Metadiscourse has been conceptualized in either the broad approach or the narrow approach (Ädel and Mauranen, 2010; Hyland and Jiang, 2022). This research follows a broad approach represented mainly by Hyland, as it is aimed at exploring not only teachers’ discourse organization but also their interaction with students in classrooms. Metadiscourse is defined as the cover term for the self-reflective expressions used to negotiate interactional meanings in a text, assisting the writer (or speaker) to express a viewpoint and engage with readers as members of a particular community (Hyland, 2005). In his interpersonal model, Hyland (2005) divided metadiscourse into two broad categories, interactive and interactional metadiscourse, as shown in Table 1. Although Hyland’s (2005) interpersonal model of metadiscourse is mainly designed to investigate written academic discourse, this model has demonstrated its robustness and effectiveness in exploring the discourse organization and audience involvement mechanisms in spoken academic discourse, in particular teachers’ classroom instructional discourse after slight modification of certain metadiscourse markers (e.g., Lee and Subtirelu, 2015; Zhang, 2017).

In the past decade or so, an increasing number of scholars have noticed the significance of metadiscourse in spoken academic genres (e.g., Zare and Talakoli, 2017; Qiu and Jiang, 2021; Zhang and Lo, 2021; Doiz and Lasagabaster, 2022; Kashiha, 2022; Nazari and Atai, 2022). In particular, some scholars have investigated metadiscourse use in monologic and dialogic types of classroom discourse (Zare and Talakoli, 2017). For example, by adopting Ädel’s (2010) taxonomy of metadiscourse, Zare and Talakoli (2017) compared the functions of personal metadiscourse in academic monologic and dialogic speech, which were represented, respectively, by classroom lectures and discussions. Employing Hyland’s (2005) interpersonal model of metadiscourse, Lee and Subtirelu (2015) made a comparison between teachers’ use of metadiscourse in EAP lessons and academic lectures. These three comparative studies of metadiscourse use in monologic and dialogic types of classrooms share rather similar conclusions. They all show that interactive metadiscourse is used more frequently in monologic speech events, while interactional metadiscourse is more common in dialogic ones due to their respective discourse functions in the two pedagogical contexts. This sheds new light on our understanding of metadiscourse use in classroom contexts. However, very few studies have been conducted in these aspects, and far more research is still needed.

Furthermore, in light of the aforementioned role in assisting lesson comprehension, interactive metadiscourse seems to deserve more scholarly attention in recent years. Indeed, this has drawn increasing scholastic attention in recent years. For instance, based on a corpus of four English medium instruction (EMI) teachers’ interactive metadiscourse use in the Chinese context, Zhang and Lo (2021) examined how different types of interactive metadiscourse expressions are used in university lectures in the science discipline and how these expressions facilitate knowledge construction. They found that transition markers and frame markers were the two most frequently used interactive metadiscourse categories. Within transition markers, the frequently use subcategory marking consequence (e.g., because and so) indicates that explaining was an important feature of classroom instruction. Frame markers were used as an important means to signal the macro-structure and stage of the lectures. Immediately following this, Doiz and Lasagabaster (2022) investigated four English teachers’ interactive metadiscourse in English classrooms in Spain and made a comparison of their research results with those of Zhang and Lo (2021). They demonstrated the overwhelmingly higher frequency of transition markers and frame markers of EMI teachers in Spain against that in the Chinese context and also pointed out some differences in the linguistic realization of specific metadiscourse markers between EMI teachers across Spain and China.

Taken together, such a crucial role of interactive metadiscourse in classroom comprehension, its specific functions in organizing classroom instruction, and the cross-contexts features make it all the more interesting and worthwhile to explore the intricate nature of English teachers’ interactive metadiscourse in various contexts. Following this line of thought, this study concentrates on the interactive metadiscourse (i.e., transitions and frame markers in the present research) used by native English-speaking EAP writing teachers in the United Kingdom and their non-native counterparts in the Chinese contexts and formulates the following two questions:

This research is part of a research project in teachers’ classroom metadiscourse use conducted across the United Kingdom and China. Prior to the data collection, ethical approval for this study was obtained from the Research Ethics and Governance Committee of the authors’ institution to ensure the rights of the participants and the integrity, quality, and transparency of the research. Then a questionnaire was delivered to get access to and select native and non-native English-speaking EAP teacher participants, respectively, for this research in both British and Chinese higher education institutions. Teachers and their students were informed fully about the purpose, methods, and intended possible uses of the research, and what their participation in the research entails. Each teacher participant’s classroom teachings, which last for 90 or 120 min, were recorded for two sessions.

