactional metadiscourse markers, is more concerned with the writer making his/her views explicit and involving readers by allowing them to respond to the text (Hyland, 2005: 49). More precisely, interactional metadiscourse is "essentially evaluative and engaging, expressing solidarity, anticipating objections and responding to an imagined dialogue with others. It reveals the extent to which the writer works to jointly construct the text with readers" (Hyland, 2005: 49-50).
As this paper focuses on the corpus analysis of students' summaries of a text they had previously read, it is clear that this type of texts produced by the students inherently requires the explication of the structure of the source paper and ideas presented there, which implies the necessity to use interactive metadiscourse markers. On the other hand, this task does not require students to make their views explicit as they are summarizing someone else's ideas and arguments, so interactional metadiscourse markers are not the subject of research presented here.
Interactive markers "are used to organize propositional information in ways that a projected target audience is likely to find coherent and convincing" (Hyland, 2005: 50) and they "anticipate readers' knowledge and reflect the writer's assessment of what needs to be made explicit to constrain and guide what can be recovered from the text" (Hyland, 2010: 128). When summaries are concerned, this actually means that the students first choose which information from the source article needs to be included in the summary (what they deem important for the reader) and then organize this selected set of arguments in a way which they assess the reader will best understand, i.e. they "manage the information flow to explicitly establish his or her preferred interpretations" (Hyland, 2010: 129). This is generally achieved with five types of interactive metadiscourse markers: transitions, frame markers, endophoric markers, evidentials and code glosses (cf. Hyland, 2005: 50-52).
Transitions are mainly conjunctions, but also adverbials which signify additive (e.g. and, moreover, furthermore, etc.), comparative (e.g. however, likewise, similarly, on the other hand, etc.) or consequential (e.g. therefore, in any case, nevertheless, etc.) steps in presenting arguments. Next, frame markers refer to text boundaries or elements of schematic text structure, hence they sequence (e.g. firstly, lastly, then, to begin with, etc.), label text stages (e.g. for the moment, in conclusion, in summary, on the whole, etc.), announce goals in the discourse (e.g. in this chapter, intention, objective, purpose, want to, etc.) and specify topic shifts (e.g. look more closely, with regard to, now, back to, etc.). Then, endophoric markers refer to other parts of the text, which allows the reader to recover the writer's meaning that may rely on a reference to the point already made or an announcement of a point yet to be made (e.g. in chapter X, Fig. X, X above, X below, etc.). Evidentials serve the purpose of indexing another source, or more precisely referring to literature from the field, which supports arguments and conclusions (e.g. to cite X, to quote X, according to X, etc.). Finally, code glosses supply additional information because they rephrase, explain or elaborate what the writer had already said, which serves the purpose of clarification for the reader (e.g. e.g., for example, in other words, namely, etc.).
When the genre of summaries is concerned, it can be clearly seen that its individual parts and aspects are covered by certain types of interactive metadiscourse markers. When the writer of the summary mentions the source text and its author, as well as any other authors he/she deems necessary, evidentials should be used. When the summary elaborates on the main ideas from the source text, the appropriate metadiscourse markers are transitions and frame markers. Finally, when the author of the summary thinks it is necessary to further explain ideas from the source text in case he/she assesses the reader needs further clarification, code glosses will serve this purpose. This paper investigates to what extent EFL majors abide by the principles of this genre when they write summaries, which is further elaborated and investigated in the sections that follow.

RESEARCH METHOD
The research for this paper was conducted on the corpus collected in the MA course Advanced Academic Writing and Research at the Department of English Studies, Faculty of Philosophy, University of Novi Sad in the academic 2016/2017. One part of the course trains students to read academic texts, identify main ideas and arguments and summarize them in order to incorporate them into their own papers and theses, most often in the section that deals with literature overview. In order to practice this skill and get feedback on their own production, the students were given an academic paper (Novakov, 2011) which they were supposed to read at home, single out the main points and write a 250-word summary. The source text that the students received did not have the abstract nor the summary from which they could adapt structure or phrases because those parts were deleted in the pdf file given to the students.
A total of 59 summaries were submitted with a total length of 13,200 words and an average length of 225 words per summary. All the summaries were then read, spelling errors were corrected and the texts were prepared for the two-step corpus analysis (cf. Hyland, 1998;Ädel, 2006): first the whole material was analyzed with the AntConc freeware for all instances of interactive metadiscourse markers listed in Hyland's (2005) appendix and then every concordance was manually filtered in order to separate metadiscourse functions of the words from Hyland's list from their other functions (e.g. the conjunction and was found to have 430 instances in the corpus, whereas it functions as a metadiscourse marker in 69 cases). Finally, after all interactive metadiscourse markers were counted and singled out in the corpus, their frequency per 10,000 words was calculated, as that is the standard procedure of presenting the occurrence of metadiscourse markers in texts.

