Theoretical background: DMs are multifunctional units which serve to establish cohesive ties in spoken and written discourse, creating an interactive relationship between speakers and listeners (Iglesias Moreno, 2001). These units play an essential role in the pragmatic and discourse competence of speakers, which has made them a popular research topic over the last few decades. Due to the diversity of approaches used to analyse these units, there is a lot of disagreement in theoretical accounts when it comes to the terminology, definitions, and classifications of DMs. However, most authors agree on some basic features of DMs. They are considered essential for creating links between utterances within the discourse, as well as between the utterance and wider context. They do not influence the truth-conditions of the utterance, as they are not a part of its propositional content, and they do not add to its semantic content, which leads to their weak connection to the utterance (Müller, 2005). Furthermore, DMs cannot be “placed” into a specific word class, as they stem from different word classes (e.g. adverbs, expressions, verbs, non-lexical elements), and their interpretation depends on the linguistic and extralinguistic context. They direct communication, allowing the listener to reach the right conclusions (Iglesias Moreno, 2001). Due to their importance for structuring the discourse and maintaining the interactive links in speech, the knowledge of DMs is crucial in order to use language in a way that is culturally, socially and situationally appropriate (Celce-Murcia & Olshtain, 2000). The importance of DMs for natural and context-appropriate communication indicates their relevance for foreign language (FL) learners. Moreover, FL learners can profit from the use of these units as they aid listeners in interpreting utterances and can also help them gain valuable thinking time (Müller, 2005). Despite these benefits, research has shown that non-native speakers generally use DMs less frequently and with less diversity than their native counterparts (Fung & Carter, 2007; Gilquin, 2016; Hellerman & Vergun, 2007; Iglesias Moreno, 2001; Müller, 2005; Romero Trillo, 2002). Their speech is often characterised by overuse of specific markers (e.g. Müller, 2005) or the use of non-targetlike functions of these units (e.g. House, 2009). However, in spite of these differences, studies have also shown that L2 speakers are able to use DMs efficiently, depending on a variety of factors – the proficiency of speakers, exposure to the target language as well as the nature of individual DMs (Buysse, 2010). The importance of DMs has also been recognised in the academic and classroom context, where they have been proven to be relevant for the comprehension and organisation of lectures (Jung, 2006; Swales & Malczewski, 2001). DMs also play an important role in the speech of foreign language (FL) teachers. In most cases, the FL classroom represents a very intensive environment, in which DMs aid teachers in marking transitions, holding the learners’ attention, and establishing better interpersonal relationships in the classroom (Walsh, 2013). On top of that, teachers represent an important source of linguistic input, and the quality of their speech influences the quality of interaction in the classroom and the availability of opportunities of acquisition (Sešek, 2007). Interaction is one of the preconditions for the successful acquisition of an FL, good organisation of the class helps learners’ understanding, and DM use can facilitate the achievement of these goals. The present study: Despite the importance of teacher talk and the essential role played by DMs within it, studies of these units in the speech of teachers is scarce. In the Croatian context, to our knowledge, there are no extensive studies focusing on the DM use of EFL teachers. As studies of teacher talk represent a valuable reflection of the existing practice in Croatian schools, and in view of the importance of DMs as a tool for organising teacher talk as well as achieving important goals in the classroom context, the aim of this thesis is to offer a detailed insight into the use of DMs by eight Croatian teachers of English as a foreign language (EFL). The individual aims of the thesis were as follows: 1) to determine the frequency of use and the types of DMs in the speech of the teachers in classroom discourse; 2) to examine the use of DMs in the speech of the teachers in relation to the students’ level of proficiency (primary and secondary school); 3) to determine the differences in DM use in the spoken production of the teachers in classroom discourse and the informal interview; 4) to examine the relationship and possible connections between DM use, sociodemographic characteristics of the participants (age, work experience, exposure to English) and their attitudes towards DMs; 5) to determine the functional distribution of most frequent DM with a special emphasis on their functions in classroom discourse; 6) to examine the attitudes of the teachers on the use of DMs and their importance for classroom discourse; 7) to determine the extent to which the textbooks the participants use in their classes offer the students and teachers information about DMs. On the basis of the stated aims and previous research, the following hypotheses are formed: H1 Croatian EFL teachers will use DMs less frequently than native speakers in classroom discourse (comparison with the CANCODE corpus). H2 Teachers working with secondary school students will use