Metadiscursive Markers and Text Genre: A Metareview

: Given the interest in the study of metadiscourse as the communication of ideas and the way people use language in different communicative situations, this paper attempted to ﬁnd the degree of conﬂuence between metadiscourse markers from different studies and to show how patterns of metadiscourse analysis based on various written genres can be applied to a wider range. The mean values for the frequency of marker use and their respective deviations were determined by comparing a signiﬁcant number of studies on metadiscourse elements. To ensure comparability, those following Hyland’s model were chosen. The units of analysis were grouped into two broad categories based on discursive characteristics: Academic genres (research articles, theses, and textbooks) and non-academic genres, which included documents ranging from newspaper editorials or opinion columns to Internet texts and other forms of digital communication. The results of our study highlight that the disparity in interactive markers between academic and non-academic texts is relatively small. This difference has been identiﬁed by previous studies, and it is conﬁrmed herein that the difference may be related to the use of academic language, the topic, or the object of study. In contrast, the mean values of the interactive markers in non-academic texts are considerably higher than those in academic texts. At the same time, the texts seem to be organised along two axes (interactional and interactive) in distinct areas. Despite our initial assumptions that the data would be subject to individual variations, that differences would be found between different sections of the same genre within the same academic discipline, and that the results would vary if certain texts were added or excluded, we observed certain trends in the behaviour of the documents, although it prevailed that, within each category, the texts should be studied individually.


Definitions
on the topic, his arguments, and the content of the text. Of course, when we understand language use more deeply, we can use our knowledge to improve the quality of our own writing. Adding these types of words can help us express ourselves better in our non-native language, help others understand what we have written, and potentially improve our academic performance.
Metadiscourse markers are words or phrases that help connect and organise text, express an attitude, provide evidence, connect the reader to the writer, and ensure that the text "flows" from one idea to the next. This term has been studied extensively by numerous linguists, including Ken Hyland and Annelie Ädel [7]. Hyland [8] showed that metadiscourse essentially refers to how we use language out of consideration for our readers or listeners "based on our estimation of how we can best help them process and understand what we are saying..." In this respect, according to Vande Kopple [9], metadiscourse is "discourse about discourse or communication about communication." Meanwhile, Hyland [5] referred to it as textual elements used to organise text, to indicate the writer's attitude, and to make the intended message more effective and easier for the addressee/reader to understand. Metadiscourse also includes authorial presence without the addition of propositional information. In a certain sense, metadiscourse is the "author's intrusion into the discourse, either explicitly or non-explicitly, to direct the reader rather than inform" [10]. Therefore, it refers not to the contents of the text but to the speaker/writer, the listener/reader, or the text.
With regard to the elements encompassed by metadiscourse, Ädel [7] distinguished between a broad view of metadiscourse, which defines as each and every linguistic resource that organises the text or contributes to communicating the author's opinions, and a narrower view, which conceptualises metadiscourse exclusively as the linguistic elements that fulfil a textual function. This narrow view is excessively constrictive and overlooks the interpersonal essence of metadiscourse, as Wei et al. [11] noted in their literature review.
In this theoretical framework, we must emphasise that analogously to how the definition of a "metadiscursive marker" is not unitary, neither are the ways that the term is classified [12,13].

