2.1Functions of PMs and their empirical studies

Discourse markers or PMs in recent studies have undergone different and convoluted functional classifications (see Aijmer 2013Aijmer, Karin 2013 “Analyzing Modal Adverbs as Modal Particles and Discourse Markers.” In Discourse Markers and Modal Particles: Categorization and Description, edited by Kirsten Fischer, 89–106. Amsterdam: John Benjamins. ; Brinton 1996Brinton, Laurel J. 1996 Pragmatic Markers in English: Grammaticalization and Discourse Functions. Berlin: Mouton de Gruyter. ; Fung and Carter 2007Fung, Loretta, and Ronald Carter 2007 “Discourse Markers and Spoken English: Native and Learner Use in Pedagogic Settings.” Applied Linguistics 28(3):410–439. ; Maschler 1994Maschler, Yael 1994 “Metalanguaging and Discourse Markers in Bilingual Conversation.” Language in Society 23(3):325–366. ; Redeker 1990Redeker, Gisela 1990 “Ideational and Pragmatic Markers of Discourse Structure.” Journal of Pragmatics 14 (3):367–381. ) as “there is apparently no consistency” among such functionally based taxonomies, and “studies of individual pragmatic markers yield conflicting and confusing results” (Brinton 1996Brinton, Laurel J. 1996 Pragmatic Markers in English: Grammaticalization and Discourse Functions. Berlin: Mouton de Gruyter. , 37). To name a few, Redeker (1990)Redeker, Gisela 1990 “Ideational and Pragmatic Markers of Discourse Structure.” Journal of Pragmatics 14 (3):367–381. classifies discourse marker into ideational marker and PM, which is reckoned to better delineate discourse coherence via representations of both proposition and non-proposition. Ideational markers refer to propositional devices serving semantic relations, while pragmatic markers relate to non-propositional devices serving pragmatic relations such as attitudinal and intentional ones. The former cover simple connectives, semantically rich connectives, and other temporal adverbials. The latter involve pragmatic employments of conjunctions, connective employments of interjections, and discourse-structuring employments of comment. Maschler (1994)Maschler, Yael 1994 “Metalanguaging and Discourse Markers in Bilingual Conversation.” Language in Society 23(3):325–366. then categorizes discourse marker into five types: referential, interpersonal, medial, structural, and silential. Drawing on Maschler (1994)Maschler, Yael 1994 “Metalanguaging and Discourse Markers in Bilingual Conversation.” Language in Society 23(3):325–366. , Fung and Carter (2007)Fung, Loretta, and Ronald Carter 2007 “Discourse Markers and Spoken English: Native and Learner Use in Pedagogic Settings.” Applied Linguistics 28(3):410–439. propose their classification, i.e., referential, interpersonal, structural, and cognitive. Overall, “the central function of PM is to express the relation or relevance of an utterance to the preceding utterance or to the context” (Brinton 1996Brinton, Laurel J. 1996 Pragmatic Markers in English: Grammaticalization and Discourse Functions. Berlin: Mouton de Gruyter. , 30) while some secondary functions in the literature are categorized as structural or interactive, or as response signals or a means to achieve conversation continuity (Brinton 1996Brinton, Laurel J. 1996 Pragmatic Markers in English: Grammaticalization and Discourse Functions. Berlin: Mouton de Gruyter. ).