DMs more frequently than those working with primary school students and will use a greater number of different DMs. H3 The teachers will use some DMs in the informal interview that will not appear in their classroom discourse. H4 Teachers who are more exposed to English outside the classroom will use DMs more frequently in the classroom and will use a greater number of different DMs. H5 Teachers with more work experience will use DMs more frequently and will use more different DMs, especially those typical for the organisation of classroom discourse. H6 Teachers whose attitudes indicate a higher awareness of the importance of DM use will use them more frequently in classroom discourse. H7 The teachers will use certain DMs in a way that they are not used by native speakers or in ways not previously described in literature. Participants, methods and instruments: To achieve the abovementioned aims and test the hypotheses, the study is focused on the classroom talk of four teachers working with students in the final years of primary (A2 level of English) and four in the final years of secondary school (B2 level of English). It is based on a corpus of teacher talk which consists of ten lessons recorded for each of the eight teachers. The teachers are all L1 Croatian speakers who have graduated in English language and literature from Croatian universities. At the time the study was conducted (spring 2018), the average age of the teachers was 40.6 (min 27, max 56, SD 10.2), and they had an average work experience of 15.4 years (min 2, max 26, SD 15.4). DM studies are commonly corpus-based, as corpora enable the examination of real language in use. The advantage of using a small corpus, designed specifically for this study, is that it allows for the examination of the required contexts and features of speech, and it also allows for manual analysis which is important not only for determining the functions of individual DMs, but also the very status of a linguistic unit as a DM. The smaller number of participants and the larger number of recorded lessons enable us to pay more attention to the specific characteristics of the teachers’ DM use, which is especially important as our focus is on the qualitative analysis, with the aim of examining a larger number of examples to gain insight into the functions of the frequent DMs. The study uses a mixed methods approach, starting from a more general quantitative examination of our corpus using statistical methods, while the main part of the study consists of a more detailed qualitative analysis of the five most frequent DMs in the classroom discourse of the teachers. The qualitative analysis is based on the methods used in conversation analysis, i.e. the analysis of spoken interaction in its natural form with the aim of examining the organisation of interaction and revealing the mechanisms that it is based on (Wong & Waring, 2010). A simplified transcription system, the so-called Jefferson light (Hepburn & Bolden, 2017) was used in the present study (Appendix 5). The lessons (3204 minutes) were recorded and transcribed. This material constitutes the main corpus, consisting of 192,027 words. The participants completed a questionnaire (Appendix 1), the first part of which consisted of sociodemographic questions, while the second part was designed to investigate the teachers’ attitudes towards DMs. This was followed by an informal interview held with each participant separately in English. The aim of the informal interview was two-fold: to gain more detailed insight into the participants’ attitudes using supplementary questions (Appendix 2), and to gain insight into the teachers’ speech in an informal context. The informal interviews make up the supplementary corpus (162 minutes/10993 words). The analysis of the corpora was based on the existing lists and descriptions of DMs by several authors (Aijmer, 2004; Brinton, 1996; Fung & Carter, 2007; Müller, 2005; Othman, 2010; Swan, 2016). The DMs were identified following three conditions (Fuller, 2003b; Liao, 2009): DMs signalise the relationship of prior discourse to the following discourse; the semantic relationship between elements connected by a DMs remains the same is the DM is omitted; the utterance remains grammatically intact if the DM is omitted. The corpora were analysed manually, as the identification of DMs depends on the context. The final list of extracted DMs can be found in Appendix 3. The qualitative and quantitative analysis was based on the functional framework of DMs in pedagogical discourse by Fung and Carter (2007). The quantitative portion of the study focused on the first four research aims, with statistical procedures (frequencies, t-tests, correlations) used to gain a deeper insight into the corpus. The more extensive part of the study included a detailed qualitative analysis of the five most frequent DMs in the corpus based on examples of their use extracted from the corpus. Special emphasis was placed on determining functions typical for the organisation of classroom discourse. The sixth research aim was tackled by providing an insight into the attitudes of the teachers towards DMs on the basis of the teachers’ responses from the questionnaire. These were elaborated on by the teachers in the course of the informal interviews and presented together. Finally, the last aim was fulfilled through a brief overview of the textbooks used by the teachers in the recorded classes with a focus on how DMs are presented within them. The textbooks were analysed by counting the pages