A Review of Recent Metadiscourse Studies: The Iranian Context

While studies of metadiscourse in a European or US context have been conducted since the 1980s and have reached a relatively wide audience, studies of metadiscourse outside of these areas have a more recent history (if any) and have not attracted much attention. One country in which research into metadiscourse has gained ground in the past decade is Iran. The first author has helped to bring this about and the second author is one of the active researchers into various aspects of metadiscourse in Iran. The purpose of this paper is to provide a review of studies of metadiscourse that have taken place in the Iranian context. These have been carried out by master‘s and doctoral-level students in the past decade. A close observation of these studies reveals three areas of metadiscourse-related studies: metadiscourse use in writing in English, cross-linguistic comparison of metadiscourse in English and Persian, and metadiscourse in EFL reading comprehension. Studies related to each area will be reviewed and evaluative summaries at the end of each section will be presented. The implications for metadiscourse research and instruction will also be discussed.


Introduction
Metadiscourse is discourse about discourse, intended to direct rather than inform readers (Williams 1981). Metadiscourse includes linguistic elements which do not refer to aspects of external reality (as propositional or referential elements do) but to the organization of the discourse itself and to aspects of the relationship that develops between the author and the reader (Crismore 1989;Vande Kopple 2002). What Vande Kopple labels ‗referential' meaning is equivalent to what Halliday (1978) calls ‗ideational' meaning. Vande Kopple (1985), using the broad definition of metadiscourse, suggests that metadiscourse conveys interpersonal and/or textual meanings. Interpersonal metadiscourse -helps writers express their personalities, their evaluations of and attitudes toward ideational material, shows what role in the communication situation they are choosing, and indicates how they hope readers will respond to the ideational material‖ (Vande Kopple 2002: 2-3). Textual metadiscourse helps writers relate and connect bits of ideational material within a text and helps the text make sense in a particular situation for readers. Crismore, Markkanen & Steffensen (1993) point out that interpersonal and textual functions are important from the point of view of teaching composition, and they use the term metadiscourse to refer to linguistic items that explicitly serve the interpersonal and textual functions of language.
In the broad definition, metadiscourse is based on a view of writing as social engagement in which writers project themselves into their discourse to signal their attitudes and commitments (cf. Hyland 2005). Metadiscourse elements are rhetorical tools that make a text readerfriendly and as such enable the writer to reach the audience. Research over the past two decades has shown that the use of metadiscourse in writing may vary from one language and culture to another and that the conventions followed in its use may be different in different cultures (Abdollahzadeh 2003;Crismore et al. 1993;Mauranen 1993). There is also burgeoning research on both the role of metadiscourse presence in text comprehension and its instructional impact on reading and writing (see Section 4 below).
This paper provides an overview of the research on metadiscourse that has been undertaken in the Iranian context. In what follows, the studies of metadiscourse carried out by master's and doctoral-level graduate students in Iran during the last decade will be examined and discussed. The graduate students who designed and carried out these studies have all been in contact with the first author concerning their work through e-mail correspondence and conferences. The studies are categorized into three areas depending on their approach and topics: metadiscourse in writing in English; cross-linguistic comparison of metadiscourse in English and Persian; and metadiscourse in EFL (English as a Foreign Language) reading comprehension. We present evaluative summaries and relevant discussions related to each area and discuss pedagogical and research implications. Table 1 summarizes the studies carried out on the use of metadiscourse in writing in English, for which both professional writers and student writers have been in focus.  Table 2 summarizes the studies that have been carried out on the use of metadiscourse in both English and Persian texts, involving a comparison of the two languages.  Table 3 summarizes the experimental studies that have been carried out on metadiscourse related to reading comprehension or the impact of instruction on metadiscourse. The ensuing sections will be organized according to Tables 1 to 3 above.

Metadiscourse use in writing in English
In the area of representing the use of metadiscourse in texts in English, a study by Abdi (2000) examines interpersonal metadiscourse following Vande Kopple (1985). The interpersonal metadiscourse categories in the discussion sections of sixty research articles in English from social science and natural science journals published in 1999 are examined.