The above functions of discourse markers or PMs are too variegated to pursue a general framework for application studies. Despite this, we can sort out some PM functions from researchers who share compatible classifications pertinent to our study. As can be seen in Fung and Carter’s (2007)Fung, Loretta, and Ronald Carter 2007 “Discourse Markers and Spoken English: Native and Learner Use in Pedagogic Settings.” Applied Linguistics 28(3):410–439. four categories above, the first three functions are based on Maschler (1994)Maschler, Yael 1994 “Metalanguaging and Discourse Markers in Bilingual Conversation.” Language in Society 23(3):325–366. , while the latter three are analogous to Bazzanella’s (2006)Bazzanella, Carla 2006 “Discourse Markers in Italian: Towards a ‘Compositional’ Meaning.” In Approaches to Discourse Particles, edited by Kerstin Fisher, 449–464. Amsterdam: Elsevier. three macro-functions, viz. interactional, metatextual, and cognitive functions. Specifically, their referential function concerns such propositional conjunctions as because, after, and whose, analogous to Redeker’s (1990)Redeker, Gisela 1990 “Ideational and Pragmatic Markers of Discourse Structure.” Journal of Pragmatics 14 (3):367–381. ideational function. The interpersonal function denotes one of the mechanisms that marks spoken grammar’s affective and social functions (e.g., yeah, you know, well). The structural function indicates the discourse in progress, the presence of which may affect the subject under discussion or even the distribution of turn taking at either textual or interactional level (e.g., yeah, well). The cognitive function signifies constructing a cognitive state of speakers, a mental representation of discourse (e.g., well, you know).

Given the characteristics of non-proposition and multifunctionality, we employ PM, a broad cover term, following Aijmer and Simon-Vanderbergen (2006)Aijmer, Karin, and Anne-Marie Simon-Vandenbergen 2006 “Introduction.” In Pragmatic Markers in Contrast, edited by Karin Aijmer, and Anne-Marie Simon-Vandenbergen, 1–10. Amsterdam: Elsevier. who hold that PMs are not only devices associated with discourse and textual functions but also signals guiding the addressee’s interpretation in the communication situation. Based on Fung and Carter (2007)Fung, Loretta, and Ronald Carter 2007 “Discourse Markers and Spoken English: Native and Learner Use in Pedagogic Settings.” Applied Linguistics 28(3):410–439. and others, we concentrate on the three pragmatic functions (i.e., interpersonal, structural, and cognitive) herein, excluding the referential function that is proposition-based and unfit to our purpose. Hence, we solely review some conflicting and confusing findings on pertinent PMs in empirical studies below, with a focus on three PMs (i.e., yeah, well, you know).

In Müller’s (2005)Müller, Simone 2005 Discourse Markers in Native and Non-native English Discourse. Amsterdam: John Benjamins. findings, NSs (Americans) utilize much more markers (like, you know) than NNSs (German) whereas the NNSs deploy well twice as often as the NSs in its discourse marker functions.11.Müller (2005Müller, Simone 2005 Discourse Markers in Native and Non-native English Discourse. Amsterdam: John Benjamins. , 138) classified well into six functions at textual and interactional levels respectively. Textual: searching for the right phrase, rephrasing/correcting, quotative well, move to the main story, introducing the next scene, conclusive well; Interactional: indirect answer, direct answer, response to self-raised expectations, contributing an opinion, continuing an opinion/answer, evaluating a previous statement. Fung and Carter (2007)Fung, Loretta, and Ronald Carter 2007 “Discourse Markers and Spoken English: Native and Learner Use in Pedagogic Settings.” Applied Linguistics 28(3):410–439. then find that Chinese NNSs’ markers are mostly referential while their employment of yeah, well, you know is much less frequent than NSs’ and restricted in terms of the category of function. More recently, Ament et al. (2018)Ament, Jennifer, Carmen Pérez-Vidal, and Júlia Barón Parés 2018 “The Effects of English-medium Instruction on the Use of Textual and Interpersonal Pragmatic Markers.” Pragmatics 28(4):517–545. examine the effects of full-English media teaching and semi-English media teaching on the application of interpersonal and textual PM functions in the writings of Spanish NNSs at different tertiary levels. Their results reveal that despite the significant gain of the full-English teaching group over the semi-English media teaching group, neither group reached the NSs’ frequency for the utilizations of interpersonal PMs, and both groups displayed more employments of textual PMs than the NSs did. Likewise, Buysse (2018)Buysse, Lieven 2018 “ ‘It Was a bit Stressy as well Actually’. The Pragmatic Markers Actually and In Fact in Spoken Learner English.” Journal of Pragmatics 11(4):1–13. finds that Dutch-speaking NNSs of English deploy much more actually and in fact than NSs while French-speaking NNSs utilize extremely more in fact but less actually than NSs, the findings of which betray an incidence from cognates in the learners’ first languages (L1s), i.e., their cognitive context or background knowledge.

