A Study of Interactional Metadiscourse Features in Chinese University Students' Prepared English Speech

| ABSTRACT Speech is an activity in which speakers convey their message to listeners continuously. A coherent, logical argument must be made by the speaker, and audience participation is essential for a successful speech. The majority of prior research on English Public Speaking (hence referred to as EPS) from the Critical Discourse Analysis (hereinafter referred to as CDA) perspective focused on political or academic speech. Although the speech of English language learners (also known as EFLs) has caught the interest of academics. There has not been much study of the interactional metadiscourse characteristics of EFLs' EPS. In this study, we investigate how the usage of metadiscourse influences speakers' performance and the ways presenters engage with their listeners. We do this by drawing on Hyland's stance and engagement framework from 2005. Based on a corpus of 90 prepared speech scripts from high-scoring and low-scoring groups, the results suggest that explicit mentions of oneself and listeners are the most frequent elements, whereas stance markers are employed more frequently than listener engagement markers. Additionally, this study found a positive correlation between the usage of hedges and speech scores and a marginally negative correlation between speech scores and the use of self-mentions. In terms of technique, the interactional metadiscourse analysis on investigating EPS is provided by this study, while in terms of pedagogy, it emphasizes the significance of interactional metadiscourse in EPS instruction.


Introduction
Speech is an act in which the speaker addresses the audience in a formal speech and expresses their own ideas about the information being presented.On the other hand, the interaction between speakers and audiences also plays a significant role in realizing communicative purposes.
EPS has attracted scholarly interest over the years, but the majority of relevant studies have focused on academic or political addresses.Notwithstanding, studies appeared focusing on EFLs' EPS, which were mainly conducted from a pedagogical perspective, either elaborating the significance and necessity of EPS in English teaching or applying different pedagogical methods to practical teaching, etc.

Literature review 2.1 Empirical research on the Speech and research gaps
In terms of research topics, academic speeches (Bu, 2014; Hu & Zeng, 2007;Recski, 2005; Zeng and Liang, 2007; etc.) and political speeches received a lot of attention in earlier speech literature (Shao, 2017; Zhang, 2017; etc.).For instance, recent studies by Zhang (2016) and Shao (2017) involved speeches delivered by leaders on the 70th Anniversary of the UN to uncover the relationship between speech characteristics and political texts.In an investigation into academic speeches, Bu (2014) examined language strategies used by lecturers to express objectively and persuasively, while E. Sheppard, Elliott, and Baese-Berk (2017) compared the awareness of college EAP teachers and content staff of international students' speech comprehensibility and intelligibility.
This research was conducted to look into the feature of part of speech in university students' EPS; despite more recent emphasis being paid to EFLs' public speech (Wang & Qian, 2009;Wang & Ma, 2009), the construction of speakers' identity (Zhao, Chen, Huang, & Qin, 2012), comparison of the rating differences of foreign and national judges (Wen, Liu, & Jin, 2005) or the link between speech and critical thinking (Sun, Yu, & Wang, 2015; Sun, 2018).
Additionally, there is a relatively small body of literature that is concerned with pedagogical theories to illustrate the significance and necessity of speech in English teaching (Ren, 2007;Zhang, 2007) or applying different pedagogical methods to EPS teaching (Deng, 2019; Ye, 2014; Zhao, 2011).
What's more, previous studies have explored public speaking anxiety, which is an important factor in the effectiveness of public speech.One study by Allan Kelsen (2019) explored the association between personality traits and the public speaking anxiety of university students.
Undoubtedly, there isn't much historical research that focuses on EPS competency in EFL learners.Zhang and Ardasheva (2019) explored the source of 263 adult Chinese English as foreign language learners' self-efficacy, which suggested course experience, gender, and academic major apparently affecting the relationship of self-efficacy to its theoretical sources.Next year, Zhang, Ardasheva, and Bruce (2020) complemented the work and verified a hypothesized model of EPS performance predictors, which provided evidence for the availability of the model and the importance of EPS instructional practices.University students' metadiscourse characteristics have been the subject of several studies.For example, Deng (2010) explored metadiscourse features employed in university students' speeches from EPS contests, while Mei (2013) compared metadiscourse features in Chinese and American college commencement speeches.The research provided metadiscourse features of university students' speech, but the study on revealing the inner difference between metadiscourse features and the relationship between metadiscourse features and moves are scarce.
Overall, the requirement for EFLs' EPS is generally highlighted by these studies.Such studies remain narrow in focus dealing only with linguistic features.The inner language characteristics and correlation between language features and moves in university students' speech are still vague to us.
However, the following conclusions are reported by other researchers who had a foundation in the interpersonal model of metadiscourse. 1) Compared with English speech, other language speech used less metadiscourse.2) The use of metadiscourse in different types of speech is differentiated; for example, Mei (2013) compared 29 Chinese college commencement speeches and 30 American ones, and Peng (2016) chose 60 pieces of G20 and APEC leaders' speeches from 2008 to 2015.
There have also been other viewpoints on this topic, such as discussing political discourse from a pragmatic perspective (Capone, 2010;Yang, 2011), exploring the link between national identity construction and the indexicality principle from a sociocultural, linguistic approach (Shao, 2017; Xin & Li, 2016; Zhang, 2011), investigating discourse metafunction and pattern of speech (Wen, 2017).
When the papers discussed above are combined, it becomes clear that researchers approached the subject from a variety of perspectives, and metadiscourse analysis has garnered interest.However, there are still many unexplored metadiscourse properties.