These video recordings were then transcribed verbatim to facilitate further analysis. Altogether 16 sessions of eight teachers’ classroom teaching, two sessions by each of the four teachers from the British and the Chinese contexts, respectively, were selected for this study. Finally, two sub-corpora including the classroom discourse of four native English-speaking EAP teachers in the United Kingdom (ET sub-corpus) and four in China (CT sub-corpus) were compiled. To calculate the normalized frequency, each teacher’s classroom talk was restored in a separate file by eliminating students’ talk and the overall information. The total verbatim transcript of their classroom discourse amounts to 70,073 words. By sifting out student talk, the total amount of teacher discourse is 66,035 words. Among them, 32,860 words constitute the teacher discourse in the ET sub-corpus, and 33,175 words in the CT sub-corpus.

A corpus-based method was employed in the current research to retrieve potential items of metadiscourse, complemented by a manual analysis of each metadiscourse marker to sift out those irrelevant linguistic items. The concordance tool of AntConc was used to observe the immediate context of each metadiscourse item. In view of the context-dependent nature of metadiscourse markers, each linguistic item was judged by the specific function it performs in its particular context and thus warrants the manual identification of metadiscourse items. Moreover, following Ädel (2010), quoted materials and dysfluencies are excluded from the metadiscourse markers. During this process, the two authors crosscheck part of the identified instances of metadiscourse markers, until a final agreement was achieved concerning the disagreed linguistic items.

The identification of metadiscourse markers also takes into account the findings from Hyland’s (2005, 2019) interpersonal model and some other relevant research. This is due to the aforementioned fact that Hyland’s research of metadiscourse use mainly focuses on written discourse. Due to the differences of linguistic features between written and spoken genres (Ädel, 2010), as mentioned before, there are inevitably some instances of metadiscourse markers specific to spoken discourse but not included in Hyland’s (2005) metadiscourse list. Therefore, the current study also makes reference to reported instances from other metadiscourse research into spoken academic discourse, for example, Lee and Subtirelu’s (2015) research into metadiscourse use by EAP teachers and lecturers.

The procedure of data analysis was followed on the basis of pertinent research questions. The frequencies, distributions, and ranges (representing the number of teachers used particular metadiscourse item) of metadiscourse markers used by teachers in the two sub-corpora are the key information needed for this study. Thus, a quantitative discourse analysis was carried out. First, the raw frequencies of certain metadiscourse items can be generated by importing the coded text into AntConc 4.0 and inputting the metadiscourse markers into the search term tool. For example, in Figure 1, the metadiscourse item still, together with its coding mark < Tconj > , is searched by clicking the “start” button. After that, by clicking the Concordance Plot tool at the upper part of the software, an overall landscape of the total number of instances used by individual teachers of this metadiscourse item is generated. However, the number of total instances generated before is the raw frequency of still. Such frequencies are then normalized against per thousand words to generate the normalized frequencies of all the metadiscourse items to make them comparable to each other.

Enlightened by Hyland’s (2005, 2019) interpersonal model of metadiscourse, two categories of interactive metadiscourse, transition markers and frame markers, were selected for analysis in this study. The sub-functional categories or pragmatic functions of these two types of metadiscourse markers and their coding examples are shown in Table 2.

Transition markers mainly refer to conjunctions or adverbial phrases that assist the addressees to better understand pragmatic connections between steps in an argument. Transitions in the current study consist of three pragmatic functions, namely, additions, comparisons, and consequences. An analysis of data reveals that the linguistic realizations of transitions are generally represented by the lexical category of conjunctions and multi-word expressions. Second, frame markers are generally used to mark text boundaries or elements of the schematic text structure, which can be divided into four pragmatic functions. Specifically, they can be used to sequence parts of a text and act as more explicit additive relations, such as first, then, and next, to explicitly label text stages, such as to summarize and in sum, to announce discourse goals, as in my purpose is and I want to, and to signal topic shifts, for example, by using well, right, and now.

This section discusses the considerable similarities in metadiscourse use between teachers in the ET and CT sub-corpora, in the sense of both individual metadiscourse categories and individual lexical items within each metadiscourse category.