One point to note in the above-reviewed studies is that all the researchers conduct a cross-sectional or immediate effect study. Such studies have been predominated in the literature, while longitudinal and quantitative studies that can examine PM changes over time are relatively deficient, as claimed by some researchers (Jones 2016Jones, Christian 2016 “Teaching Spoken Stance Markers: A Comparison of Receptive and Productive Practice.” The European Journal of Applied Linguistics and TEFL 5(2), 83–100.; Jones and Carter 2014Jones, Christian, and Ronald Carter 2014 “Teaching Spoken Discourse Markers Explicitly: A Comparison of III and PPP.” International Journal of English Studies 14(2), 37–54. ; Magliacane and Howard 2019Magliacane, Annarita, and Martin Howard 2019 “The Role of Learner Status in the Acquisition of Pragmatic Markers during Study Abroad: The Use of ‘Like’ in L2 English.” Journal of Pragmatics 146 (6):72–86. ). Given this factor, some researchers have commenced to fill the lacuna. Tracking a Chinese adult learner of English over a year in a naturalistic setting (via the researcher’s conversations with the adult learner at home), Polat (2011)Polat, Brittany 2011 “Investigation Acquisition of Discourse Markers through a Developmental Learner Corpus.” Journal of Pragmatics 43(15):3745–3756. finds that the learner’s markers (you know, like, well) are limited to certain use frequencies, with well not employed at all. In Magliacane and Howard’s (2019)Magliacane, Annarita, and Martin Howard 2019 “The Role of Learner Status in the Acquisition of Pragmatic Markers during Study Abroad: The Use of ‘Like’ in L2 English.” Journal of Pragmatics 146 (6):72–86. study, significant differences in the PM use (i.e., like) are found not only between the university study group and the au-pairs group during a six-month sojourn in Ireland but also between the NS group and the two NNS groups. Additionally, both NNS groups have a significant gain over time, though. Nonetheless, the au-pairs group’s function typology is closer to that of the NS group. More recently, Magliacane (2020)Magliacane, Annarita 2020 “Erasmus Students in an Irish Study Abroad Context: A Longitudinal Analysis of the Use of ‘well’ and ‘like’” Study Abroad Research in Second Language Acquisition and International Education 5 (1):89–117. conducts a longitudinal and quantitative study that tracked the frequency and production of the two PMs, like and well, in study-abroad contexts. By comparison with English NSs, the results indicate an increase in frequency and constraints on the output of the PMs.

A critical point to infer from the above longitudinal studies is that even studying abroad in the target language country, different social contexts cause discrepancies in interlanguage PMs’ learning or development. The results may indicate more interactional alignments of the NNSs with NSs in the homestay settings than in the university settings. More situational variables such as social roles and activities are probably involved in the diversified homestay settings, even in contrast with Polat’s single home setting for data collection. What seems constructive herein needs more evidence for the effects of situational contexts on interlanguage PM development. More importantly, how PMs relate to specific situational contexts, especially EFL classroom contexts, in the learning process needs to be unraveled, and the process of interlanguage PMs’ dynamic development remains nebulous.