Research Significance
According to Halliday (1994), language is structured to achieve three metafunctions-ideational function, interpersonal function, and textual function-that work together rather than separately.Drawing on Halliday's metafunctions, many scholars in Applied Linguistics further developed it, among which Hyland (1998) put forward interpersonal function is mainly realized by interactional metadiscourse, which assists lecturers in making statements and meanwhile engages with audiences.And the paramount factors in determining whether lecturers clearly express their views and successfully persuade audiences are lecturers' stance expression and interaction with audiences while delivering the speech.Despite the fact that previous studies examined discourse analysis from the perspective of genre analysis, few studies were carried out comprehensively to reveal the interactional metadiscourse features in speeches' moves.As a consequence, the thesis is of theory and practice significance.One way to test the viability of interactional metadiscourse is theoretically to look for interactional metadiscourse elements in prepared speeches of university students.For another, this study hopes to uncover the relation between interactional metadiscourse features and speech scores in EPS so as to provide implications for future research.Previous studies have examined the metadiscourse features in political or academic speeches, but a scarcity of studies focuses on EFLs' speech.This study, to some degree, hopes to arouse academic attention to the study of EFLs' EPS skills in this field.
Practically, learning the interactional metadiscourse characteristics of an effective speech gives instructors reason to value interactional metadiscourse markers as significant teaching content and explain their relationship to students, which not only helps students deliver a speech with a clear stance expression and interaction with audiences in order to achieve communicative effectiveness.

Research questions
To achieve this purpose, three key research questions were formulated below: 1.To what extent did university students make use of interactional metadiscourse features in their prepared public speeches?2. Is there a correlation between the score of the speech and the frequency, type, and effectiveness of the interactional metadiscourse markers used? 3. What are the differences between high-scoring speakers and low-scoring speakers in terms of frequency, type, and effectiveness of the interactional metadiscourse markers used?

Corpus
The corpus we created includes the 90 transcripts of prepared speeches in the final round of the "21st Century Cup" National English Speaking Contest and the Uchanllege "FLTRP Cup".To examine the differences in metadiscourse use between top and average outstanding speakers, the authors selected 45 speeches of the national first and second runners-up and 45 speeches of the national second and third-place winners for comparison.
The author manually converts the audio recording into textual words, excluding repetition, false starts, hesitation, and filled pause since they are not the focus of our analysis.Then the author checked all the transcribed words verbatim to ensure the accuracy of the transcriptions.At last, the author manually marks the various metaverse languages that appear in the corpus.The reason for selecting the two competitions is that both of them are considered high-level English public speaking competitions, with students' English proficiency closer to those of native English speakers.The length of the prepared speeches ranges from 321 words to 510 words, and the mean length of them is 352 words.This corpus contains 45992 words in total.