His corpus was approximately 80,000 words in total-half the words from the social sciences and half the words from the natural sciences. The interpersonal metadiscourse categories of hedges (modal verbs such as might and would and words such as likely, suggest, possibly), emphatics (words such as strongly, definitely, clearly, truly), and attitude markers (such as unfortunately, surprisingly, it is noteworthy) were examined. His quantitative and qualitative analysis demonstrated that some metadiscourse expressions have different functions depending on the context. One of the main results was that writers used emphatics not to show arrogance, as suggested in some literature (Vande Kopple 1985), but to reveal their limitations and show humility, as in this example: -Quite clearly, this single study is not sufficient to demonstrate that...‖. Hedges were used by these writers to discuss their findings and what the findings denoted and implied. Abdi also found that emphatics were used almost as frequently as hedges, the most frequently used of the three types of interpersonal metadiscourse he studied. Beighmohammadi (2003) examined the extent to which the use of intensity markers varies across three domains: the hard sciences, social sciences, and TEFL. Seventy-five randomly-selected introductions from prestigious journals were selected. He employed the Quirk et al. (1985) model for intensity markers. He found that social science writers used twice as many intensity markers as hard science writers. The TEFL writers' performance was similar to that of hard science writers. He argued that social science writers depend more on discursive and rhetorical strategies in presenting their findings rather than on the mere reporting of facts.
To examine the impact of metadiscourse knowledge and use on student writing, Simin (2004) investigated the metadiscourse used in the writing of ninety undergraduate Iranian EFL learners. The students were divided into upper-intermediate, intermediate, and lower intermediate proficiency levels. For a period of one semester, their sample essays, written on argumentative topics assigned to them, were collected and analyzed using Vande Kopple's (1985) model. The proportion of appropriate uses of metadiscourse was counted across the given tasks. Significant differences were found in metadiscourse use across different levels of proficiency. Proficiency level was found to affect the use of metadiscourse; the more proficient the learners were, the more they used metadiscourse in their writing. All students in the three proficiency groups used both textual and interpersonal metadiscourse in their argumentative writing. The upper-intermediate group used far more metadiscourse than the intermediate and lower-intermediate groups.
However, Simin noted that the three groups were similar regarding the variety of metadiscourse types used. All groups used more textual than interpersonal metadiscourse, and text (logical) connectives were the most frequently used textual metadiscourse subtype. Another finding from this study was that there was some improvement in the use of metadiscourse during this one-term period of writing instruction. Thus, Simin concludes tentatively that writing instruction had a positive effect on the use of metadiscourse represented in her data.
Simin's study is interesting as it looks into the use of metadiscourse markers in more persuasive discourse genres such as in argumentation. Simin's findings confirm the predictions of Williams (1981) that argumentative writing lends itself well to the use of metadiscourse but disconfirmed his predictions about the use of the interpersonal type, which allows writers to make their ethical, logical and emotional appeals. On the other hand, Simin's results may indicate a distinction between the way novice and professional writers project themselves into texts to establish more interactional persona with their readers. We see more interpersonal metadiscourse in professional writing. Consequently, it is important to analyze the texts of professional writers from various countries, comparing their metadiscourse use to those of inexperienced writers.
One misconception with reference to the use of metadiscourse may be ‗the more metadiscourse use, the better'. Overuse or misuse of such markers can make the text long-winded and clumsy, which may be a sign of poor writing. Excessive use of metadiscourse can be as disadvantageous as a limited use or no use of such expressions since they may interfere with the reading process and may look imposing and condescending (Rahman 2004). Like many other rhetorical devices, metadiscourse can be used both effectively and ineffectively. Therefore, pedagogically speaking, we need to teach all types of metadiscourse rhetorically not as a panacea (Crismore et al. 1993). The increased use of metadiscourse by learners cannot by itself be a sign of language development.
One of the main issues with the metadiscourse studies reported here is that researchers have adopted different models of metadiscourse (e.g., Hyland 2004;Vande Kopple 1985) as their point of departure. The advantage of this situation is that we can get a more elaborate spectrum of these meta-communicative markers. However, the different models make the comparability of the results more difficult, especially when we compare data from different genres, registers, cultures and disciplines. This problem is prevalent in most available published research on metadiscourse.