2.2Complexity theory and dynamic development in interlanguage PM studies

Complexity theory (CT), though hard to trace its source, is generally presumed to emanate from natural sciences (Larsen-Freeman 2017bLarsen-Freeman, Diane 2017b “Complexity Theory: The Lessons Continue.” In Complexity Theory and Language Development: In Celebration of Diane Larsen-Freeman, edited by Lourdes Ortega and Zhaohong Han, 11–50. Amsterdam: John Benjamins. ). While acknowledging the existing individual differences (Larsen-Freeman 1997Larsen-Freeman, Diane 1997 “Chaos/Complexity Science and Second Language Acquisition.” Applied Linguistics 18(2):141–165. ), Larsen-Freeman (2019)Larsen-Freeman, Diane 2019 “Second Language Development in Its Time: Expanding Our Scope of Inquiry.” Chinese Journal of Applied Linguistics 42(3):267–284. asserts that CT as an ecological theory is a metatheory or a transdisciplinary conceptual framework, viz. a holistic perspective on ways of thinking, context-dependent systems view, and dynamic systems view. CT views language as a dynamic complex system in which various system components are interconnected and changeable in interaction. The language system adapts itself to environmental stimuli through self-organizing behaviors, and then new language structures and functions or meanings emerge (Larsen-Freeman 1997Larsen-Freeman, Diane 1997 “Chaos/Complexity Science and Second Language Acquisition.” Applied Linguistics 18(2):141–165. , 2006Larsen-Freeman, Diane 2006 “The Emergence of Complexity, Fluency, and Accuracy in the Oral and Written Production of Five Chinese Learners of English.” Applied Linguistics 27(4):590–619. , 2017bLarsen-Freeman, Diane 2017b “Complexity Theory: The Lessons Continue.” In Complexity Theory and Language Development: In Celebration of Diane Larsen-Freeman, edited by Lourdes Ortega and Zhaohong Han, 11–50. Amsterdam: John Benjamins. ). Dynamic, emergentism, and adaptivity are three essential tenets of CT. Furthermore, other tenets in line with ecological theory include timescales as an exploration mechanism for cognitive purposes, fractals referring to the patterns of activities and events that have self-similarity at different levels of dimensionality, open-endedness in the open or inconclusive analysis of discrete events, and nonlinearity in language development (cf. Kramsch and Whiteside 2008Kramsch, Claire, and Anne Whiteside 2008 “Language Ecology in Multilingual Settings. Towards a Theory of Symbolic Competence.” Applied Linguistics 29(4):645–671. ; Larsen-Freeman 1997Larsen-Freeman, Diane 1997 “Chaos/Complexity Science and Second Language Acquisition.” Applied Linguistics 18(2):141–165. , 2006Larsen-Freeman, Diane 2006 “The Emergence of Complexity, Fluency, and Accuracy in the Oral and Written Production of Five Chinese Learners of English.” Applied Linguistics 27(4):590–619. , 2017bLarsen-Freeman, Diane 2017b “Complexity Theory: The Lessons Continue.” In Complexity Theory and Language Development: In Celebration of Diane Larsen-Freeman, edited by Lourdes Ortega and Zhaohong Han, 11–50. Amsterdam: John Benjamins. ).

In empirical studies, Larsen-Freeman (2006)Larsen-Freeman, Diane 2006 “The Emergence of Complexity, Fluency, and Accuracy in the Oral and Written Production of Five Chinese Learners of English.” Applied Linguistics 27(4):590–619. applies CT to L2 studies and explores the dynamic development process of speaking and writing abilities via accomplishing the same topic task in a time-series design experiment. Since then, researchers have been interested mostly in the dynamic development of grammatical or lexical competence or writing ability (Martín-Laguna and Alcón-Soler 2018Martín-Laguna, Sofía and Eva Alcón-Soler 2018 “Development of Discourse-pragmatic Markers in the Multilingual Classroom: A Mixed Method Research Approach.” System (75):68–80. ; Larsen-Freeman 2006Larsen-Freeman, Diane 2006 “The Emergence of Complexity, Fluency, and Accuracy in the Oral and Written Production of Five Chinese Learners of English.” Applied Linguistics 27(4):590–619. ; Verspoor et al. 2012Verspoor, Marjolijn, Monika S. Schmid, and Xiaoyan Xu 2012 “A Dynamic Usage-based Perspective on L2 Writing.” Journal of Second Language Writing 21(3):239–263. ). Other contextual factors to be measured in a time series design for spoken language learning (let alone for PM learning) remain understudied. Contextual factors may influence the dynamic development of learners’ pragmatic competence, which requires preferably tracking in a time series design experiment given the role of different contexts in shaping different PM employments. Even Larsen-Freeman (2017aLarsen-Freeman, Diane 2017a “Complexity and ELF” In The Routledge Handbook of English as a Lingua Franca, edited by Jennifer Jenkins, Will Baker, and Martin Dewey, Chapter 4 in Part 1. London: Routledge. , 2019Larsen-Freeman, Diane 2019 “Second Language Development in Its Time: Expanding Our Scope of Inquiry.” Chinese Journal of Applied Linguistics 42(3):267–284. ) herself states later (but not verifies) that dynamic development depends on the context.

