Metadiscourse within a discipline: A study of introduction and literature review chapters of sociology masters’ theses

It has been established that there are inter-chapter variations regarding metadiscourse use in masters’ thesis. This paper, therefore, investigates the differences and similarities in metadiscourse use between the Introduction and Literature Review (LR) chapters in Sociology masters’ thesis from an English-medium university. The Introduction and LR Chapters of ten theses constituted the corpus for this study. The metadiscursive devices were manually coded, drawing on the modified version of the Hyland’s (2005a) interpersonal model of metadiscourse. It was realized that there are both differences and similarities between the two chapters in terms of metadiscourse use. In the ranking, the interactive and interactional devices, transitions, and hedges respectively recorded the highest frequencies in the two chapters. Besides, there were significant variations across the two chapters in terms of both interactive and interactional subcategories. The findings have implications for the teaching of English for Research Purposes, and the theory of metadiscourse.

The above studies affirm intra-disciplinary variation in metadiscourse use in the thesis genre. Generally, the studies reveal that metadiscourse use is influenced by the communicative and rhetorical functions of thesis parts. For example, Zahra et al. (2015) found interactional metadiscourse (more than the interactive ones) to be higher in the conclusion Hyland's (2005a) model of metadiscourse originally comprised two main categories: interactive and interactional, with each comprising five subcategories. Generally, the interactive devices help to guide readers through the text, while the interactional ones engage the reader in the text. Hyland's model has been widely applied in studies on academic genres in general and the masters' thesis in particular (e.g. Bal-Gezegin, 2016; Lee & Casal, 2014;Ozdemir & Longo, 2014;Zadeh et al. 2015). Akoto (2012) revealed that Hyland's (2005a) model does not account for certain metadiscourse features, and accordingly added a new subcategory labeled continuants to the interactional category. The modified version of Hyland's model of metadiscourse can be seen in Table 1.

Examples Transitions
Express semantic relation between main clauses. In addition/but/thus/and

Frame makers
Refer to discourse acts, sequences, or text stages Finally/to conclude/my purpose is to

Endophoric makers
Refer to information in other parts of the text Noted above/see Fig./in Section 2 Evidentials Refer to source of information from other texts According to X/(Y, 1990)/Z says

Code glosses
Help readers grasp meanings of ideational material Namely/e.g./such as/in other words

Interactional resources
Involve the reader in the argument

Hedges
Withhold writer's full commitment to proposition Might/perhaps/possible/about Boosters Emphasize force or writer's certainty in proposition In fact/definitely/it is clear that

Attitude markers
Express writer's attitude towards proposition Unfortunately/I agree/surprisingly

Engagement markers
Explicitly refer to or build relationship with reader Consider/note that/you can see that

Self-mentions
Explicit reference to author(s) I/we/my/our Continuants Create a space for reader intrusion into text And so on, etc., among others, … Akoto posited that continuants which involve 'and so on', 'among others'. 'etc.' and 'many more' are metadiscursive but were not accounted for by Hyland (2005a) and even those models that predated it (e.g. Vande Kopple 1985). Akoto (2012) observed that like metadiscursive nouns (Jiang & Hyland, 2016), continuants had been neglected as features of metadiscourse, or had suffered metadiscoursal silence (Swales, 2004). Generally, continuants provide textual space for readers to continue the knowledge production task. He noted that these devices draw on Economics' concept of demand and supply whereby writers consider their reader as knowledge co-producer, and so they (the writers) supply limited knowledge of what the readers demand and expect them to also make contributions. In fact, Hyland somehow allude to these device as he noted that metadiscourse 'helps them[readers] to better understand the cognitive demand that texts make on readers…" (2005a, p. 178). Scholars' apparent neglect of the metadiscursivity of these devices possibly stem from the fact these rhetorical resources are classified as part of vague language (e.g. Lin, 2012Lin, , 2013. Surprisingly, hedges are also regarded as part of vague language (Lin, 2012(Lin, , 2013, but their metadiscursivity for the past years had received scholarly recognition (e.g. Hyland, 2004Hyland, , 2005aHyland & Tse, 2004).