RESULTS
Out of five Hyland's categories of interactive metadiscourse markers, four were identified in the material analyzed (code glosses, evidentials, frame markers and transition markers). The fifth category, endophoric markers, was not identified in the material because this type of metadiscourse markers refers to other parts of the same text (in chapter X, in section X, in part X, etc.); since the summaries were short texts without intricate structure, it was neither possible nor necessary for students to refer to their internal parts or segments.
The first category of interactive metadiscourse markers are transition markers, which are formally mostly conjunctions or adverbial phrases, but functionally "they signal additive, causative and contrastive relations in the writer's thinking, expressing relationships between stretches of discourse" (Hyland, 2005: 50). After the results presented in Table 1. below are analyzed, it can be concluded that there are a few metadiscourse markers that have a relatively high frequency (and, because, but, however, since, so), while most of the others have less than ten occurrences in the entire corpus. Frame markers as the second category analyzed in the corpus are of great importance for summaries in general since they order, sequence, label, predict and shift arguments (Hyland, 2005: 51). In the summaries they serve the purpose of sequencing the parts of the source paper in order to help the reader understand its internal structure, but the quantitative analysis reveals that there are actually very few instances when students use them (see Table 2). Hyland (2005: 219-220) further subdivides them into frame markers of sequencing, frame markers that label stages, frame markers that announce goals and frame markers that shift topic (separated in the table with dotted lines) and on the basis of the results in Table 2 it can be deduced that the students sequence the arguments in their summaries, but do not label stages, announce goals or shift topics as much. The third category of interactive metadiscourse markers that was analyzed are evidentials, which are "metalinguistic representations of an idea from another source" (Thomas and Hawes, 1994: 129). When the corpus analyzed in this paper is concerned, it seems that the students resorted to these discourse devices quite often (see Table 3). This can be explained by the fact that in their summaries they referred to the source paper or other sources it cited. Even though this is not always necessary in summaries, the students seemed to have applied what they had learnt in the course dealing with academic writing, where citing sources is an imperative and serves the purpose of avoiding plagiarism. The final category of interactive metadiscourse markers, code glosses, "supply additional information, by rephrasing, explaining or elaborating what has been said" (Hyland, 2005: 52), which in the case of the summaries from the corpus means that students offered alternative explanations in the parentheses or introduced examples from the source article to illustrate some of the theoretical claims made by its author (see Table 4). However, a closer comparison of the summaries from the corpus and the source text reveals the following: in most cases the students lifted phrases and parenthesized examples from the source text, which explains the relatively high number of the '( )' code gloss (41 per 10,000 words). As has already been mentioned, the category of endophoric markers was not identified in the corpus, which is logical because the short summaries that were analyzed did not have an intricate internal structure, parts of which could be referred to in the text. If the other four categories of interactive metadiscourse markers are observed in comparison (see Table 5), it can be seen that transition markers are the most numerous and are followed by frame markers and code glosses, with a low frequency of evidentials. The results presented in this section need to be further contextualized and compared with other research studies in order to determine to what extent the students from the sample abide by the standard conventions of the application of interactive metadiscourse markers in academic writing.