In relation to individual metadiscourse categories, teachers in both sub-corpora use transitions and frame markers with comparatively higher frequencies than endophoric markers and code glosses. This finding is in line with Yan (2010) and Lee and Subtirelu (2015). It also evinces that teachers in both sub-corpora attach considerable attention to organizing and guiding students through the classroom discourse at both local (realized by transition markers) and global (realized by frame markers) levels (Chaudron and Richards, 1986; DeCarrico and Nattinger, 1988; Zhang, 2017). Both transition markers, such as and or but, and frame markers, such as first or so, may contribute to a coherent classroom discourse and help signal shifts in discourse trajectories (Crawford Camiciottoli, 2005). Since students in both contexts of the current study are at the stage of learning language skills, teachers in both sub-corpora are sensitive to students’ needs for assistance in navigating through the instructional process. Teachers’ use of such interactive markers may serve as signposts to help relieve the cognitive burden on the part of students in processing the classroom instruction that normally lasts for approximately an hour (Cazden, 2001; Bu, 2014). As such, students would benefit from the effort made by teachers in achieving optimal relevance with minimal processing effort in the interpretation of academic information.

In terms of individual lexical items within each metadiscourse category, the first aspect concerns that both groups of teachers make frequent use of transition markers such as and, but, and so, respectively, to display the additive, comparative, and consequential transitions between discourse segments. In effect, previous research also found that and is so prevalent that it is sometimes left out from rhetorical analysis because it is regarded as the default option of “marking conjunctive relations” of addition (Hyland and Jiang, 2018, p. 21). The current research may develop this view and demonstrate that but and so likewise are so prevalent that they may also be deemed as the default options of expressing comparative and consequential relations in academic speech contexts.

Second, teachers in both sub-corpora use the frame markers then, first, and second with considerably high frequencies to indicate sequencing relations. These sequencing frame markers can serve as an important navigational aid for the students to “locate learning in time and space” (Walsh, 2011, p. 208). This confirms Fung and Carter’s (2007) view that discourse markers like firstly, secondly, and then are used frequently in teachers’ classroom discourse to signal and segment the logical sequence. However, the current research reports exceptionally higher frequencies of then compared with first and second, which seems to be inconsistent with Yang’s (2014) finding that first is predominantly widespread in Chinese teachers’ classroom discourse, sometimes without using subsequent logical connectors like secondly and thirdly. Nevertheless, this might be compensated for by other forms of expressions with a similar meaning to first, either in freestanding forms such as firstly, start, and begin, or in the multiword expression first of all in the current research.

The third feature shared by teachers in ET and CT sub-corpora is the extensive use of framing markers okay, so, and now. These are the top three frequently used lexical items by every teacher in both ET and CT sub-corpora. This is partly in conformity with previous studies (Sacks et al., 1974; Schiffrin, 1987) which found that okay and/or so are common pre-closing devices to open another round of talk prior to conversational closure. Moreover, Carter and McCarthy (2006) also noted that these frame markers marking shifting topics are commonly used at the opening/closing positions of a topic. They can perform both interactive and social function at the same time in classroom discourse (Walsh, 2006; Fung and Carter, 2007). This enriches previous studies on discourse markers (e.g., Hellermann and Vergun, 2007; Evison, 2009). These signposting devices can function as a lubricant in teacher–student interaction to reduce understanding difficulties, incoherence, and social distance between teachers and students (Yang, 2014).

The current research has sought to make an in-depth analysis of two other aspects of variations in metadiscourse use between teachers in the two sub-corpora. The first difference relates to the use of some metadiscourse markers with different functions. The second aspect concerns salient metadiscourse devices which may occur with strikingly high frequencies in one sub-corpus but with very few or no instances in another.

The first difference of transitions and frame markers across teachers in the ET and CT sub-corpora is the use of consequence transition marker so in the ET sub-corpus as opposed to the frame markers first and then in the CT sub-corpus. As mentioned earlier, teachers endeavor to construct a coherent classroom discourse at both the local (realized by transition markers) and global (realized by fame markers) levels. However, results from the current research reveal that the teachers in the ET sub-corpus are more inclined to organize classroom discourse at the local level typically by virtue of consequential relationships. This is realized by the significantly high frequencies of the transition marker so. Yet the case is opposite on the part of teachers in the CT sub-corpus, who mainly frame the sequence of the discourse by the frame markers first and then. This may extend previous research on teachers’ classroom metadiscourse use (e.g., Lee and Subtirelu, 2015; Zhang, 2017) by demonstrating that Chinese EAP teachers and those in the United Kingdom show variations in their logical preferences when delivering classroom instruction. Admittedly, even though transitions and frame markers are employed with varying frequencies in the two sub-corpora, these two categories constitute the major interactive devices.