Procedure and materials
The Introduction and Literature Review (LR) chapters of ten theses submitted to the Department of Sociology, University of Cape Coast (UCC), constituted the data for this study as seen in Table  2. The words in the Introduction and LR were approximately 33, 500 and 99, 500 respectively. The significant difference in the sizes of the two datasets is attributable to the differences in the length of the two chapters, which invariably are informed by their varied communicative and rhetorical functions (see Bitchener, 2009;Olmos-Lopez, 2015).
Although in recent times there are plethora of softwares such as textinspector (https://textinspector.com) MetaPak (Abbas et al., 2017) that assist in the study of metadiscourse. Given the highly fuzzy and subjective nature of metadiscourse (Ädel, 2006;Hyland, 2005a), I read the datasets to determine all actual metadiscoursal resources, being guided by the principle of compositionality (see Pelletier, 1994) and manually coded all such items.
Given the uneven number of words from the two datasets, I normalized them at 10, 000 running words, using Saberi's (2017) normalized frequency calculator (http://www.writeabstract.com/calculator.php). For fairness, the normed frequencies (NFs) were used in the discussion instead of the raw frequencies (RFs) and percentages. Besides, the log-likelihood test was employed to test whether the observed differences were statistically significant. The normed frequency (NF) and raw frequency (RF) are provided in Tables 2, as 'frequency is…an essential feature for making general claims about the discourse" (Teubert, 2005, p. 5).

RESULT AND DISCUSSION Variation in Interactive Subcategories across SIC and SLRC
This subsection discusses the variations across SIC and SLRC in terms of interactive resources. The interactive Metadiscourse in SIC and SLRC can be seen in Table  3. It is shown in Table 3 that the scale of preferences (Akoto, 2012) in Sociology Introduction Chapter (SIC) and Sociology Literature Review Chapter (SLRC) are: transitions, code glosses, frame markers, evidentials, and endophorics; and transitions, evidentials, code glosses, frame markers and endophorics respectively. In both chapters, transitions occurred as the most frequent interactive subcategory, while endophorics appeared as the least. Transition ranking as the first subcategory in the two chapters is in tandem with previous studies (e.g. Cherk, 2016;Hyland, 2004;Lee & Casal, 2014;Zahra et al., 2015). Aside from the similarities, we notice that while code glosses recorded the second position in SIC, it was third in SLRC. More so, while frame markers occupied the 3 rd position in SIC, it was 4 th in SLRC. The differences point to the effect of the varied communicative functions of the two chapters on their metadiscoursal preferences.
While the ranking of the subcategories reveals differences and similarities between the two chapters, there are also marked statistical differences in individual interactive subcategories across them.

Transitions
In the case of transitions, per 1000 running words, SIC and SLRC recorded 14.7 and 18.5 respectively, suggesting that SLRC is dense in transition markers than SIC. Table 4 indicates that the observed difference is statistically significant with a log-likelihood (LL) value of 20.61, as against the LL 3.84 statistical significance threshold.
Thus, participation in the MTS is expected to improve the socio-economic conditions of participants compared to that of non-participants as a result of higher yields for participants. (SIC 0001) Conflict theory is problematic because it succeeds in explaining only a portion of social life ignoring the other side. (SLRC 0001) This finding implies that master's thesis writers tend to write a more pragmatically cohesive argument in the Literature Review (LR) than they do in the Introduction. The Sociology apprentice scholars (Lee & Casal, 2014) probably require more of the transition markers in the LR to sythesize the ideas of the various scholars they have reviewed in order to present a more persuasive text. It further reveals the universality and indispensability of transitions in written academic communication. The log-likelihood scores for interactive metadiscourse across SIC and SLRC can be seen in Table  4.