However, regarding the studies of dynamic development of interlanguage PMs, to our best understanding, only Martín-Laguna and Alcón-Soler (2018)Martín-Laguna, Sofía and Eva Alcón-Soler 2018 “Development of Discourse-pragmatic Markers in the Multilingual Classroom: A Mixed Method Research Approach.” System (75):68–80. investigate the dynamic development of interpersonal and textual functions of ‘discourse-pragmatic markers’ (in their term) in three compositions written respectively in three languages (L1, L2, and L3) three times in a year. Little research has considered the effect of aligning with more specific contextual factors such as role types on the dynamic development of spoken PMs, i.e., the effect of “variable performance that could be due to differences in topic tasks or contexts” (Larsen-Freeman, 2006Larsen-Freeman, Diane 2006 “The Emergence of Complexity, Fluency, and Accuracy in the Oral and Written Production of Five Chinese Learners of English.” Applied Linguistics 27(4):590–619. , 595). Given the role of situational contexts in L2 development, what follows is to introduce the alignment theory, which further explains why the complex system of L2 development should be adaptive to the changes in the learning environment.

4.1Design

For a dynamic description, a time-series design is desirable (Larsen-Freeman 2006Larsen-Freeman, Diane 2006 “The Emergence of Complexity, Fluency, and Accuracy in the Oral and Written Production of Five Chinese Learners of English.” Applied Linguistics 27(4):590–619. ; van Geert and Steenbeek 2005van Geert, Paul, and Henderien Steenbeek 2005 “Explaining After by Before: Basic Aspects of a Dynamic Systems Approach to the Study of Development.” Developmental Review 25(3–4): 408–42. ), which inspired the present longitudinal and quantitative study. Given Larsen-Freeman’s (2006)Larsen-Freeman, Diane 2006 “The Emergence of Complexity, Fluency, and Accuracy in the Oral and Written Production of Five Chinese Learners of English.” Applied Linguistics 27(4):590–619. adoption of performance variability (four times’ oral and written production on the same topic tasks accomplished by five Chinese learners of English in a semester), our design focused on variable performance in a more extended time-series design (about 1.5 year). Given the lack of authenticity, e.g., the fragmentary aspect of spontaneous spoken language in the textbook (script)-bound EFL classrooms, we chose film-situated conversations as they contained a proper amount of PMs. The film-based conversations took place between social roles with un/equal role relationships in situational contexts. The two types of role relationships had four tasks, respectively (see Table 1). Learners were organized in class each time to see and listen to a short English film-situated conversation (2–3 minutes’ play) first, and then role-play with the assistance of follow-up prompt content tasks33.The sample task material may be available from the corresponding author if not attached behind. (turn-oriented reminders of fragmentary keywords or indirect speech descriptions) in a laboratory.