Models
In her discussion of Williams's and Meyer's classifications, Crismore [10] classified metadiscourse into two general categories: Informational, the aspects of discourse intended to improve the reader's understanding, and attitudinal, those that reveal the author's attitude towards the content or structure of the text or towards the listener/reader. Crismore further divided each category into several subtypes (see Figure 1).
According to Vande Kopple [9], metadiscourse can be divided into two vague, functionally overlapping categories: textual metadiscourse and interpersonal metadiscourse. Thus, he established seven types of metadiscursive markers. As illustrated by Figure 2, the first four types are textual (text connectives, code glosses, illocution markers, and narrators), whereas the remaining three are interpersonal (validity markers (hedges, emphatics, and attributors), attitude markers, and commentaries). According to Vande Kopple [9], metadiscourse can be divided into two vague, functionally overlapping categories: textual metadiscourse and interpersonal metadiscourse. Thus, he established seven types of metadiscursive markers. As illustrated by Figure 2, the first four types are textual (text connectives, code glosses, illocution markers, and narrators), whereas the remaining three are interpersonal (validity markers (hedges, emphatics, and attributors), attitude markers, and commentaries).   According to Vande Kopple [9], metadiscourse can be divided into two vague, functionally overlapping categories: textual metadiscourse and interpersonal metadiscourse. Thus, he established seven types of metadiscursive markers. As illustrated by Figure 2, the first four types are textual (text connectives, code glosses, illocution markers, and narrators), whereas the remaining three are interpersonal (validity markers (hedges, emphatics, and attributors), attitude markers, and commentaries).  Crismore and Farnsworth [14] added the category of "scientific commentaries", and Crismore et al. [15] divided metadiscourse into textual metadiscourse (text markers and interpretive markers) and interpersonal metadiscourse (hedges, certainty markers, attributors, attitude markers, and commentaries).
Focusing on academic texts, Hyland [16,17] proposed a modification of the classification in order to distinguish between textual and interpersonal metadiscourse. In addition, Hyland offered a more interpersonal perception of metadiscourse [5,18]. Studies following this approach reject the dichotomy of textual and interpersonal functions and develop a model tailored to texts characterised as an academic writing ( Figure 3). This model presumes two macro-categories: Interactive, which helps guide the reader through the text, interpretive markers) and interpersonal metadiscourse (hedges, certainty markers, attrib-utors, attitude markers, and commentaries).
Focusing on academic texts, Hyland [16,17] proposed a modification of the classification in order to distinguish between textual and interpersonal metadiscourse. In addition, Hyland offered a more interpersonal perception of metadiscourse [5,18]. Studies following this approach reject the dichotomy of textual and interpersonal functions and develop a model tailored to texts characterised as an academic writing ( Figure 3). This model presumes two macro-categories: Interactive, which helps guide the reader through the text, and interactional, which helps involve the reader in the text. Several authors have adapted this model to fit the genre they study [19][20][21][22]. For a more in-depth discussion of the various approaches to both the theoretical framework and the categorisation of marker types see Wei et al. [11], who reviewed metadiscourse theories and models from the past 15 years.

Research Objective and Hypotheses
The aim of this paper was to determine whether the presence of metadiscourse markers in different texts follows a similar behaviour depending on the type of academic or non-academic text in which it is found. In categorising the studies focused on academic writing we grouped these studies into three broad categories: (a) research articles, (b) university theses and dissertations, and (c) textbooks. In order to analyse non-academic texts, we used four categories: newspapers, Internet texts, tourist promotion materials, and texts that we denominated "oral texts". However, we did not make distinctions regarding the section of the text analysed (e.g., abstract, introduction, or conclusion), the subject matter (hard versus soft sciences), or the type of study, i.e., or whether the study was empirical or non-empirical.
This study aimed to compare the results obtained from other research works whose analysis of metadiscursive markers was grounded in Hyland's metadiscourse model. Comparisons of metadiscursive marker patterns within the same genre have been studied by several researchers, as has become evident in the referenced bibliography. Earlier For a more in-depth discussion of the various approaches to both the theoretical framework and the categorisation of marker types see Wei et al. [11], who reviewed metadiscourse theories and models from the past 15 years.

Research Objective and Hypotheses
The aim of this paper was to determine whether the presence of metadiscourse markers in different texts follows a similar behaviour depending on the type of academic or nonacademic text in which it is found. In categorising the studies focused on academic writing we grouped these studies into three broad categories: (a) research articles, (b) university theses and dissertations, and (c) textbooks. In order to analyse non-academic texts, we used four categories: newspapers, Internet texts, tourist promotion materials, and texts that we denominated "oral texts". However, we did not make distinctions regarding the section of the text analysed (e.g., abstract, introduction, or conclusion), the subject matter (hard versus soft sciences), or the type of study, i.e., or whether the study was empirical or non-empirical.
This study aimed to compare the results obtained from other research works whose analysis of metadiscursive markers was grounded in Hyland's metadiscourse model. Comparisons of metadiscursive marker patterns within the same genre have been studied by several researchers, as has become evident in the referenced bibliography. Earlier studies [16,18,23] have suggested that the use of metadiscursive markers in academic texts varies by linguistic and cultural communities and academic discipline.
The proposition that the use of metadiscursive markers can differentiate genres and text types to discern whether a generalisable pattern exists seems, to a certain extent, reasonable.
The hypotheses of the research are as follows: Hypothesis 1 (H1): The data will be subject to individual variations, that is, differences will be found between different sections of the same genre within the same academic discipline.