Code glosses
Code glosses are required for reformulating, exemplifying, and defining propositional information in text (Hyland, 2007). It is shown in Table 3 that Introduction (13.5) significantly used more code glosses than the LR (11.7). The more frequent use of code glosses in the Introduction chapter than the LR is quite justifiable since most of the Introduction chapters analyzed had a section captioned 'Definition of terms', which explained the key words either adopted or adapted in the studies. Colonialism is seen as essentially an economic phenomenon, the earliest period of which was associated with…the slave tradethe period of socalled primitive capital accumulation. (SLRC 0002) Unlike the LR where readers may be familiar with the theoretical, conceptual and empirical literature, the Introduction mostly requires the writers' conceptualization of the study and so requires that key terms are clarified, explained, and defined (Hyland, 2004b) to aid readers to comprehend the ideational materials in the text. Unsurprisingly, it is said that in evaluating thesis Introduction one of the key questions raised is whether the researcher has provided conceptual definitions of key terms (Bitchener, 2009).

Frame markers
Frame markers are meant to signal discourse acts, sequences, or text stages. It is indicated in Table 3 that the Introduction and the LR chapters recorded 9.1 and 3.5 respectively, per 1000 running words. Hyland and Jiang (2018) assert that the length of a document positively correlates with the use of frame markers, and that "longer papers require more explicit structuring (p. 24)". However, the Introduction employing greater use of frame markers debunk this assertion, as the LR was found to be lengthier than the Introduction.
Chapter Five concerns itself with the nature of development plans and how they have been implemented. (SIC 0003) The purpose of the review is to help create the grounding for the study of women's perceptions of their roles in households in Yendi, Ghana. (SLRC 0003) Most likely, 'Overview of the thesis', a common section in the Introduction, which announces the focus of proceeding chapters, accounted greatly to the high occurrence of frame markers in the Introduction chapter. Moreover, the highest use of frame markers in the Introduction also shows that writers aimed to achieve brevity, orderliness and clarity in the first chapter since 'frame marking is an effort on the part of the writer to be orderly, brief and clear…' (Abdi et al., 2009, p. 149).

Evidentials
A generally-held position has been confirmed by the predominance of evidentials in the LR chapter (see Kawase, 2015;Salek, 2014). Salek (2014) describes "evidentials as a symbol of 'review of the literature' (p. 55) and this is consistent with Bitchener (2009) who also argues that LRs are dense in evidentials. As expressed by one of the interviewees in Hyland's (2010) study, the LR mainly reports the works of others and so requires a lot of citations to help in the critique and synthesis of ideas.
Part of the village has now moved beyond the boundary into the reserve (Agyemang et al, 2003). SIC 0004 Frank (1967, p. 34) argues that there is a 'chain of dependency' running down from the highly advanced centres of the world… (SLRC 0004) The large use of evidentials in the LR further supports Hyland's (2004a) assertion that 'citation is central to the social context of persuasion in academic writing as it helps provide justification for arguments and demonstrates the novelty of the writer's position' (p. 141). Unsurprisingly, it is shown in Table 4 that the difference between the Introduction and the LR chapters in the use of evidentials is statistically significant.

Endophoric markers
Writers require endophoric markers to make prospective and retrospective references in text. In effect, they facilitate "auto-referentiality of the discourse" (Teubert, 2005, p. 7). It is shown in Tables 3 and 4 that endophoric markers are significantly more frequent in the Introduction than in the Literature Review (LR). The finding supports Kawase's (2015) observation that endophorics typify thesis introduction chapters. The frequent use of endophorics in the Introduction chapter is attributed to the relatively more use of visuals such as tables, and figures. Kawase (2015) then noted that the introduction chapter has a responsibility to present previews of all chapters within a thesis. Hyland and Jiang (2018) argued that visuals attract more use of endophorics as writers are required to often point to them to make information salient to their readers. They then argue that "the increase in these features perhaps indicates greater awareness of readers' processing needs, or at least greater care in constructing arguments" (p. 24).
In Chapter Six, a critical look is taken on how the development plans have been informed by the structuralist paradigms. (SIC 0005) Aluko and Alfa explain that the above observations between the colonized and the colonizer can be used as a situational analysis to the low status of women in the household. (SLRC 0005) The marked difference in the use of endophorics across the Introduction and the LR chapters calls for some concerns. It suggests that writers scarcely refer to other parts of their texts in the LR, given that endophorics in the LR are thrice the occurrences in the Introduction chapter. This marked difference is corroborated by the overwhelming use of frame markers in the Introduction as most of the frame markers also performed endophorics functions since language users employ metadiscoursal elements for polypragmatic functions (Hyland, 2005b). Table 5 presents the raw frequencies (RFs), normed frequencies (NFs), percentages, and ranks of the interactional subcategories in the Sociology Introduction Chapter (SIC) and Sociology Literature Review Chapter (SLRC). It is shown that the scale of preference for interactional subcategories in SIC is: hedges, attitude markers, boosters, engagement markers, continuants, and self mentions. On the other hand, SLRC recorded hedges, boosters, attitude markers, engagement markers, continuants, and self mentions. Some differences and similarities exist between this study and previous ones, especially Rezaei et al., (2015). The interactive Metadiscourse in SIC and SLRC can be seen in Table 5. Apart from self mentions whose rank corresponds with Rezaei et al. (2015), every other rank differs. While hedges ranked first in the present study, it ranked 2 nd in Rezaei et al. (2015), and while boosters ranked 2 nd here, it ranked 4 th in Rezaei et al. (2015). The similarity between the current study and the previous one is largely attributable to the genre factor -the master's thesis. The differences can also be attributed to the varied disciplinary conventions, norms, practices, and epistemologies which invariably have effect on rhetorical choices.