Another potential problem in the study of metadiscourse is that most comparative studies dealing with native vs. non-native writing conventions consider US and British conventions as similar in terms of their argumentation patterns and rhetorical conventions. However, recent corpus-based studies demonstrate that cultural conventions may differ even within the English-speaking world. For instance, Ädel's (2006) comparison of the argumentative writings of American and British writers demonstrated significant differences in terms of personal metadiscourse use. Similarly, Precht's (2003) study of stance differences between American and British English conversations showed less frequent uses of emphatics and emotive, affective markers by British than US speakers. These results have serious implications for second language writing, especially ESL/EFL composition courses. The pedagogical implication is that we need to make decisions as to which English ‗norm' for metadiscourse use we should adopt and teach in L2 composition courses for non-native speakers of English. Research in this area, focusing more on intercultural than intracultural differences, is in its infancy however and further research is needed.
In addition to intracultural rhetorical differences in written discourses, it is important to note such differences within disciplines. Applied linguistics as a discipline has grown substantially in terms of its domain, subdisciplines, and research frontiers. Therefore, experts practicing in subdisciplines such as computational linguistics, discourse analysis, language testing, pragmatics, TESOL, etc. may have different priorities and rhetorical norms which could vary depending on the size of their discourse community, the gatekeepers in that community, and how ‗conventionalized' the generic practice is (Swales 1990). Therefore, corpus-based studies in general and metadiscourse studies in particular need to take into account these intradisciplinary variations in rhetorical practice that may affect the results of studies and comparisons. Thus, selecting an applied linguistics corpus (as in Rahimpour's study for instance) without controlling for such intradisciplinary variations may confound the validity of the results and comparisons. Careful corpus selection in this regard is needed.

Cross-linguistic comparison of metadiscourse in English and Persian
The first study in this section is by Marandi (2002) who investigated the use of metadiscourse in the introduction and discussion sections of 30 master's theses written after 1990 by Persian-speaking and Englishspeaking graduate students. She compared three sets of texts: (a) texts by British English writers, (b) texts written in Persian by Iranians, and (c) texts written in English by Iranians. Marandi analyzed the first 1,000 words in each introduction and discussion section of the master's theses to determine the amounts and the subtypes of metadiscourse that the graduate students used. She used her own model of metadiscourse developed from different established models. She found that textual metadiscourse subtypes were used significantly more in the introductions but that interpersonal metadiscourse subtypes were used more in the discussion sections. In addition, the results showed that, of all groups, the native speakers of Persian used text/logical connectors the most while the native speakers of English used them the least.
Using a model of metadiscourse from Crismore et al. (1993), Azizi (2001) looked at the use of interpersonal and textual metadiscourse in the English and Persian writings of Iranian university students. A set topic (-What should be done to increase the quality of education?‖) was rated by 106 upper-level EFL students as the most popular topic from among several given topics, and then the students were asked to write at least 150 words on the topic both in English and in Persian. Forty-eight papers (24 in English and 24 in Persian) were selected through various judgment procedures and reviewed for metadiscourse use by independent raters. The corpus amounted to 6,000 words. Participants used more textual markers in Persian and more interpersonal markers in English. Attitude and commentary markers were used significantly more in their English writings while the use of hedges, emphatics and text connectives were significantly higher in their Persian essays. Azizi argues that these participants' English language learning experience and their awareness of rhetorical preferences of the foreign language compelled them to produce more interpersonal markers while writing in English. Thus, their English learning probably impacted their English thinking process and thus helped them to develop a second identity while writing in a second language.
Abdollahzadeh (2003) investigated whether there was any significant difference between Iranian and English academic writers in their use of interpersonal metadiscourse and its relevant subcategories in the discussion and conclusion sections of ELT papers. Applying a model from Vande Kopple (2002), his purpose was to find the extent to which academic writers project themselves into texts to assert their personal involvement and how they accomplished this projecting. The materials randomly selected for the study were 65 articles (32 articles by native speakers of English and 33 by Iranian academics writing in English) published during the years 2000-2002 in the field of English Language Teaching (ELT). The research focused on the subcategories of hedges, emphatics, and attitude markers in these materials as it was assumed that the discussion and conclusion sections have a greater possibility of using interpersonal metadiscourse and author projection in them. The results showed a statistically significant difference between native and nonnative writers in their use of interpersonal metadiscourse. Anglo-American writers used more (56%). There was no significant difference in the use of hedges, but Iranian academics used more (65%). On the other hand, the Anglo-American writers used more certainty and attitude markers than the Iranian academics.