Hypothesis 2 (H2):
The use of metadiscourse markers in academic and non-academic texts will be different according to the categories of Hyland's model.

Hypothesis 3 (H3):
There will be similar behaviour of metadiscourse markers according to the selected classification of academic and non-academic texts.

Corpus
The corpus of this study comprised 334 documents obtained from the ScienceDirect and Scopus databases and Internet searches used to obtain book chapters. The selected texts are accessible with the University of Vigo library resources from the Spanish Foundation for Science and Technology.
The search criteria were the inclusion of the term "metadiscourse" in the text's title, abstract, or keywords. After selecting the documents, those that addressed discourse analysis or rhetorical analysis were excluded because these topics were not the focus of this study.

Data Collection-Selection Criteria
Considering the metadescriptive nature of this paper, rigorous article selection criteria were established to limit the number of studies and to obtain a sample that would guarantee comparability, thus facilitating the comparison of values among various types of metadiscursive markers in each document. The selection criteria used were the following:

1.
Articles that adhered to Hyland's model were chosen because this model is the most frequently cited and most widely used in studies since the mid-2000s.

2.
Articles that differentiated interactive and/or interactional metadiscursive markers were chosen.

3.
Texts that used a classification system-based differentiation between transition, frame markers, endophoric markers, evidentials, and code gloss markers were included.

4.
Texts that used a classification system-based differentiation between hedges, boosters, attitude markers, self-mentions, and engagement markers were included. 5.
Documents that described and analysed the presence of markers in the texts were excluded when their corpus size was unknown because their data (number of appearances per 1000 words) could not be normalised prior to comparison.
Only the papers that belonged to one of the categories in Figure 4 were included. The selected studies (44; see Appendix A) were categorised into two broad groupings, namely, academic texts and non-academic texts, with the following subcategories (  In the case of Bachelor's, Master's, and doctoral theses, we are aware that the former two should exhibit a lower degree of complexity than should the latter. Nevertheless, considering that the terms "thesis" and "dissertation" are used differently in academic contexts, we grouped these different documents together to obtain aggregate comparative averages. As Hyland [23] noted, theses are treated "differently in different countries and sometimes even in different universities in the same countries". Works that analyse re-  In the case of Bachelor's, Master's, and doctoral theses, we are aware that the former two should exhibit a lower degree of complexity than should the latter. Nevertheless, considering that the terms "thesis" and "dissertation" are used differently in academic contexts, we grouped these different documents together to obtain aggregate comparative averages. As Hyland [23] noted, theses are treated "differently in different countries and sometimes even in different universities in the same countries". Works that analyse research articles were also grouped together, regardless of whether those texts were article abstracts, conclusions, or bodies and whether they addressed topics related to the hard or soft sciences. Documents associated with oral presentations, such as lessons and sermons, were grouped together due to their expository nature. In the case of internet texts, all texts associated with wikis, the web, chats, and online forums were grouped together.

Data Analysis
In this section, the values regarding the presence of metadiscursive markers obtained from the selected studies are compared and analysed. Our overall approach adhered to the distinction between interactive and interactional markers.