To address the afore-reviewed inconsistent findings, we hypothesized that to align with the two different types of role relationships in situations where other variables are the same would result in different effects on interlanguage PM use and development. For this, utilizing the situation types with un/equal role relationships as independent variables was to test their effects on the dependent variable, namely, the use/learning of the PM functions via Independent Sample T-Test. Furthermore, the learner corpus data needed comparing with the pedagogic sub-corpus data from CANCODE,44.CANCODE (Cambridge and Nottingham Corpus of Discourse in English) which forms part of the 1 billion word CIC (Cambridge International Corpus) is a five million-word spoken corpus developed at Nottingham University in the late 1990s, made up of recordings from a variety of settings in the countries of the United Kingdom and Ireland. Its pedagogic sub-corpus has 460,055 words in size (see Fung and Carter, 2007Fung, Loretta, and Ronald Carter 2007 “Discourse Markers and Spoken English: Native and Learner Use in Pedagogic Settings.” Applied Linguistics 28(3):410–439. ; McCarthy, 1998McCarthy, Michael 1998 Spoken Language and Applied Linguistics. Cambridge: Cambridge University Press.). a British spoken corpus built according to types of contexts (see Fung and Carter 2007Fung, Loretta, and Ronald Carter 2007 “Discourse Markers and Spoken English: Native and Learner Use in Pedagogic Settings.” Applied Linguistics 28(3):410–439. ; McCarthy 1998McCarthy, Michael 1998 Spoken Language and Applied Linguistics. Cambridge: Cambridge University Press.) in that both corpora were context-dependent. Such a measure would enhance the comparability of the two data as well as the reliability and validity of the study.

4.2Participants

In this longitudinal study, the participants (N = 28), 24 females and 4 males ranging in age from 19 to 23, were from a class of 32 Business English majors in a university in China. At the beginning of data collection, they were sophomores who had a two-period class once a week for the visual-audio-oral course. Upon completion of the data collection, they were in junior year. Thanks to some learners’ leave-takings and invalid recordings, the final number of participants was 28 who completed all the eight times’ tasks. They all had more than 10 years of learning experience in EFL classrooms. All of them spoke standard Chinese, EFL, and a local dialect. None of them reported having stayed in English-speaking countries for more than two weeks. In order to reveal more clearly the learning process and data analysis, the 28 students were categorized into two subgroups, i.e., the higher and the lower proficiency learners, according to the median score of an intermediate comprehensive English proficiency test55.This test consisting of grammar and vocabulary, listening, reading and writing was based on a standardized sample test for TEM 4, a national proficiency level test for English majors (for Year 2 college students) in China. they took shortly before the project.

4.3Materials and tools

The researchers chose the audiovisual materials (conversations) from films or TV series66.The film-situated conversation each lasted 3 minutes or so, chosen from films such as Glengarry Glen Ross, The Devil Wears Prada, and a TV series, i.e., House of Cards, available from the website: www​.youku​.com. for the input based on which they wrote the corresponding follow-up prompt content tasks for learners’ output. The NSs’ and NNSs’ conversations then share the same eight different business situations in films (including one TV series). Each follow-up task (handout) provides turn-taking, prompt content information, contextual (background) knowledge/situational variables such as topic, social roles and role relationships, location and time, purpose, and language style, nearly identical to the multidimensional representations of the situational models. The eight situational topics and role relationships are presented in Table 1, with four equal role relationships and four unequal ones alternatively assigned. These equal and unequal role relationships were treated as different situational variables to examine whether the learners’ alignment with the two types of role relationships would result in different productions of interlanguage PMs. The film-situated conversations and role-play tasks were utilized as tools for collecting data. They were carried out as usual practice, with both listening comprehension and speaking ability involved. Each of the eight tasks (including both film watching and follow-up role-playing or dyadic conversation) followed the same requirements or procedure as elaborated below.

4.4Procedure

In the longitudinal learning process, the students randomly formed dyads, and were required to conduct a follow-up prompt content task after exposure to a short clip of film in the visual-audio-oral class.

Each time, the teacher firstly allowed all the students to see a film clip twice in the laboratory and explained where students had difficulties understanding. The teacher then assigned the follow-up prompt content task in A/B role electronic version separately to the dyads’ mobile phones. After all the dyads’ preparation for the prompt content task, one or two dyads were selected randomly to perform it at the teacher’s desk. Finally, all the dyads’ conversations were recorded simultaneously in the laboratory, the whole process of which lasted about five minutes.