In a similar study, Abdollahzadeh (2001) examined the use of textual metadiscourse in the introduction sections of ELT papers by Iranian and Anglo-American academic writers. The subtypes of text connectives (including logical connectives and sequencers), code glosses (the subtype which helps readers understand the meaning of discourse elements, e.g., what I meant to say, or in other words) and illocution markers (elements which make explicit for readers what specific action the writer is performing in the text, e.g. to sum up, we claim that, I argue that) were examined in these papers. He selected 73 introductions (37 articles written by native English writers and 36 by Iranians). For purposes of rater consistency a panel of raters (MA and Ph.D. graduates) with nativelike proficiency and sufficient knowledge about the function of each metadiscourse instance reviewed the corpus. Abdollahzadeh found that the native Anglo-American writers used significantly more textual metadiscourse (54%) than their Iranian academic counterparts (46%). Thus the Anglo-American texts provided more guidance to readers. Both groups used more text connectives than code glosses and more code glosses than illocution markers. The non-native writers used a few more text connectors than the native writers, who used more code glosses and illocution markers than the non-native writers.
A study by Rahimpour (2006) focused on metadiscourse use in the discussion sections of 90 (British and US) English and Persian applied linguistics research articles. Her assumption was that, due to differences in cultural values, the metadiscourse use in these two languages would be different. The discussion sections of the articles were selected from three groups: those written in English by Iranians as non-native speakers of English; those in Persian written by Iranians; and those written by native speakers of English. The researcher selected 30 discussion sections by each group of applied linguistics writers published between 1998 and 2005. The study used metadiscourse sub-types adopted from Hyland's (2004) model which consists of textual metadiscourse (the subtypes include transitions, frame markers, endophoric markers, evidentials, code glosses) and interpersonal metadiscourse (the subtypes include hedges, boosters, attitude markers, engagement markers, and self-mentions). Also see Hyland (this volume) for further details. Rahimpour found that writers of all three groups of applied linguistic discussion sections used all sub-types of metadiscourse. Transitions and hedges were the most frequently used subtypes. Native speakers of English used significantly more textual metadiscourse than the two groups of Iranian writers did. Furthermore, textual metadiscourse was used significantly more than interpersonal metadiscourse by all groups. She argues that teachers must teach students how to identify metadiscourse and then use it for different audiences and genres. She also argues that teachers themselves must learn more about metadiscourse use in different disciplines and contexts.
Abdollahzadeh (2007) studied the use of metadiscourse in 53 Persian and English (British and US) newspaper editorials in order to see how writers in different languages and cultures tone down and organize their writings in their attempt to gain solidarity and community acceptance. Based on Vande Kopple (1985), he examined instances of hedges, emphatics, attitude markers, person markers (I, my, our, we, etc.), text connectives and code glosses, to ascertain if there were significant differences in the use of these subtypes of metadiscourse. Twenty-six editorials (16,144 words) by Persian-speaking editors and columnists and 27 editorals by English-speaking editors (16,190 words) were examined qualitatively and quantitatively. The newspapers were selected randomly from March to June 2003 issues from Iran, the United States and Britain. The results demonstrated no significant difference between Persian and English editorials for the metadiscourse subtypes of text connectives, attitude markers, and person markers. However, significant differences were found for the subtypes code glosses (more were used in English editorials), hedges (English editorials used more) and for emphatics (Persian editorials used more). According to Abdollahzadeh, the heavy use of emphatics by the Persian editorial writers was due to an Iranian tradition of valuing and abiding by the rules of those in power without questioning them or without expressing doubt or uncertainty about social and, specifically, religious issues. The heavy use of hedges by the English editorial writers was ascribed to their being more considerate and polite to their readers. The significant use of code glosses by the English editorial writers was believed to show a reader-oriented attitude. It is concluded that not all cultures sanction the same degree of author projection and author presence in order to be persuasive and that metadiscourse use is influenced by personal, interpersonal, institutional, and socio-cultural factors.
An important finding of the studies in this section is that Iranian writers and academics tend to use more textual than interpersonal markers, while their Anglo-American counterparts tend to use more interpersonal markers. Further, the significantly more frequent use of textual glossing and illocution markers by Anglo-American writers can imply a more writer-responsible tradition among these writers in comparison to the apparently more propositional-oriented, readerresponsible Iranian writers (see Hinds 1987 for the origin of this distinction).