DISCUSSION
When Table 5 is analyzed, it can be noticed that transition markers make up over 50% of all interactive metadiscourse markers in the summaries, as well as that frame markers and code glosses have an almost equal share of around 20%. Because there are no other papers which analyze metadiscourse markers in this genre, precise comparisons cannot be made in order to observe how the writing of EFL majors who are non-native speakers compares to that of native speakers. However, since the source paper was in the field of applied linguistics, a comparison with Hyland's (2010) analysis of interactive metadiscourse in dissertations by discipline, particularly in the field of applied linguistics, will be presented first (see Table 6). Namely, Hyland (2010) has found that dissertations in applied linguistics have 265.9 interactive metadiscourse markers per 10,000 words, which is less than in the corpus of summaries, which has 328.03. Furthermore, Hyland's transition markers and evidentials appear to almost the same extent (95.1 and 82.2 per 10,000 words), which differs from the analyzed summaries (169.70 and 23.48 per 10,000 words). This could, of course, be explained by the differences in genre, because dissertations are by all means expected to cite other sources and do that frequently, which is not really the case with summaries. Another noticeable point of difference is the frequency of frame markers, which is much higher in the corpus of summaries (71.97 vs. 25.5). This could be explained by the content of the courses EFL majors took during their studies, which insist on the use of transition and frame markers in general, but also by the genre, since summaries condense a lot of information in a short text and therefore need to be "reader-friendly" in terms of ordering and sequencing information that can be found in different parts of the source paper. Finally, code glosses are again more frequent in summaries than in dissertations (62.88 vs. 41.1) and a closer look at Table 4 reveals why. Namely, 65% of all code glossed that EFL majors used in their summaries are parentheses '( )' which further explain or exemplify information from the source paper, while all other code glosses are very infrequent. All these differences could also be explained by other factors such as the length of the texts in the corpus (very long dissertations vs. very short summaries), but also by the differences between native and non-native writers, which should definitely be further explored in another paper.
However, what is more interesting in the results of the analysis is the uneven distribution of particular interactive metadiscourse markers within each subcategory. Like already noted for transitions, markers like and, because, but, however, since, so have a relatively high frequency, while instances of other markers (whereas, thus, still, etc.) have just a few occurrences in the corpus. The same could be noted for other categories of interactive metadiscourse: the most frequent frame markers are first, then and aim, but second and third, or goal and focus appear just once or twice in the entire corpus; in the category of evidentials (name) (date) is much more frequent than according to; and finally, in the category of code glosses parentheses '( )' outweigh all other code glosses combined. Lack of variety of interactive metadiscourse could be illustrated with an excerpt from a summary in the corpus: Some terms were not even adapted into Serbian (low fat) and the usage of borrowed words was often unjustified because of the existence of the domestic terms (implementacija -sprovođenje), and the borrowed words were not always understood. Afterwards the terminology started to be integrated into Serbian and the terms were naturalized.
As it can be seen, in the two sentences from a summary that have a total of 51 words the transition marker and was used three times even though it could have been replaced by another marker with some syntactic changes. This feature of EFL majors' summaries is considered to be poor academic writing, which is characterized by the lack of variety of metadiscourse markers (Intaraprawat and Steffensen, 1995: 265). What can also be noticed about the excerpt above is that it resembles spoken language with a repetitive use of coordination and lack of subordination, which is another feature of poor ESL writing that Intaraprawat and Steffensen (1995: 262) detected. These findings imply that there is a need to pay more attention to metadiscourse markers in courses that teach academic writing to EFL majors and other non-native speakers, so some of the principles that could help improve their skill are outlined in the section that follows.

PEDAGOGICAL IMPLICATIONS
If summaries are taken to be a genre, then it is logical that they could be the subject of a genre-based approach to teaching writing. This method implies a focus on two different sides of summaries: the source text and the summary itself. As some research studies (Kongpetch, 2006;Cheng, 2008;Chen and Su, 2012) have proven, if students understand how a source paper is structured, they could more easily distinguish between important and unimportant points, which will later help them structure their summary. According to Widodo (2006), the genre-based approach to writing requires students to (1) understand why they are writing a text, i.e. focus on its purpose; (2) ask themselves who they are writing for, i.e. who the audience will be; and (3) think about how to write a text, i.e. what its organization should be like. When summaries are concerned, students should be taught that they are writing a summary in order to show their understanding of the source text and present in short form its main points and their relevance for students' own research if the summary is part of an MA thesis. They should also discuss with their teacher of academic writing what kind of audience they are writing the summary for and it should be made clear that the audience is usually an expert in the field. Finally, when discussing organization, they should pay attention to the distinction between important and less important aspects of the source text and decide how they are going to present a selection of important points in their own text. The results of the research presented in this paper could be of great use in this case. Hyland (2007) describes a teaching-learning cycle of a genre-based approach, which could be applied with some modification to teaching EFL majors how to write a summary. The first stage is setting the context, where students explore the purpose of summary writing. This is followed by modelling, where students analyze discourse features of a sample summary, which should definitely put an emphasis on metadiscourse markers and possible variations in their use. Then students do controlled tasks with teacher guidance in order to learn and practice the organizational and grammatical features of the genre such as markers, which is followed by independent tasks without teacher control. Finally, students collaborate in pairs or groups and compare their products, do peer assessment and state what they had learnt in the process and what aspects still need improvement. When this cycle is repeated enough times, students will gradually acquire the conventions of this genre and hopefully increase the skill of using adequate and varied metadiscourse markers in their own writing.

CONCLUSION
Using a corpus-based approach, this paper has set out to analyze how EFL majors use interactive metadiscourse markers when writing summaries of a paper from the field of applied linguistics. The analysis of 59 summaries has established that students used four categories of interactive metadiscourse and that more than half of overall markers were transitions. Furthermore, the analysis has established that despite relatively high frequencies of markers in the corpus, especially in comparison to Hyland (2010), the variation of markers is low, i.e. students used just a few markers very frequently, while the majority of other markers were essentially underrepresented, which confirms the initial hypothesis. This has established ground for pedagogical improvements in the courses that tackle academic writing, especially courses that prepare students for writing their MA theses. What remains to be investigated is an idea mentioned several times in the papers, related to the students' ability to distinguish important and unimportant information, which has a direct influence on how summaries are structured and which metadiscourse markers are used. However, since that falls well beyond the scope of this paper, it could be the topic of a future research study.