Hedges
The statistics on the interactional subcategories across SIC and SLRC manifests significant variation in all the interactional subcategories.
On hedges, whereas Introduction used nearly 20, LR employed approximately 30, indicating a pronounced difference of almost 10, in the normed frequencies. This finding challenges the assertion that "the introduction chapter of a thesis is heavily hedged because it creates a platform for readers to analyze background of the study which includes aspects such as the contexts, reasons, and purpose of the study" (Seskauskien, 2008 as cited in Haufiku & Kangira, 2017, p. 1) It is estimated that 65.1 million hectares (ha) of forests were destroyed in developing countries between 1990 and1995. (SIC 0006) This, in turn may lower the opportunity cost of inactivity by reducing the foregone current income associated with the educational attendance. (SLRC 0006) The statistically significant difference (see Table 6) between SIC and SLRC implies that Sociology apprentice scholars more often present their facts as opinions in the LR chapter than in the Introduction chapter. Further, they more often "convey deference and respect for readers' views" (Hyland, 2005a, p. 68), and "judicious option for writers as they allow them to mark their claims as provisional and subject to current objections and future revisions" (Hyland & Jiang, 2018, p. 26), as exemplified in SLRC 9, where the verb 'may' is used. Table 6 shows the log-likelihood scores for interactional metadiscourse across SIC and SLRC. The frequent use of hedges in the LR therefore is conventional given that the 'the pool of knowledge' in the literature from which the researchers draw information is also known to the readers whose right to judgement ought to be respected. Thus, Sociology master's theses writers minimize threats (Martin-Martin, 2008;Nivales, 2011) to the face of their readers when critiquing, evaluating and synthesizing the views of previous researchers. This probably explains why boosters used in the LR are less than hedges. Also, the comparatively limited use of hedges in the Introduction suggests that writers in Sociology are careful in conceptualizing their studies with uncertainty, caution and tentativeness (Hyland, 1998a) in order to show deference (Hyland, 2004a) to the views of their readers. Interestingly, the finding supports Kawase' (2015) claim that more use of evidentials requires more use of hedges. Thus, in this paper LR reviewed recorded more evidentials than Introduction, and consistent with the claim of Kawase (2015), more hedges were recorded in the LR.