The Iranian corpus-based studies of metadiscourse have mainly analyzed academic research articles. None the less, analysis of metadiscourse in other genres such as books, manuscripts, and nonacademic promotional genres from a cross-cultural perspective would broaden our knowledge of the extent and role of this rhetorical device. One of the strengths in Iranian research is the examining of metadiscourse use in theses and dissertations. Theses and dissertations are less competitive than research articles and may be less analyzed and less studied regarding metadiscourse use. It would be interesting to compare the more competitive and promotional genres with the less competitive ones such as theses and dissertations with respect to the use of a rhetorical device such as metadiscourse.
Awareness of audience and purpose pushes writers to be rhetorically more effective, particularly so for authors publishing in leading journals, given the high-stakes nature of article publication and the critical stance of the readers. Therefore, it seems that Anglo-American writers publishing in leading international journals need to create more forcefully a research space for themselves in order to persuade an expert audience of a new interpretation or need to anticipate the consequences of being proved wrong. These situations may account for the significant use of interpersonal metadiscourse found in Anglo-American writing. On the other hand, writers publishing in local journals may not need to compete for a research space because of the much smaller size of the discourse community and the decreased possibility of audience rejection. Consequently, the status of the journals (local vs. international) can be another reason for the significant use of interpersonal language in order to gain community acceptance and solidarity with their audience by English writers.

Metadiscourse effects on students' reading comprehension
The third area of metadiscourse studies undertaken in Iran has to do with the role of metadiscourse presence in texts and its explicit instruction in reading comprehension. Dastgoshadeh (2001) investigated the question of whether there were positive effects of metadiscourse use on the reading comprehension of EFL university students with high and low levels of English language proficiency. He selected his subjects from different genders, ages, and religions. In appropriate places, he inserted a variety of different subtypes of metadiscourse into a reading passage, on an unfamiliar topic. Dastgoshadeh found that students at both high and low levels of English language proficiency used metadiscourse to comprehend the passage more effectively. English language proficiency was a powerful factor regarding the degree of comprehension achievement. An interesting implication of studies of this kind is the need for further research to examine the percentage contributions of textual and interpersonal types to reading comprehension across different language proficiency levels.
In a similar study, Daftary Fard (2002), taking account of all the theoretical views on 24 different reading skills, tried to find if there were any implicational relationships among those skills. Among them were skills relevant to metadiscourse: guessing; interpreting cohesive devices; understanding the source of the text; understanding the opinion of the author; text organization; and choosing the main idea of the text. She gave several reading tests to 650 Iranian students of varying reading abilities. She used expository, descriptive, and instructional texts in order to come up with a model of reading comprehension. She noted that recognizing and understanding metadiscourse is one skill among many others that a reader should have in order to be called an effective reader. She found that reading comprehension is not a general reading ability but a multidimensional construct and that metadiscourse knowledge and use is part of this multidimensional reading skill model advocated.
Another study is by Khorvash (2008) who investigated the differential impact of explicit instruction of types of metadiscourse on Iranian EFL learners' achievement in reading comprehension. She used four groups of Persian learners of English (three experimental groups and one control group) as the participants in her study. All 80 students (20 in each group) were at the intermediate level of English in a language institute. The first experimental group received instruction in both textual and interpersonal metadiscourse; the second, instruction in only textual metadiscourse; and the third, instruction in only interpersonal metadiscourse. The comparison group received no instruction in metadiscourse, only relevant exercises for reading in general. Analyses of the post-tests revealed a positive effect for instruction in metadiscourse. The findings clearly showed that the types of metadiscourse do not similarly affect learners' reading comprehension. The first experimental group (the one that received instructions in both textual and interpersonal metadiscourse) did much better than the other three groups on the post-test. The second experimental group performed better on the post-test than the third experimental group, and the third experimental group preformed better than the control group, which scored the lowest on the test.