Modified Model of Stance and Engagement
We follow the modified model comprising about 500 potential markers of stance and engagement generated by Xuyan Qiu and Kevin Jiang (2019).They not only draw on Hyland's (2005) model of stance and engagement for textual analysis but also piloted the interactional markers with the British Academic Spoken English (BASE) and the Michigan Corpus of Academic Spoken English (MICASE), which are widely used corpora of academic speech.This informed our decision to exclude those categories which are not likely to appear in spoken academic discourse.
Therefore, four important stance resources are utilized in this analysis to describe "an attitudinal dimension," which refers to the methods by which speakers present themselves and communicate their judgments, attitudes, and commitments (Hyland, 2005, p. 176): hedges, boosters, attitude markers and self-mention.
In addition, four kinds of engagement are looked at to see how speakers use rhetoric to draw listeners into their message and move them along with it (Hyland, 2005, p. 178): listener mention, directives, questions and appeals to shared knowledge.
We must keep in mind, though, that interaction is essentially an open concept that may be represented through a variety of modalities, including body language in a spoken presentation.Instead, our concentration is on explicit textual methods because they have been found to be the most pedagogically effective (e.g.Boldt, 2019).While not exhaustive, these items provide a basis for comparing rhetorical resources between speakers.Furthermore, we are less concerned with the assessment process of prepared speech speakers as a contest than how students project themselves and engage the audience with language features

Analysis procedure
AntConc was used to look for the aforementioned interactional markers (Anthony, 2019).Some features are very easily identified through a corpus word-search (we, of course), while others entail a regular expression search (imperatives, it is adj to þ verb).After the concordancing search, we then manually checked the concordance lines containing every occurrence of these items to ensure that they were functioning as markings of either stance or engagement and excluded extraneous examples.This process allowed us to avoid double coding if a word could be seen as having more than one function in context.

Statistical Analysis
The paper provides statistics on the types and frequency of metadiscourse markers used in the corpus after editing the entire document and coding the metadiscourse markers (for research question one).The length of speeches in this corpus ranges from 321 words to 510 words.If the raw data is used to compare only according to word count, the results may be distorted by the length of the scripts.As a result, the number of subcategories interactional metadiscourse was standardized by computing "frequency per 1000 words (hereinafter referred to as ptw)" so as to be compared on a common scale.To address question one, the descriptive analysis was performed to summarize the overall distribution of eight subcategories of interactional metadiscourse markers.
Based on the statistical results, the author conducts a correlation analysis of the resulting statistics using SPSS to determine whether there is a significant correlation between the use of metadiscourse in speeches and the scores of the speakers in terms of content, organization and persuasiveness (for research question 2).Following the statistical analysis, a qualitative analysis of the speech corpus is conducted in order to reflect the general characteristics of high and low-scoring speakers in terms of their metadiscourse use patterns (for research question 3).

The overall distribution of interactional metadiscourse markers used in university students' prepared speeches.
To answer research question one: To what extent did university students make use of interactional metadiscourse features in their prepared public speeches?Table 2 shows the overall distribution of interactional metadiscourse in prepared speeches.

Table 2 Overall distribution of interactional metadiscourse in prepared speeches
As shown in Table 2, it is found that 6130 occurrences of interactional metadiscourse markers appeared in the corpus of 45992 words, which means that there are 133.28cases of interactional metadiscourse resources per thousand words.
Moreover, the frequency of stance markers (86.78 ptw) used in university students' prepared speeches is nearly double the frequency of engagement markers (46.51 ptw), with respectively 65.10% and 34.89 % proportion, which indicates that university students employed much more stance markers in prepared speeches compared with engagement markers.