Interactive Markers
The research studies analysed for our meta-study showed that the linguistic and cultural context in which texts are written can impact the use of metadiscursive markers, even within the same genre. Below, we hypothesise about the possible causes of this variation.
As shown in Figure 5, academic texts (Master's theses, research articles, and textbooks) exhibited significant variation. Such variation may be owing to several factors. First, plotting the frequency of interactive markers as a function of text type showed that these markers are unevenly distributed. The significant internal variability within each subcategory is also evident, particularly for theses. This high variability can be explained by differences in content and length and those resulting from the type of work entailed by the subject matter. Doctoral students use more interactive markers and engagement markers, and self-mentions appear much more frequently in doctoral theses than in Master's theses [23,24]. Likewise, we believe The significant internal variability within each subcategory is also evident, particularly for theses. This high variability can be explained by differences in content and length and those resulting from the type of work entailed by the subject matter. Doctoral students use more interactive markers and engagement markers, and self-mentions appear much more frequently in doctoral theses than in Master's theses [23,24]. Likewise, we believe that the variation observed in the research articles fundamentally stems from the fact that marker usage may be determined by the discipline [5].
Second, if we focus on interactive markers, we can see that, on average, transitions/logical markers are the most frequently used. This result holds regardless of whether the analysed texts are theses, textbooks, or research articles or whether their disciplines are considered soft (social sciences) or hard (hard and technological sciences), though significant differences exist among several text types. Additionally, frame markers stand out in theses, and code glosses stand out in textbooks and research articles. Moreover, code glosses are the least-used marker type in theses ( Figure 6). A plethora of research studies have been conducted on the analysis of MD markers in different disciplines and different languages [25][26][27][28][29]. The overall results of these studies indicate that there is a strong association between the distribution of metadiscourse markers and the specific discipline.
Furthermore, a large proportion of evidentials is present in Internet texts (twice the percentage of other text types), likely because of the intertextual, multi-author, and social nature of Internet texts in terms of both their initial creation and their distribution, adaptation, and comment generation. The formation of relationships by referring to information in other parts of the text also follows logical patterns: Endophoric markers appear more frequently in long texts that require such references, such as academic texts or internet texts. However, endophoric markers are inconsequential in shorter texts (newspapers) and texts meant to be read aloud (sermons and lessons). Conversely, the more significant presence of frame markers in oral texts (sermons and lessons) is logical because the oral nature of such texts makes references to discourse shifts or text stages even more necessary.

Interactional Markers
As mentioned earlier, interactional markers, such as the elements that refer and relate to interactional practices [30], are used to establish speaking turns and linguistically comply with certain sociocultural conversation norms, among other functions. In this respect, interactional markers are clearly directed at the listener/reader, and such markers may arguably function to monitor message transmission [31]. In other words, interactional The greater influence of transitions/logical markers is evident in all texts, regardless of textual characteristics, as shown in Figure 6, which shows the percentage distribution of items per 1000 words.
A plethora of research studies have been conducted on the analysis of MD markers in different disciplines and different languages [25][26][27][28][29]. The overall results of these studies indicate that there is a strong association between the distribution of metadiscourse markers and the specific discipline.
Furthermore, a large proportion of evidentials is present in Internet texts (twice the percentage of other text types), likely because of the intertextual, multi-author, and social nature of Internet texts in terms of both their initial creation and their distribution, adaptation, and comment generation. The formation of relationships by referring to information in other parts of the text also follows logical patterns: Endophoric markers appear more frequently in long texts that require such references, such as academic texts or internet texts. However, endophoric markers are inconsequential in shorter texts (newspapers) and texts meant to be read aloud (sermons and lessons). Conversely, the more significant presence of frame markers in oral texts (sermons and lessons) is logical because the oral nature of such texts makes references to discourse shifts or text stages even more necessary.