Along the same lines, Jalilifar & Alipour (2007) examined the impact of metadiscourse presence and instruction on TOEFL reading passages for three groups of students with pre-to intermediate reading proficiency. One group received the original passages; the second, the same passages with metadiscourse removed (otherwise unmodified), and the third group received the more coherent, modified metadiscourse-free version. The significant result was that -the omission of metadiscourse markers from a text does not hinder the comprehensibility of the propositional content presented in the text, once that enough structural modifications are made in the text‖ (Jalilifar & Alipour 2007: 43). On the other hand, performance on the original texts was significantly higher than that on unmodified texts (i.e. the ones from which metadiscourse ties were removed without making any other changes to the text). Meanwhile, the group which had received the modified texts was explicitly instructed about metadiscourse, and this group outperformed the other two groups on the post-test. The explicit metadiscourse instruction was argued to have helped participants ‗notice' and become aware of these language forms and their functions while reading. The removal of these markers broke the propositional chains in the texts and thus made them confusing.
Similarly, Parvaresh (2008, later published under Nemati & Parvaresh, 2008 investigated the effect of metadiscourse on the comprehension of texts in both English and Persian. Hyland's (2005) model of metadiscourse was used. The research attempted to find whether there was a significant difference between the comprehension performance of Iranian EFL learners on the English texts and their translated Persian versions with and without metadiscourse in them. Parvaresh also tried to examine the participants' awareness of the metadiscourse used and their interactions with those texts in both languages by using a follow-up questionnaire. Based on an original English text, a set of true/false questions were given about both the English text and the translated Persian text. The EFL learners were limited to higher and lower intermediate learners in language institutes. The results indicated that both higher and lower level EFL learners performed significantly better on the texts with all the metadiscourse items left in than on the texts with removed metadiscourse items. Thus, lower proficiency EFL learners might benefit more from the presence of metadiscourse in texts. His questionnaire results also suggested that when Iranian EFL learners have problems understanding a text (whether English or Persian), it is the presence of metadiscourse which can help them both comprehend and remember the propositional content of the text more effectively.
Amiri (2007) examined whether metadiscourse consciousnessraising had any significant effect on Iranian EFL learners' improvement of writing skill. Sixty senior university students majoring in English literature served as subjects. The subjects were enrolled in two classes and, at the outset of the study, were given a TOEFL test to determine whether they were homogeneous. In the second class session, all students (the experimental and control classes) as a pretest wrote an essay about knowledge and power. For seven weeks the teacher used sample texts (e.g. from Vande Kopple's 1985 study), lectures, and exercises to make the experimental group aware of the role and function of metadiscourse in writing. The control group read and did assignments from a textbook on general composition during the seven weeks. The experimental group wrote essays for each class at home, some of which the teacher discussed in class. At the end of the seven weeks, a post-test was given to both groups, using the same topic as in the pretest. The results showed that the experimental group benefited from the metadiscourse consciousness-raising and produced essays that received significantly higher grades than those in the control group. In the experimental group, the essays appropriately used metadiscourse, which made the texts more accommodating for readers. Amiri argues that metadiscourse is an effective rhetorical device for writing because it combines a readercentered approach with a text-centered approach by giving adequate attention to the text.
The above-mentioned findings on the role of metadiscourse in reading comprehension reflect the significant impact of the presence of metadiscourse markers and instruction of these markers on reading comprehension for different language proficiency groups. However, determining language proficiency levels is a major problem in most such studies as there are few standardized tests available to determine proficiency levels. Further, these studies have employed different language proficiency tests, such as the TOEFL test, the Oxford Placement Test, and the Comprehensive English Language Test, which makes comparison of the results of such studies somewhat difficult, especially when we make cross-level comparisons. Generally, all of these metadiscourse studies represent a broad picture of the cumulative impacts of these markers or lack of them on reading performance. What seems to be missing in these studies is the impact of each of the subtypes of these markers (e.g. textual and interpersonal subtypes) on reading comprehension. That is, what these researchers consider as textual or interpersonal metadiscourse, or their subtypes, is not fully clarified: for instance, whether they included visual metadiscourse, capitalization, and circled words as instances of metalanguage is not sufficiently dealt with. Furthermore, the percentage contribution of each of the subtypes of metadiscourse to reading comprehension could have been examined by using more rigorous research designs and procedures. A significant finding of the studies in this section, however, is the relationship between text manipulation through inserting or removing metadiscourse markers and reading performance. It appears that the removal of metatextual markers with proper modifications such that the text's coherence is not ruined, as shown in Jalilifar & Alipour's study (2007), will not significantly affect reading comprehension. This finding can have significant implications in terms of the relationship between metadiscourse use and textual coherence. To what extent metadiscourse manipulation can affect (in the readers' minds) textual coherence on the one hand and cognitive coherence on the other hand, is yet an open question. More research is needed to find out, through different tasks (free recall, written recall, summary, etc), in what way coherent or minimally coherent texts with or without metadiscourse affect the comprehension of readers with little knowledge of the domain of the text, or vice versa. In addition, research is needed to discover to what extent this coherence manipulation would affect the text-based or situational understanding of readers.