There were twelve tasks and recordings conducted in the project. Recordings of four data collection points were excluded because of some dyads’ invalid recordings or absences. Only eight data collection points’ recordings were valid. Few were even re-recorded due to one or two malfunctioning laboratory recorders on student desks; some were recorded after returning from short leaves of absence. For this, these individuals’ complementary recording was arranged one or two weeks later, which was supposed to decrease repetitive effect. The last two tasks were completed after the visual-audio-oral course ended.

4.5Data collection and analysis

Each dyad completed a follow-up conversation task after seeing a film clip each time. By the end of the project, 111 copies of intact recordings were collected. Three researchers were involved in transcription and checking, given the load of data of eight data collection points and pragmatic tagging. Functional annotations were exclusively confined to the three pragmatic functions (interpersonal, structural, and cognitive) and the three selected PMs (yeah, well, you know) that incorporate the functions based on Fung and Carter’s (2007)Fung, Loretta, and Ronald Carter 2007 “Discourse Markers and Spoken English: Native and Learner Use in Pedagogic Settings.” Applied Linguistics 28(3):410–439. classification. After printing the transcription, two researchers separately annotated the functions on printed copies first according to the categories in Table 2. They marked the PM functions with abbreviations in the square bracket, e.g., [WIF] stands for ‘well, interpersonal function’ (see Table 2). To make the annotations more consistent and objective, they calculated the inter-rating score, for which they got 97.9%. The discrepancies were discussed and checked by a professor until a final agreement was reached. The annotated PM functions in the printed copies were then tagged in the corpus’ electronic documents for the convenience of retrieving these annotated PM functions via a concordance. The tools used included Word for tagging and editing, Excel for calculation and figures, AntConc for corpus analysis, and SPSS for statistical analysis.

5.1The overall picture of interlanguage PM development

The total words of the learner corpus are 39,409. As for the PM production, on the one hand, the results cumulatively from all the diachronic data in Table 3 show that the most frequently used marker of the students is yeah (0.87%), followed by well (0.47%), while the frequency of you know (0.19%) is the least among the three. Then, are the learners’ PM productions overused or underused? This question necessitates comparing the NNSs’ data with NSs’ data as NSs’ normalized PM frequencies function as a baseline, conducive to gauging the extent to which the learners’ PMs were developing. Following Fung and Carter (2007)Fung, Loretta, and Ronald Carter 2007 “Discourse Markers and Spoken English: Native and Learner Use in Pedagogic Settings.” Applied Linguistics 28(3):410–439. , the word frequency ratio is comparable between +0.14% and −0.14%. Therefore, as indicated in Table 3, the frequencies of the first two markers are comparable to those of NSs (except you know).

On the other hand, as far as a single task is concerned, a calculation of the scarce PM utilizations in Task 2 (yeah = 0, well = 0.22, you know = 0.26) or Task 3 (yeah = 0.71, well = 0.30, you know = 0.02) invites a fundamental question. Namely, why can frequency percentages from a single task be so low?77.Cf. Fung and Carter’s (2007)Fung, Loretta, and Ronald Carter 2007 “Discourse Markers and Spoken English: Native and Learner Use in Pedagogic Settings.” Applied Linguistics 28(3):410–439. low results: yeah = 0.47, well= 0.01, you know = 0.16 (Cf. Section 6.1)

Concerning the dynamic development, Figure 1 shows that the three markers’ respective frequencies manifest themselves in fluctuations despite their mutual differences, indicating that the development trajectories of the three PMs in the learning process are nonlinear.

Figure 2, then, indicates that all the PMs’ functional frequency tends to demonstrate a fluctuating but rising trend in the longitudinal learning process. Given the fluctuation in this three-in-one trajectory, the learners’ development and utilization of the PM functions may have been reshaped by or readapted to the different contexts distributed at the different data collection points in the tasks. Overall, the learners’ ability to use the PM functions may have developed despite its regression and progression in the longitudinal process.