containing DMs in speaking-focused activities (e.g. dialogues), as well as those containing material directing students in the use of DMs. Results and discussion: In line with the first aim of the study, the quantitative analysis of the corpus has shown that Croatian teachers used DMs in their classroom discourse with the frequency of 116.76/1000 word, with emphasised differences between individual participants (see Table 8, min 75.71, max 187.56, SD 41.10). The differences can be ascribed to the personal style of the speakers, which is a feature of speech also found in native speakers (Gilquin, 2016). It was determined that the teachers use a wide range of DMs in their classroom discourse, with 61 different units extracted from the corpus (see Table 7), which is comparable to the native speakers and advanced students analysed in the study by Neary-Sundquist (2014). The teachers also differed when it came to the number of different DMs they utilized (min 35, max 51, SD 4.96). However, only a limited number of these DMs (18) are used with a frequency greater than 1/1000 words. Five DMs (okay, mhm/uh-huh, yes, and, so) stand out with their frequencies and are also among the most frequent markers in the supplementary corpus. The wide range of functions of the extracted DMs is evident from Table 32. The highest number of them belong to the interpersonal and structural categories, such as expressing confirmation or organising discourse, which can be considered functions typical for the classroom context. This is in line with Hellermann and Vergun’s (2007) conclusions that the classroom context leads to a prevalence of DMs serving primarily for the organisation and management of discourse. The differences between the individual teachers are partially caused by their preferences for specific DMs. For example, the DM okay is the most frequently used marker for teachers OŠ1 and OŠ2. However, while OŠ1 uses it with a frequency of 51.5/1000 words, OŠ2 uses it almost five times less frequently (11.86/1000 words). Similarly, the marker mhm/uh-huh is the most frequent marker for both SŠ2 and SŠ3, with frequencies of 53.01/1000 and 18.8/1000, respectively. The comparison with the native speakers (comparison with the pedagogical sub-corpus of the CANCODE corpus as described by Fung and Carter (2007)) showed that native speakers use 70% (16 of 23) of compared DMs more frequently. This means that our hypothesis H1 has been confirmed. Of the six DMs used more frequently by our participants, okay and yes stand out as the differences exceed 15/1000 words. The latter is evidently commonly used by our teachers as a replacement for yeah, which is more frequent in the native speaker corpus. The high frequency of okay in our teachers’ discourse could be explained by the fact that our corpus contains exclusively teachers’ speech, which might mean that their role as managers of discourse results in the high frequency of this marker, called “the lecturer’s okay” by Levin and Gray (1983). Students of English as a foreign language also require more explanations and support (Aijmer, 2011), which could explain the high rate of use of this marker. On the other hand, this unit is also commonly used in Croatian, and it is a marker with a variety of functions, which might mean that our teachers prefer it for its familiarity and ease of use. In line with the second aim, the DM use of primary and secondary school teachers was compared. The results of the analysis show that there are no statistically significant differences between the two groups when it comes to the frequency and the diversity of DM use, i.e. the students’ level was not shown to be relevant to the teachers’ DM use. This means that the hypothesis H2 was not confirmed. However, a more detailed insight into the corpus indicated that there are some differences, with the secondary school teachers using more DMs that are typical for speech outside the classroom (e.g. actually, well). The differences in the frequencies of DM use are lower than expected, which might be due to the individual preferences of the teachers or the fact that both advanced and beginner learners can benefit from teachers’ use of these units. In line with the third aim, a comparison of the main and supplementary corpora was performed, showing statistical differences regarding the teachers’ frequency and diversity of DM use. The teachers used DMs more frequently in informal interviews, while they used a greater diversity of these units in the classroom. However, the exclusive occurrences of DMs in the supplementary corpus were negligible, which means that our hypothesis H3 was not confirmed. Finally, in line with the fourth aim, no correlation has been found between the frequency and diversity of DM use and the teachers’ age, years of work experience, levels of exposure to English outside the classroom or their attitudes towards DMs, which means that our hypotheses H4, H5 and H6 were not confirmed. There was not much variance in the teachers’ levels of exposure to English outside the classroom as well as their attitudes towards DMs, which was probably the reason why these factors were not related to the frequency and diversity of DM use. All in all, it seems that individual factors, which were not explored in the present study, were more influential in this regard. The second, more extensive part of the study contained a detailed analysis of the most frequent markers in our corpus. The goal of this analysis, in line with the fifth aim, was to provide insight into how these markers