A related issue is the relationship between readability and metadiscourse. Dictionary definitions of readability define ‗readable' as ‗interesting', ‗easy to read', and ‗legible' (Neufeldt & Guralnick 1991). None of the above studies have examined the impact of metadiscourse insertion and/or removal on the readability of the texts and consequently its relation to reading comprehension. Readability formulas provide a quick, easy, and practical way of estimating the difficulty of a text, focusing on word difficulty and sentence length. The goal of readability measures is to find out the best match between readers and texts. Nonetheless, the point is that metadiscourse signaling makes sentences longer and consequently affects readability scores. However, metadiscourse signaling can ease the difficulty of a text for readers. Readability formulas seem to ignore the degree of vividness, exposition, organization, and writer presence in the text and the interactions of these factors with the reader (de Beaugrande & Dressler 1981;Meyer 2003). Therefore, a comprehensive readability formula, among other things, needs to consider metadiscourse variables and their impact on text difficulty.

Conclusion
We have reviewed fifteen studies of metadiscourse by Iranian graduate students. No doubt more metadiscourse studies in Iran are in progress at the time of writing. These studies demonstrate differences in methodology (experimental vs. descriptive), approach (contrastive, causal-comparative, etc.), and research questions. These differences, of course, make generalizations somewhat difficult. However, when we relate the results to the broader picture of metadiscourse research, we can offer more constructive comments, especially from a cross-cultural rhetorical perspective.
All of the studies reported here, save for Abdi's (2000) and Beighmohammadi's (2003) studies, fall within the domain of contrastive rhetoric, i.e.they focus on cultural differences in textual preferences. They deal with texts from so-called ‗soft' sciences such as applied linguistics, social sciences, political sciences, ELT, etc., and have been mainly concerned with academic writing at the undergraduate, graduate, and professional levels. In their quest for new knowledge, they have employed both quantitative (frequency counts) and qualitative (functional-contextual analysis) approaches to text analysis which, in fact, adds more value to such studies. Nonetheless, one main consideration of these research studies is the extent to which the functional contextual analyses that were done are reliable. When dealing with cross-cultural analysis of such data, consistent coding is extremely important. This requires multiple raters and analysis over time (Crismore et al. 1993). In most of the reports, the main researchers and some graduate students that the researchers trained did the coding of the metadiscourse items. It would be easy to say that they might have coded the way that they did in order to see what they wanted to see. Therefore, it is recommended that future studies of this kind take into account more seriously multiple ratings and rating over time. The views of different raters need to be solicited and the degree of consistency in their analyses need to be reported. This is especially important given the multifunctionality and elusiveness of some metadiscourse items having more than one function at a time and different functions in different contexts.
Another consideration in these studies is the reference made to cultural values (British, American, and Iranian) and the differences these values make in the use of rhetorical devices in the texts. Cultural preferences will undoubtedly affect the style of discourse organization and the degree of rhetorical uncertainty or assertiveness of the writers and the degree of their reader-oriented or writer-oriented discourse (Ädel 2006;Hyland 2004). On the other hand, some of the studies refer to the important role of schooling as well as second language learning experiences. Azizi (2003), for instance, refers to the important role of the extent of second language learning experience and the significant impact it has on learners' adopting a second language identity and having thoughts different from those thoughts in the native language. This second language identity can distinguish the rhetorical practices of the same learners' written productions in English from that of their Persian writings. Further, the possible similarities between Iranian and British/American professional writers in the use of certain rhetorical devices can be attributed to the degree of training and familiarity with the rhetorical norms and preferences of their disciplines. In other words, although they come from different ‗national cultures' (Iranian, British/US English), they belong to the same ‗disciplinary culture' (Mauranen 1993), or to particular knowledge disciplines. Thus, the implication would be that we need to examine the written practices of novice and professional writers at different levels of proficiency and expertise, who are from different cultures and disciplines to see if there is any developmental pattern of effective pragmatic development in their use of rhetorical devices like metadiscourse.