Interactional Markers
As mentioned earlier, interactional markers, such as the elements that refer and relate to interactional practices [30], are used to establish speaking turns and linguistically comply with certain sociocultural conversation norms, among other functions. In this respect, interactional markers are clearly directed at the listener/reader, and such markers may arguably function to monitor message transmission [31]. In other words, interactional markers concern "the way that writers conduct interaction by intruding and commenting on their message" [5]. Figure 7 clearly illustrates that interactional markers are more widely used in nonacademic than in academic texts. Interactional markers are particularly prevalent in internet texts. The data on interactional markers also vary less than the data on interactive markers, as demonstrated by the corresponding standard deviations. Additionally, hedges are the most frequently used marker regardless of the type of academic text or scientific field. Boosters constitute the next most frequently used marker in research articles (hedges and boosters combined account for 59.84% of markers per 1000 words), and engagement markers are the second most frequently used markers in theses (hedges and engagement markers together account for 62.34% of markers per 1000 words) and textbooks (with the two types combined representing 62.41% of markers per 1000 words). Self-mentions are the least frequently used markers. As illustrated by Figure 8, the data showing the frequency of the different interactional markers for these three text types are highly similar.
Hedges appear much less frequently in non-academic genres than in academic texts, with the number of hedges in non-academic texts being approximately one-third of that of academic texts. Several other significant differences between academic and non-academic texts in the use of interactional markers are apparent: Self-mentions and engagement markers. Additionally, hedges are the most frequently used marker regardless of the type of academic text or scientific field. Boosters constitute the next most frequently used marker in research articles (hedges and boosters combined account for 59.84% of markers per 1000 words), and engagement markers are the second most frequently used markers in theses (hedges and engagement markers together account for 62.34% of markers per 1000 words) and textbooks (with the two types combined representing 62.41% of markers per 1000 words). Self-mentions are the least frequently used markers. As illustrated by Figure 8, the data showing the frequency of the different interactional markers for these three text types are highly similar. Publications 2021, 9, x FOR PEER REVIEW 10 of 20 The differences in the distribution of interactional markers that, broadly speaking, distinguish academic texts from non-academic texts are evident in Figure 8, which shows the percentage of appearances of each type of interactional marker for different text types. The percentage distribution of each marker type appears to not only differentiate academic from non-academic texts but also to typify each of the distinct types of non-academic texts analysed in this study.
Given that engagement markers function to engage listeners/readers and involve them with the text, this marker type is frequently used in both texts with an oral nature that are intended to convince listeners/readers (sermons and lessons) and Internet texts. In the case of the latter, we believe that the frequent use of engagement markers is linked to the aim to express opinions and convince individuals who visit websites and participate in online forums. Similarly, the significant presence of self-mentions in Internet texts likely stems from the intertextual, multi-author, and social nature of Internet texts in terms of both their initial creation and their distribution, adaptation, and comment generation.
The expression of opinions, which is common in newspapers, Internet texts, and tourism promotion materials because of their communicative intent, is reflected by a higher proportion of attitude markers. By contrast, attitude markers rarely appear in texts intended to convince or explain, such as sermons and lessons. Tourist promotional materials and texts of a more scholarly nature clearly exhibit a high percentage of boosters, which help emphasise certainty about the ideas expressed in these texts.

Discussion
This article examined studies on metadiscourse in academic and non-academic texts. In categorising the studies focused on academic writing, we grouped these studies into three broad categories: (a) research articles, (b) university theses and dissertations, and (c) textbooks; we divided non-academic texts into four categories: newspapers, internet texts, Hedges appear much less frequently in non-academic genres than in academic texts, with the number of hedges in non-academic texts being approximately one-third of that of academic texts. Several other significant differences between academic and non-academic texts in the use of interactional markers are apparent: Self-mentions and engagement markers.
The differences in the distribution of interactional markers that, broadly speaking, distinguish academic texts from non-academic texts are evident in Figure 8, which shows the percentage of appearances of each type of interactional marker for different text types. The percentage distribution of each marker type appears to not only differentiate academic from non-academic texts but also to typify each of the distinct types of non-academic texts analysed in this study.
Given that engagement markers function to engage listeners/readers and involve them with the text, this marker type is frequently used in both texts with an oral nature that are intended to convince listeners/readers (sermons and lessons) and Internet texts. In the case of the latter, we believe that the frequent use of engagement markers is linked to the aim to express opinions and convince individuals who visit websites and participate in online forums. Similarly, the significant presence of self-mentions in Internet texts likely stems from the intertextual, multi-author, and social nature of Internet texts in terms of both their initial creation and their distribution, adaptation, and comment generation.
The expression of opinions, which is common in newspapers, Internet texts, and tourism promotion materials because of their communicative intent, is reflected by a higher proportion of attitude markers. By contrast, attitude markers rarely appear in texts intended to convince or explain, such as sermons and lessons. Tourist promotional materials and texts of a more scholarly nature clearly exhibit a high percentage of boosters, which help emphasise certainty about the ideas expressed in these texts.