are used by the teachers to structure and manage classroom discourse. Of the five markers, okay and so are commonly described as features of the speech of lecturers and teachers (e.g. Biber, 2006; Othman, 2010; Rendle-Short, 2003), while the remaining markers (mhm/uh-huh, yes, and) were not previously extensively described within this context. The teachers used all five markers with a wide range of functions in all four categories described by Fung and Carter (2007) – structural, referential, interpersonal and cognitive. All five DMs, but especially okay, and and so, were found to have functions in the structural category – managing topics, marking transitions etc. DMs okay and so have been found to function in similar ways when it comes to focusing the students’ attention on instructions, explanations and relevant information. These functions are especially relevant in the EFL classroom, where communication mostly takes place in a language that is not the learners’ L1. DMs and and so are also used to introduce questions, while and is generally used to mark the connection between steps or their sequence, reflecting the way a class is planned by the teachers and thus the way discourse and interaction is used to achieve set goals. The DM so is also used for summarising, reformulating and introducing examples and is frequently used by our teachers to explain a task, an unfamiliar word or similar. The DMs okay, mhm/uh-huh and yes have numerous functions in the interpersonal category, where they are used to acknowledge and evaluate students’ responses. However, in the third step of the IRE/F (initiation, response, evaluation/feedback) structure they also help teachers perform more complex tasks, allowing them to introduce recasts, repetitions, reformulations and corrections of the students’ contributions, at the same time providing them with valuable feedback. The DM mhm/uh-huh, and to some extent okay and yes, are also used by the teachers as backchannel signals, to provide support for students’ utterances. All five markers were also used in the cognitive category, helping the teachers gain thinking time, especially at transition points. The train of thought of the speakers is commonly reflected in the use of these markers, while they are also used to self-affirmation (yes, mhm/uh-huh) and selfcorrection (yes) as well as for keeping one’s turn in discourse (and, so). Finally, the lowest number of functions of these markers has been found in the referential category. Only and and so have functions in this category, expressing relations such as result/consequence, contrast (and), addition/elaboration (and) and conclusion (so). Some of the described functions of DMs have not been previously separately or comprehensively described in literature. For example, the teachers use the DM yes with a wide range of functions which are often ascribed to yeah in literature, which is a marker avoided by our teachers in the classroom context. The DM mhm/uh-huh was used less by our teachers as a marker of passive listening and more as an active marker of acknowledgement, evaluation, and support for the interlocutor. The DM okay is also used by the teacher with an especially wide range of roles specific for the classroom environment, especially to manage interaction in the classroom e.g. by giving turns or interrupting disturbances such as students talking out of turn and similar. There were other instances in which the use of a specific DM by an individual teacher could be described as atypical (e.g. the frequent use of ah by OŠ1) or influenced by L1 (e.g. the use of very well for evaluation by OŠ3). However, these were rare, and we can say that our hypothesis H7 was only partially confirmed. In line with the sixth aim, the attitudes of the teachers towards DMs were shown to be predominantly positive, which was largely confirmed in the informal interviews. The teachers showed their awareness of the roles these units play in the organisation of discourse and maintaining interpersonal relations in the classroom. On the other hand, the teachers expressed certain prejudices towards DMs, often describing at least some of them (such as like, you know, yeah) as pertaining to informal speech and thus not appropriate for the classroom environment. In that regard, the teachers also mentioned specific DMs as problematic, regarding them as signs of disfluent speech which is of lower quality. Although this is to some extent true, disfluencies are normal in everyday speech and should be treated as such. The results point to the need to raise awareness of how DMs can actually help speakers deal with disfluencies in a natural way by, for example, indicating they need more thinking time. The teachers were unanimous in expressing that DMs should be included into teacher education programmes in a more comprehensive manner. Finally, in line with the seventh aim, the analysis of the textbooks used in the recorded classes showed that DMs appear more frequently in the two analysed primary school textbooks (23% and 19% of pages) than in the three analysed secondary school textbooks (6% - 11%). However, the primary school textbooks contain only a very narrow range of DMs, and these are not in any way emphasised. In the secondary school textbooks DMs do occasionally appear on the lists of expressions to be used by students in tasks related to spoken interaction, although they do not contain any explicit directions for their use. Conclusions, implications, and limitations: In our study, a mixed methods approach was used