Percentage in metadiscourse
Frequency per 1000 words Frequency With regard to the stance markers, it is found that self mentions (50.53 ptw) appeared most commonly in university students' prepared speech, which accounted for the greatest proportion (37.91% out of the total amount of interactional metadiscourse markers).In addition, self mentions also rank first in terms of the frequency per text.These findings imply that university students emphasize speaker visibility during delivering the speech, which may stem from speakers in front of the audience delivering the lecture.
Moreover, the percentage and normalized frequency of hedges (15.29 ptw) and boosters (12.65 ptw) are evenly distributed, respectively occupying 11.47% and 9.49% percentage, while the attitude markers (8.31 ptw) is relatively less than epistemic stance, occupying 6.23% proportion, which signifies that the epistemic stance used in the corpus is more than attitudinal stance.
Hedges and boosters are to express the degree of certainty towards the propositional information.Hedges are to withhold commitment and open dialogue, while boosters are to express the certainty of propositional or close dialogue (Hyland, 2005b).
(1) But there seems darkness behind the prosperous technology because… (Text 13) (2) Adults often break their marriage and only find love in….(Text 25) (3) Personally, I love all kinds of music for it...

. (Text 65)
Attitude markers express speakers' attitudes to the proposition (Hyland, 2005b), and there appeared various resources to transmit speakers' attitudes.For example, adjectives like happy, important, wrong, perfect, etc. and attitude verbs and nouns such as prefer, agree, hope, essence, etc., occurred frequently.What's more, sentence adverbs like even, no matter and prepositions like against also appeared to act as attitude markers.
With respect to the engagement markers, it is found that listener mentions (33.48 ptw) appeared most frequently, and the frequency of listener mentions per text (17.71 ptw) occurred commonly.These findings suggest that university students highlight the addressee visibility by using a large number of listener mentions.
(4) We can always find a better ourselves and communicate with other people and even..? (Text 28) (5) When scientists are bragging about generalizing 5G 6G, the humanists say you are providing easier ways to infringe on personal privacy.(Text 56) (6) From the lack of laughter from the audience, I am sure that you all feel the embarrassment I had when I first heard the joke and did not understand it.(Text 43) It is noteworthy that the engagement markers account for nearly 35% percent out of total interactional metadiscourse markers; however, as can be seen in Table 2, listener mentions take up 25.12% while the engagement markers occupy 34.89%, which indicates that university students mainly use listener pronouns to engage with audiences, questions (3.38 ptw), appeals to shared knowledge (9.12 ptw) and directives (0.65 ptw) are employed seldom.Together these results provide that university students use a single way to engage with audiences.Taken together, these results reveal that there appeared to be a high frequency of interactional metadiscourse markers in the corpus, but stance markers are exploited more frequently than engagement markers.Additionally, university students have a balanced way of employing stance markers but a tendency to mainly use listener pronouns to engage with audiences.

Correlation analysis between metadiscourse use and speech scores
Using SPSS software, this paper analyzes the possible correlation between speakers' metadiscourse use and speech scores, and the results show a weak correlation between interactional metadiscourse and speech scores.The data in Table 2 shows that there is a significant difference between the high-scoring and low-scoring groups in the use of interactional metadiscourse markers.The use of attitude markers is 3.25% higher in the high-scoring group than in the low-scoring group (7.75% v.s.4.60%); the use of hedges is 1.64% higher in the high-scoring group than in the low-scoring group (12.23% vs. 10.59%); the use of Booster is 0.82% higher in the high-scoring group than in the low-scoring group (9.89% vs.The use of Booster is 0.82% higher in the high-scoring group than in the low-scoring grouping (9.89% vs. 9.07%), while the use of Self-mentions is 6.93% lower in the high-scoring group than in the low-scoring grouping (34.83% 41.22%).In terms of engagement, the overall number of engagements used by the highscoring group and low-scoring group is comparable, but the specific types of metadiscourse use show significant differences.The use of questions is 1.3% higher in the high-scoring group than in the low-scoring group (4.00% vs. 2.70%), and the use of appeals to knowledge is 2.37% higher in the high-scoring group than in the low-scoring group (10.17% 7.80%); in the use of listener mentions and directives, their frequencies are comparable.
Table 3 Overall distribution of interactional metadiscourse in high-scoring group and low-scoring group