Boosters
The presence of boosters in a text has implications for writer's degree of commitment to their propositions, or conviction or claims (Hyland, 2004(Hyland, , 2005a. Consequently, a significant use of boosters correlates with an author's self-image. I noticed a significant use of these resources in both SIC and SLRC. Aside from this similarity, there is a pronounced difference between the two chapters, as SIC and SLRC employed 10.3 and 26.7, per 10, 000 words, respectively. This seems to strengthen the observation that there is a positive correlation between hedges and boosters (e.g. Hyland, 2004;Hyland & Tse, 2004;Hyland & Jiang, 2018;Salek, 2014;Jin & Shang, 2016). See the extracts below for the use of boosters.
These psychological states are strongly influenced by certain features of the job called "core job dimensions". (SLRC 0007) Critics conclude that many infants in less developed countries suffered malnutrition and death as a result of promotion efforts by formula and related food products companies. (SLRC 0007) Akbas (2014) assert that the hedges-booster balance helps writers to establish a 'cautious confident' ethos, affirming their balanced stance on the commitment-detachment cline. The finding suggests that Sociology writers show more commitment to their proposition in the LR than in the Introduction. Table 5 shows that the normed frequencies for attitude markers are 11.7 and 19.7 for the Introduction and the LR chapters respectively. The difference suggests that more attitudinal expressions are preferred in the LR to those in the Introduction. This finding is justifiable because criticism which requires expression of attitudes is central in the LR (Hart, 1998). Accordingly, writers are required to express their agreement or disagreement, surprises, etc. in relation to the findings of previous work, the theories, and the methodologies they employed (Creswell, 2003;Bitchener, 2009).

Attitude markers
It interests the researcher to find out the kind of roles women in the conflict zones played during the times of the Konkomba-Nawuri-Gonja conflict. (SIC 0008) The dichotomy between the urban and rural becomes more striking in relation to expenditure on food. (SLRC 0008) Hyland (2005a) argues that attitude markers indicate writers' affective attitude to propositions, as demonstrated in SLRC 13 where the writers find a particular assertion 'more striking' to themselves and possibly to the readers.

Engagement markers
Just like discourse community and discourse itself, writer-reader interaction is not an ontological reality (Teubert, 2005); rather, it is an authorial construct. This construct is sanctioned and conditioned by disciplinary values (Hyland, 2005a and b), and (sub)generic conventions (Olmos-Lopez, 2015). Engagement markers remain part of the direct strategies to realizing writer-reader interaction, which enhances discourse internal negotiation (Teubert, 2005). The present study found that these networking resources are peculiar to both Introduction and LR chapters, thus, corroborating Olmos-Lopez (2015) and Rezaei et al. (2015) studies. The extracts below illustrate the use of engagement markers.
What do the results of the first SSSCE tell us about the attainment of the set objectives? (SIC 0009) On a regional basis, the primary school enrolment statistics from 1994/95 to 1996/97 show that… (see Table 2.3) (SLRC 0009) Aside from the inter-chapter similarity in the use of engagement markers, there exists a significant quantitative difference (see Tables 5 and  6). A significant difference exists between engagement markers in the two chapters, as Introduction and LR respectively employed 2.2 and 0.7. This suggests that Sociology masters' thesis writers appear more reader-sensitive in the Introduction than in LR. Being the 'welcoming' chapter, the Introduction creates a friendly atmosphere for the readers since it determines whether readers may read a text or not (Olmos-Lopez, 2015). Hyland and Jiang (2018) assert that engagement markers "enable writers to explicitly step into the text to focus readers on a particular aspect of the data or argument and guide their interpretations" (p. 27), as evident in extracts SIC 0009 and SLRC 0009. Further, since the Introduction conceptualizes and problematizes the study (Bitchener, 2009), Sociology writers engage their readers in the process but in the LR, the writers may not require the 'presence' of the readers since the writer may decide to 'blow his/her horn' (evidenced by the frequent use of boosters in SLRC, see Table 5), which is somehow expressed by the high use of boosters in the LR.