Discussion
This article examined studies on metadiscourse in academic and non-academic texts. In categorising the studies focused on academic writing, we grouped these studies into three broad categories: (a) research articles, (b) university theses and dissertations, and (c) textbooks; we divided non-academic texts into four categories: newspapers, internet texts, tourist promotion materials, and texts that we designated as oral texts. Nevertheless, we did not make distinctions regarding the section of the text analysed (e.g., abstract, introduction, or conclusion), the subject matter (hard science versus soft sciences), or the type of study, that is, whether the study was empirical or non-empirical. Instead, we grouped these studies into three broad categories: (a) research articles, (b) university theses and dissertations, and (c) textbooks.
This study aimed to compare the results obtained from other research works whose analysis of metadiscursive markers was grounded in Hyland's metadiscourse model. Comparisons of metadiscursive marker patterns within the same genre have been studied by several researchers. Earlier studies have suggested that the use of metadiscursive markers in academic texts varies by linguistic and cultural communities and academic discipline.
Academic writing, particularly with respect to research articles, is possibly the most researched field. Nearly all studies indicated that the use of metadiscursive elements is influenced by such aspects as the context, field, genre, and the linguistic and cultural background of the author(s). Research has shown (Figure 9) a higher average of interactive markers compared to interactional markers in academic texts, independent of standard deviation. This aspect distinguishes academic texts from non-academic texts, in which interactional markers are predominantly in contrast. The standard deviations, however, indicate greater variability in academic texts than in non-academic ones, which, as earlier studies have indicated, reflects the influence of non-linguistic factors: the discursive community, language, subject matter, or the part of the texts that was analysed. tourist promotion materials, and texts that we designated as oral texts. Nevertheless, we did not make distinctions regarding the section of the text analysed (e.g., abstract, introduction, or conclusion), the subject matter (hard science versus soft sciences), or the type of study, that is, whether the study was empirical or non-empirical. Instead, we grouped these studies into three broad categories: (a) research articles, (b) university theses and dissertations, and (c) textbooks. This study aimed to compare the results obtained from other research works whose analysis of metadiscursive markers was grounded in Hyland's metadiscourse model. Comparisons of metadiscursive marker patterns within the same genre have been studied by several researchers. Earlier studies have suggested that the use of metadiscursive markers in academic texts varies by linguistic and cultural communities and academic discipline.
Academic writing, particularly with respect to research articles, is possibly the most researched field. Nearly all studies indicated that the use of metadiscursive elements is influenced by such aspects as the context, field, genre, and the linguistic and cultural background of the author(s). Research has shown (Figure 9) a higher average of interactive markers compared to interactional markers in academic texts, independent of standard deviation. This aspect distinguishes academic texts from non-academic texts, in which interactional markers are predominantly in contrast. The standard deviations, however, indicate greater variability in academic texts than in non-academic ones, which, as earlier studies have indicated, reflects the influence of non-linguistic factors: the discursive community, language, subject matter, or the part of the texts that was analysed. Academic texts feature a higher percentage of interactive markers than they do interactional markers, which is in stark contrast to non-academic texts. The unique characteristics of academic texts, particularly in terms of such features as the diversity of subjects they discuss and the diverse scientific and linguistic cultures that shape them, may explain the observed variability of the data. To this end, Hyland [16] singled out abstracts as an Academic texts feature a higher percentage of interactive markers than they do interactional markers, which is in stark contrast to non-academic texts. The unique characteristics of academic texts, particularly in terms of such features as the diversity of subjects they discuss and the diverse scientific and linguistic cultures that shape them, may explain the observed variability of the data. To this end, Hyland [16] singled out abstracts as an example. Abstracts are drafted to capture the attention of potential readers and therefore entail greater use of elements meant to strengthen reader interaction. Jimenez [20] suggested other reasons to explain the variability of marker usage in academic texts, including the writer/reader relationship, the degree of competition, the emphasis on authorial contributions, and the national scholarly culture.
Internet texts show higher percentages of evidential marker use, derived from the intertextual, multi-author, and social nature of these texts in terms of both their initial creation and their distribution, adaptation, and comment generation. Endophoric markers likewise appear more frequently in academic texts and Internet texts because references to information in other parts of the text are necessary, yet this type of marker is nearly undetectable in texts that are briefer or are meant to be spoken aloud. This finding also seems consistent with the higher frequency of frame markers in oral texts, which tend to indicate discursive shifts and the stages of the text.
These variations in the usage of different markers can also be observed in the case of interactional markers (Figure 8). Academic texts revealed a high frequency of hedges (slightly less than 40% of all interactional markers per thousand words) and boosters (approximately 20% of all interactional markers per thousand words), compared to other types of markers, highlighting its use in research articles. The significant presence of these two marker types in academic texts may be owing to the need to establish logical relationships in complex texts and emphasise authorial certainty about the ideas expressed in the texts. Tourist promotion materials share this need to emphasise certainty, but the need to use emphasisers to convince and convey certainty is even greater in tourist promotion materials due to the social and economic influences involved in their development. The high percentage of boosters reflects this fact, as boosters can represent nearly 45% of the total interactional markers per thousand words.
When comparing oral texts intended to explain contents and ideas or convince listeners and/or readers (sermons and lessons) to other text types, the high frequency of engagement markers stands out, accounting for nearly 60% of interactional markers per thousand words. Conversely, the greater proportion of attitude markers in newspapers reflects the expression of personal opinions and attempts at persuasion, which is common for this text type [32][33][34][35]. Attitude markers, however, hardly appear in texts that are clearly intended to convince or explain. Meanwhile, tourist promotion materials feature few engagement markers, attitude markers, self-mentions, or mitigators. These results broadly coincide with the findings of previous studies [12,[36][37][38][39].
Internet texts exhibit a high frequency of self-mentions and engagement markers (these two types constitute over 60% of all interactional metadiscursive markers per thousand words), and this seems to distinguish Internet texts from the other text types. However, differences exist based on the website and the culture in which the website is produced, as some researchers have noted [40]. In our data, such differences resulted in high standard deviations, which can be explained by the fact that Internet texts are highly dependent on the culture and discourse community of their intended audience and on the intertextual, multi-author, and social nature in terms of both their initial creation and their distribution, adaptation, and comment generation. As the author of [41] stated, the digital environment alters the message, the individual who issues that message, and the individual who receives it. According to Jensen [42], user participation is also characterised by a higher frequency of engagement markers (approximately 32% per thousand words in our metadescription), though this percentage varies by the type of newspaper text analysed, which the author of [42] believed results from unplanned digital communication (i.e., comments) and the subject matter.
Although previous studies have detected metadiscursive similarities that allow journalistic genres to be considered as an entirety, various authors have identified several extratextual explanations that can justify the variability of the data regarding newspaper texts in our metadescription. These explanations include rhetorical and cultural preferences [43] that influence persuasiveness, the newspaper type [44], the subject matter [36,45], the newspaper section [6], and the columnist [46].