The use and persuasive power of interactive metadiscourse
The high frequency of interactional metadiscourse use in this paper indicates, to some extent, that speakers intend to use more interactive metadiscourse to get the audience's attention, clarify their ideas, and thus enhance the persuasiveness of their speeches.
The use of hedges and speech scores in this paper are positively correlated.The fact that the high-scoring group uses more hedges also indicates, to some extent, that hedges play an important interpersonal interaction role.In a speech, from the time the speaker presents an idea to the time the idea is accepted by the audience, a complex psychological interaction process takes place between the speaker and the audience.It is not an easy task to have a logical, systematic presentation of a point of view in a limited time because the audience often has a critical attitude to examine and judge the correctness, reasonableness and acceptability of this point of view, which makes the speakers are particularly careful with their statement.In order to avoid an overly absolute and subjective viewpoint, speakers often use hedges to slow down the clarity and certainty of their new argument expressions in order to minimize opposition and thus win the audience's acceptance.
From Table 3, it can be found that the use of self-mentions is weakly negatively correlated with speech scores, which indicates, to some extent, that the use of self-mentions may reduce persuasiveness, especially in some academic discourses or formal speeches, and the excessive use of self-mentions is often avoided by speakers' efforts.However, in the corpus of this paper, the self-mentions appear 2324 times, accounting for 37.91% of the total metadiscourse, of which 34.82% of the total metadiscourse is used for highscoring grouping, while 41.22% of the total metadiscourse is used for low-scoring grouping.This frequency is more common in Chinese students' English speech.On the contrary, the frequency of self-mentions is much lower for native English speakers.This may be due to the fact that speakers are keen to tell their own stories and share their lives in their prepared speeches to bring them closer to the audience and elicit empathy, resulting in a higher frequency of self-mentions than native English speakers.The use of self-mentions, on the one hand, in persuasive speeches tends to leave the audience with arguments that flow into personalization and subjectivity, thus negatively affecting the persuasive power of their speech; on the other hand, the use of selfmentions can also make the speech more intimate, thus bringing the audience closer.

Conclusion
In this study, the author has demonstrated how presenters of prepared speeches use a wide variety of interactional techniques to explicitly indicate an authorial perspective while enticing and engrossing audiences.In the corpus, stance markers are substantially more common than engagement markers.In particular, self-mentions had the highest frequency, whereas hedges, boosters, and attitude markers had similar frequencies.This indicated that university students were using their epistemic and attitudinal stances in a balanced manner.Regarding audience involvement indicators, there is a propensity for listener pronouns to be used more frequently than inquiries, appeals to shared knowledge, and instructions.
Also, the results show a weak correlation between interactional metadiscourse and speech scores, which indicates that there is a significant difference between the high-scoring and low-scoring groups in the use of interactional metadiscourse markers.
Furthermore, the author has also observed a close bond between the use of interactional resources and speech score in terms of the use and persuasive power of interactive metadiscourse.The high frequency of interactional metadiscourse use in this paper indicates, to some extent, that speakers intend to use more interactive metadiscourse to get the audience's attention, clarify their ideas, and thus enhance the persuasiveness of their speeches.The fact is that the use of hedges and speech scores in this paper are positively correlated, and the use of self-mentions is weakly negatively correlated with speech scores.

( 7 )( 8 )
Imagine if the robot gets the power of technology; they can also rule the world.So the only thing can distinguish us from other creatures is humanism.(Text 46) What we need to do is to dare to dream about the possibilities in spite of the unchangeable boundaries in front of us.(Text 79)

Table 1
per text