In addition to cultural differences, the impact of gender on the use of rhetorical devices is yet another significant factor which was not an issue for these researchers yet, although -[…] the gender of the writer could influence how much or what type of metadiscourse is used‖ (Ädel 2006: 198). The reviewed studies were mainly concerned with the overall picture of similarities and differences between writers in the use of the types or subtypes of metadiscourse. However, the corpora selected for analysis were from both male and female writers and students. Crismore et al. (1993) found gender and cultural differences between Finnish and American male and female writers. Finnish females used the most hedges and US males the least. Finnish females also used more hedges than US females. Moreover, some research shows that males employ more emphatics than females and demonstrate a more confident writing style (Francis, Robson & Read 2001;Tse & Hyland 2008). Research in this area is scant. Consequently, future cross-cultural research should reveal more about possible gender-specific rhetorical practices. If consistent gender differences appear in many studies across different text types, cultures, and contexts, this finding can suggest that there might even be causal connections to social or biological gender and the propensity to use metadiscourse. If so, these causes may have considerable effects in the foreign language classroom and may also affect the individual teacher's syllabus and teaching methodology.
The reviewed studies show that learners at different language proficiency levels benefit from effective metadiscourse instruction and awareness-raising in their comprehension and written production. As was found in the Dastgoshadeh and Parvarseh studies, it seems that metadiscourse knowledge and use can compensate for learners' inadequate pragmalinguistic competence and can boost their comprehension and memory of the propositional content of the text. However, whether these findings are appropriate for learning strategies in both LI and in L2 classes is yet to be closely examined though the use of questionnaires and think-aloud protocols. Of course, we need to also examine the correlation for metadiscourse use between the written productions of EFL learners and their written fluency. Lack of background knowledge may lead to overuse of metadiscourse to disguise the learners' gaps in knowledge. Therefore, examining the proportion of appropriate metadiscourse use in easy, moderately challenging, and difficult tasks can shed more light on the extent to which L2 learners at different proficiency levels use their metadiscoursal rhetorical awareness to overcome their inadequate knowledge of form. This awareness might also help L2 learners to perform illocutionary acts in a more effective manner. From such studies, we can explore non-native EFL learners' use, misuse, and overuse patterns of metadiscourse.
We know that the choice of metadiscourse expression is highly dictated by the overall structure of the discourse, communicative purpose, and the level of tentativeness or universality of our claims as writers (Salager-Meyer 1994). In addition to the genre, learner level and task familiarity, the contribution of these metadiscourse markers is also a function of the language skill we practice, the text type we produce (e.g. argumentative, narrative, etc.), and the constraints of the communicative situation. In the reviewed studies we have noticed the cumulative effects of these markers in improving discourse comprehension in the reading and writing practices of experimental and control groups. However, to establish a pedagogical theory of metadiscourse, we need to further our knowledge of the percentage contribution of each of the categories or individual types of metadiscourse in different genres, disciplines, text types, and skills among different populations (e.g. native and non-native speakers). Future studies in line with that of Daftary Fard can be instrumental in demonstrating the psycholinguistic validity of metadiscourse as a significant construct in different language skills.
Our final point relates to the design problems with corpus-based studies of metadiscourse in general and the studies reviewed here. In brief, future research would need to make sure of the comparability of corpora for research and comparison purposes (cf. Ädel 2006:201ff). That is, researchers working on metadiscourse need to make sure that their data are comparable in terms of length, purpose, setting, writer groups, the status of the journals examined, journal prestige, competitiveness of the context leading to the particular discourse production, and the level of research space required from writers.
The continuing interest in metadiscourse by students and researchers worldwide is evidence that metadiscourse -is a distinctive characteristic of language, ubiquitous in our speech, and it deserves close attention from linguists.‖ (Mauranen, this volume). It is strongly expected that future studies of metadiscourse will add to our knowledge of effective rhetorical strategies for various cultures and contexts.