Conclusions
As noted in the introduction of this paper, metadiscourse is crucial to text creation because metadiscursive markers help establish relationships between the speaker/writer, the listener/reader, the sociocultural context, and the specific communicative situation. Metadiscourse guides how a message is interpreted because it also reflects the linguistic expression of sociocultural reality and the use of metadiscursive markers can differentiate genres and text types to a certain extent.
This study indicated that each type of text shows a preference for the use of one or more types of metadiscursive markers. These preferences seem to differ between text types, being especially visible in the case of interactional markers. One of the clearest findings of this study emerged when we presented the number of interactional versus interactive markers, as illustrated in Figure 10. The texts appear to be organised along the two axes in distinct zones.

Conclusions
As noted in the introduction of this paper, metadiscourse is crucial to text creation because metadiscursive markers help establish relationships between the speaker/writer, the listener/reader, the sociocultural context, and the specific communicative situation. Metadiscourse guides how a message is interpreted because it also reflects the linguistic expression of sociocultural reality and the use of metadiscursive markers can differentiate genres and text types to a certain extent.
This study indicated that each type of text shows a preference for the use of one or more types of metadiscursive markers. These preferences seem to differ between text types, being especially visible in the case of interactional markers. One of the clearest findings of this study emerged when we presented the number of interactional versus interactive markers, as illustrated in Figure 10. The texts appear to be organised along the two axes in distinct zones. Despite our initial assumptions that the data found are subject to individual variations (e.g., the same text evaluated by another researcher), that there may be differences between different sections of the same genre within the same discipline of study, and that the addition or subtraction of some of the texts could have caused the results to vary significantly, a certain trend in the behaviour of the documents was observed, as seen in the previous paragraphs. This does not obviate the fact that, in order to know the individual Despite our initial assumptions that the data found are subject to individual variations (e.g., the same text evaluated by another researcher), that there may be differences between different sections of the same genre within the same discipline of study, and that the addition or subtraction of some of the texts could have caused the results to vary significantly, a certain trend in the behaviour of the documents was observed, as seen in the previous paragraphs. This does not obviate the fact that, in order to know the individual behaviour of each document, it is necessary to study each specific case, given the heterogeneity in terms of category, subject matter, language, etc., and the authors' preference regarding the use of one or another marker.
The metadiscourse analysis was carried out using the list of words identified by Hyland in his book, but other potential options discovered by various researchers were also added to this list, as Hyland confirmed that this is an open category that allows researchers to contribute to the identification of metadiscursive elements. This reasoning explains why the level of specialisation of the target audience is relevant when characterising the text type since the members of a community share a common knowledge that the author of the text may take for granted [47]. Figure 9 showed that in the categories we selected and, more specifically, in academic texts, interactive markers tended to have a slightly higher mean value than interactional markers, regardless of the standard deviation, as can be seen in the different studies of the referenced researchers.
Similarly, it was clearly observed that in non-academic texts, practically the opposite occurred, the interactional mean values, in this case, being much higher, even tripled. The only case in which this did not occur is the grouped genre of journalism, these are similar and depend on the deviation. Moreover, in the case of tourism promotion, in which we did not find documents that meet the limitations, we established for the case of interactive markers that it is not possible to try to generalise as in the previous cases.
In response to the hypotheses Hypothesis 1 (H1): The data will be subject to individual variation, that is, differences will be found between different sections of the same genre within the same academic discipline.