Continuants
Continuants constitute part of the neglected features of metadiscourse (Jiang & Hyland, 2016). Their neglect by researchers into metadiscourse might have emanated from their limited usage in academic writing. However, this reason is not tenable given that in this study continuants outnumbered self-mentions, whose metadiscurisivity is undoubted (e.g. Hyland, 2004Hyland, , 2005aÄdel, 2006. In this study, it was found continuants were used across the two chapters, although they were more frequent in LR (0.6) than in the Introduction (0.2). The log likelihood significance test also indicates the observed difference is significant (statistically) with LL 6.34. The fact that these devices hitherto had been denied metadiscourse function requires some detailed discussion, as it shows how 'knowledge demand and supply' is displayed in the two chapters. Knowledge demand refers to the amount of information required by readers while knowledge supply implies the quantum of information provided by writers to their potential readers. The statistical difference between the Introduction and LR chapters in terms of continuants implies that Sociology writers consider their readers as co-knowledge producers who are conversant with the literatureresearch and nonresearch (Bitchener, 2009) in the field of study. It, therefore, means that writers are not exhaustive in their knowledge-creation, but rather give room to the readers, thereby relinquishing some writer responsibility on the readers. This indicates that Sociology masters' thesis writers are inclined more to writer responsible style in the LR, and reader responsible style in the Introduction. Continuants are exemplified in the extracts below.
They are charged with the daily upkeep of the house -sweeping…and helping put up buildings, and running countless errands of various kinds. (SIC 0010) This research aimed at discovering whether a variety of socio-economic and school factors (location-urban/rural, … etc.)… (SLRC 0010) The Introduction relatively recognizes the readers as those who are not quite familiar with the field and hence required to be provided with more information than in the LR. This is appreciable because mostly the knowledge in the LR appears to be 'given' or what Woodward-Kron (2002) refers to as 'core disciplinary knowledge' while a chunk of the content in the Introduction looks 'new'. Hence, Sociology thesis writers, to a large extent, move their potential readers from the 'uncommonsense disciplinary knowledge' in the Introduction to the 'common sense disciplinary ones' in the LR, given that the responsibility of readers in knowledge construction in the LR is greater than what is expected in the Introduction.

Self mentions
Authorial presence or visibility is key in theses as they constitute part of the stance taking resources (Ädel, 2005Hyland, 2002aHyland, , 2002b. We notice an interesting case on the use of self mentions across the two chapters. Table 5 indicates that both chapters recorded the same normed frequencies of 0.1, which shows that there is no significance difference (statistically) between the two chapters. This finding is in contrast to Samraj (2008). It also challenges the position that the Introduction chapter is characterized by a high authorial visibility (e.g. Kawase, 2015, p. 122), and thus appears self mention laden.
In view of this it is necessary for the researcher to find out what happened to women during the conflict times. (SIC 0011) I capture this multi-linear approach in a diagrammatical form as illustrated…. (SLRC 0011) The present finding suggests that Sociology masters' theses writers create a balanced authorial presence in their Introduction and LR chapters. The relative similarity in the intrusion of the authorial selves in the two chapters somehow weakens the claims of some scholars who still defend the 'law' of impersonality in academic writing (e.g. Morley, 2015).

CONCLUSION
This paper sought to investigate the use of metadiscoursal expressions in the Introduction and the LR chapters in Sociology masters' theses. The study mainly aimed to find out the variation in the use of interactive and interactional devices across the two chapters. It was discovered that the LR chapter employed slightly more use of interactive devices than the Introduction, and largely more use of interactional devices than the Introduction as well. Specifically, while the Introduction chapter used more code glosses, frame markers, endophoric markers, engagement markers, more than the LR; transitions, evidentials, hedges, attitude markers, boosters, continuants were more frequent in the LR than in Introduction. The variations in the use of metadiscoursal devices in the two chapters are mainly attributed to their respective communicative functions, and somehow their positioning in the thesis.
The findings have a number of implications. Theoretically, it confirms Akoto's (2012) modification to Hyland' (2005a) model of metadiscourse, that continuants are part of the academic language. The discovery of continuants as part of metadiscourse is a step towards helping metadiscourse to achieve theoretical rigour and establish its boundaries more clearly (Ädel, 2005). More so, the study affirms the chapter-specific approach to the teaching of the various rhetorical component of the masters' thesis. Most textbooks (e.g. Murray, 2006;Swales & Feak, 1994) approach the teaching of the thesis from 'macro' perspective without a recourse to the specific communicative function and therefore rhetorical choices of the individual chapters. Finally, it has implication for further research where different chapters can be examined to ascertain variation in the use of metadiscourse devices, drawing on the modified model of Hyland's metadiscourse model.