With regard to the second aspect, an obvious discrepancy concerns the frame marker well, which is used as a shifting topic device. It occurs with a relatively high frequency in the ET sub-corpus, however, with no instance in the CT sub-corpus. Such a discrepancy may be explained by the development order of acquisition (Hays, 1992; Cf. Hellermann and Vergun, 2007). In a study on the use of different types of discourse markers by Japanese learners of English in their first, second, or third year of study, Hays (1992) found that while discourse markers but, and, and so are used frequently, very few learners use well. This led him to speculate that there might exist a developmental order for the acquisition of discourse markers. Discourse markers that are on the ideational plane have greater semantic weight and are taught and used first, whereas those that are more purely pragmatic, interactional discourse markers appear later in the subjects’ speech. This speculation is supported by Trillo’s (2002) corpus-based study that compares the use of discourse marker usage between native speakers and L2 learners of English and also finds that learners of English use the discourse markers well, among others, with a much lower frequency than native speakers.

The aforementioned two sections illustrated the similarities and differences of metadiscourse use between teachers in the ET and CT sub-corpora and explored their possible reasons, respectively. Taken together, however, this vast similarities and differences of teachers’ rhetorical characteristics may be explained under the constructs of discourse community and speech community.

Based on Swales (1990), a discourse community generally has a broadly agreed set of common goals, and its members share a suitable degree of content and discursive expertise. Following this thought, it is arguable that the EAP teachers in both ET and CT sub-corpora can be regarded as belonging to a specific type of EAP teaching discourse community. Members, such as teachers, in one discourse community may demonstrate more or less similar patterns of behavior in their classroom discourse in order to sustain their professional membership, such as the use of metadiscourse markers to help students navigate through the lesson. This may lead to extensive similarities in their metadiscourse use in classroom teaching.

Specifically, this research demonstrated considerable similarities in interactive metadiscourse use across the two groups of teachers, in the sense of both individual metadiscourse categories and individual lexical items within each metadiscourse category. Such similar features may be due to the fact that teachers’ classroom discourse in both sub-corpora, although located in difference cultural and educational settings, falls into the same discourse community. That is to say, they both belong to spoken academic discourse or, more specifically, classroom teaching discourse. This is also corroborated by the fact that these both of two groups of teachers are delivering EAP writing courses, which are specifically selected for the present study to reduce the effects of other factors apart from cultural and educational settings. Previous research (Swales, 1990; Arminen, 2005; Abdi et al., 2010; Lee, 2016) has noticed that specific genres can restrict the discourse conventions of communication. Thus, it might be concluded that the norms and conventions of the classroom teaching discourse genre constrain the metadiscourse use of these teachers, irrespective of their variant cultural and educational backgrounds.

Meanwhile, discourse community is often contrasted with a speech community, which is defined as “a group of people who naturally share a language (e.g., native speaker of English) in terms of grammar, lexicon, etc.” (Abdi et al., 2010, p. 1670). In other words, a speech community refers to a group of people whose membership is naturally formed due to factors such as geographical locations and largely cannot be chosen. Specifically, this research also demonstrated considerable discrepancies in the metadiscourse use between teachers in the ET and CT sub-corpora in terms of both some metadiscourse markers with variant or roughly contrastive functions, and some salient metadiscourse devices which may occur with markedly high frequencies in one sub-corpus but with very few or no instances in another. In addition to the aforementioned possible reasons such as logical preferences and development order of acquisition, another more general reason may be that teachers in the two sub-corpora belong to two different speech communities.

As noted by Abdi et al. (2010), people may fall into different speech communities due to their geographical locations and cultural backgrounds, which cause them to share a language and cultural norm that differ from another group of people. Among other factors, the former group of teachers belongs to a part of the Western world in which English is spoken as the first language, whereas the latter belongs to the eastern world in which English is learned and spoken as a foreign language. Such a discrepancy in the speech community may result in concomitant distinctive features to differentiate them from each other. Moreover, previous research (Swales, 1990; Abdi et al., 2010; Lee, 2016) argues that specific genres have a bearing on discourse use, the present research may further demonstrate that variation in speech communities can also lead to discrepancies in patterns of communication.