which allowed us to gain insight into the use of DMs of Croatian EFL teachers from two perspectives. The quantitative part of the study focused on teachers as a source of linguistic input in the classroom, and the overall aim of this part was to determine if their language presents a rich source of DMs, comparable to the way they are used by native speakers. Although teachers are clearly not the only source of input for the students, and that what is offered to them in the input is not necessarily going to be acquired, teacher talk is still of great value as it should ideally be well-structured and directed to the needs of a specific group of learners (Sešek, 2007). From this perspective, the language of the teachers is not particularly rich in frequent occurrences of diverse DMs, especially at the secondary level. The second perspective was explored through the qualitative analysis in which the language of the teachers is observed as the engine of interaction which is the basis of everything taking place in the classroom, including learning. From this perspective, the teachers are successful at using the most frequent DMs, which reflect the teachers’ care for the structuring of interaction, planning the class, facilitating the learners’ understanding and scaffolding learning. In other words, the main functions of teacher talk are, to some extent, realised through their use of these units. Through the use of DMs, the teachers show their understanding of their students’ needs and the importance of their participation in interaction. The teachers use their language and DMs within it to encourage their students to speak and participate in the foreign language, using an inductive approach instead of always supplying answers and maintaining a positive atmosphere. On the other hand, the analysis also revealed some less favourable procedures such as frequent repetitions, the impact of which is still unclear, comprehension checks or the use of this units for simple evaluation that might restrict the possibilities for further interaction in question-answer sequences. The implications of this study primarily relate to the need for including DM-related topics in the education of present and future EFL teachers in Croatia. This should be tackled in three segments. The first should focus on the importance of DMs for speakers of English in general, and specifically for EFL learners. Teachers could be encouraged to analyse corpora of fluent speech in different contexts in order to notice the ways these units are used and how they contribute to fluency and interaction in general. Secondly, teachers should be instructed in how to teach these units by using more authentic materials in the classroom and drawing their students’ attention to discourse markers and their effect on speech. Of course, this should be done in an age-appropriate manner, focusing on activities such as role-play and games which could help learners acquire the natural use of these units in speech. Finally, awareness needs to be raised on the importance of DMs for teacher talk and the organisation of classroom interaction. In our study, we have described a whole range of vital functions of these units in the speech of teachers, and the teachers themselves have stated that insight into their own classroom practices has been beneficial. Thus, practicing and future teachers would greatly benefit from similar opportunities to analyse their own speech and the speech of other teachers, with a special emphasis on discourse markers and how they can be used to facilitate understanding and organise interaction in that context. The limitation of the present study lies in the small sample, meaning that the results of the statistical analysis cannot be generalised to the wider population. However, the main focus of the study was the detailed qualitative analysis, which was one of the main reasons why the sample was small. The second limitation lies in the fact that the study only covers the speech of the teachers, which prevents us from making any firm conclusions about its effect on the students. Despite the stated limitations, we believe that the study provides a needed and interesting insight into how Croatian EFL teachers use discourse markers in their classroom discourse, and, even more broadly, how these units can be used to structure this discourse and direct and support interaction. There are two main conclusions of this study. The first is related to the deficiencies that were found in the sense of the limited range of DMs used frequently in the teachers’ speech and the neglect of these units in the analysed textbooks and the education of teachers in general. The second conclusion is that the most frequent DMs are used by our teachers in a way that illustrates the diverse possibilities of these units and provides information on their importance for classroom discourse in general. Although many of these functions have been previously described in literature, our results are valuable because they are based on a comprehensive corpus and, as such, they provide a more detailed insight into the specificities of their use in the classroom context. Finally, further research is necessary to better explain the impact of DM use on EFL learners in Croatia. These studies should examine the DM use of Croatian EFL learners, and additional comprehension tests could be used to investigate whether the teachers’ use of these units has an effect on students’ comprehension, in addition to examining their potential emotional effect on the students.