Hypothesis 2 (H2):
The use of metadiscourse markers in academic and non-academic texts will be different according to the categories of Hyland's model.

Hypothesis 3 (H3):
There will be a similar behaviour of metadiscourse markers according to the selected classification of academic and non-academic texts.
we found the following:

R1:
We did not achieve sufficient uniformity in the number of studies to be analysed to ensure this statement. What was found is a very important variability among the studies of academic texts, so that this initial hypothesis could be accepted, as corroborated by many researchers.

R2:
As can be seen in Figure 9, in non-academic texts, the use of international markers is very markedly increased compared to academic texts, even taking into account their variability.

R3:
It was not possible to discern the behaviour according to the classification chosen. We have to take into account that within academic texts, the deviation is already important due to the different characteristics of the studies analysed.
If we adopt a more dynamic perspective that considers not only the direction of most of the data but also their dispersion and variability, our metadescriptive analysis confirms that these relationships are considerably more complex, and genre/text type boundaries are blurry and are constantly being validated by language users who belong to specific discourse communities. This complexity largely originates from the heterogeneous nature of the society in which these communities arise. In analysing certain characteristics of academic and non-academic discourse, we cannot neglect how relationships between communities are negotiated. In this sense, specific features of the discourse community reflect the unique characteristics produced through discourse, with the text understood as the relationships of resonance within an academic and/or a sociocultural community.
Our metadescription confirms this idea. For example, as we mentioned in our description of the data on interactive markers, the proportion of different marker types is linked to the text type. Internet texts show higher percentages of evidential marker use, which, as mentioned earlier, derives from the intertextual, multi-author, and social nature of these texts in terms of both their initial creation and their distribution, adaptation, and comment generation. Endophoric markers likewise appear more frequently in academic texts and internet texts because references to information in other parts of the text are necessary, yet this type of marker is nearly undetectable in texts that are briefer or are meant to be spoken aloud. This finding seems furthermore consistent with the higher frequency of frame markers in oral texts, which tend to indicate discursive shifts and the stages of the text. These variations in the usage of different markers can also be observed in the case of interactional markers.
We should also point out a limitation of this study that has to do with the corpus used in the different documents on which our metadescription was based. This limitation is threefold: (a) the number of studies used for our meta-analysis, which allows generalisation at this time but should nevertheless be increased in future studies; (b) the discrepancies in the size of the corpora of the studies analysed here; and (c) one of the most important limitations in the analysis of metadiscourse, both in this study and in others, lies in the knowledge of the list of words identified at first by Hyland, but which are also added to by researchers. It should be remembered that Hyland himself confirmed that this is an open category that allows contributing to the identification of metadiscursive elements since different words or grammatical turns of phrase can be used to communicate the same thing.
As several researchers have mentioned, future research could also extend this investigation to other domains to discern whether the new genres follow the same patterns as traditional genres. In order to a better knowledge of the behaviour of these metadiscursive markers and its correlation with the different discourse communities, it is necessary to conduct a more extensive and cross-cultural analysis based on bigger corpora that contain more textual genres.
Author Contributions: Both authors contributed equally to this work. All authors have read and agreed to the published version of the manuscript.

Conflicts of Interest:
The authors declare no conflict of interest.

Appendix A
The following documents were used to categorise academic texts and to create the tables and figures in this study.

Author-Reference
Year