In order to confirm the learners’ PM progress in the learning process, we deployed the Paired Samples Test to verify whether there is any significant difference between the first time (beginning) task and the last task. Table 4 shows such a significant difference (t = −4.032, p = .000), indicating that the learners have made progress or developed their PM competence. To date, however, we still wonder whether there is any nonlinearity of the change trajectory in terms of the PMs’ specific functions. Thus, it is necessary to examine such dynamic nonlinear development of specific PM functions that manifest themselves in each PM’s function types.

5.2Dynamic changes and development of interlanguage PM functions

Regarding RQ 1, our results indicate the dynamic changes and development not only in the above overall data but also in the ones of the individual PMs below. We deployed figures to uncover such a conundrum relating to each PM’s multi-functions in that figures are visually apparent and highlight each PM’s trajectories of functions.

5.2.1The frequencies and functions of ‘yeah’

Figure 3 shows that the interpersonal and the structural functions of yeah have the highest frequencies in the fourth and sixth tasks, respectively, with little fluctuation in the other tasks. Overall, the interpersonal function frequency of yeah is characterized by a high middle and low sides, while the structural function is less frequently deployed and has little fluctuation. It is worth noting that the students did not use any yeah in the second task.

Figure 3.Dynamic development of yeah

5.2.2The frequencies and functions of ‘well’

Figure 4 reveals that the frequencies and functions of well changed dramatically with a relatively broad fluctuation range in the learning aligned with different movie situations. Understandably, the lowest frequencies of both interpersonal and cognitive functions of well emerged in Task 1 at the beginning. However, the frequency of the cognitive function of well peaked in Task 5, and so did the other two, indicating the relation of a particular function of the marker to a specific context. The frequencies of all the three function types remained the second in the last task. The fluctuation of the interpersonal function of well was not enormous but relatively stable, indicating that the learners deployed this function more frequently than the other two. Conversely, the cognitive and structural functions had larger fluctuations throughout the whole learning process. In sum, the frequency of each function of well has fluctuation, but the trend of the learning trajectory ascends.

Figure 4.Dynamic development of well

5.2.3The frequencies and functions of ‘you know’

Figure 5 shows that the use frequencies of the different functions of you know were continually fluctuating with the movie situation changed in each task. In a situational alignment, generally, the students only utilized a single function rather than two functions of the pragmatic marker in the same situation. As a result, the two functions manifested themselves in considerable fluctuations in the learning process. Overall, the frequency and functions of you know rise alternatively and are inferior to those of the other two PMs.

Figure 5.Dynamic development of you know

In sum, the PMs manifest themselves in more interpersonal functions (IF) than other functions (especially in the first four times). Longitudinally, one point to note is that the first half-learning process’ interpersonal function frequency gradually decreases in the second half, whereas the structural and cognitive functions gradually increase. This, however, differs from the case of you know.

5.3The influence of types of role relationships on the learning of interlanguage PMs

Regarding RQ 2, our results indicate that the mean of the unequal role relationship is 8.321 (SD = 2.789) and the mean of the equal role relationship is 13.464 (SD = 3.553), which means that the latter had more effect on the learners’ production of PMs than the former. Table 5 then shows that the PM learning in alignment with the two role relationships demonstrates a significant difference (t = −6.024, p = .000), indicating that the situations with equal role relationships are more facilitative for the production and alignment of the PMs. Thus, equal role relationships promote more effectively the learning of PMs in EFL learning contexts.

5.4Differences between the subgroups in the use of interlanguage PMs

Regarding RQ 3, our results show that there are significant differences between the two subgroups in the functional frequencies of the PMs in the last and the sixth tasks (F = 29.885, p = .000; F = 5.200, p = .031), whereas there are no significant differences in the rest of the tasks (Table 6). A tendency of increasing differences can be observed in the longitudinal development despite some fluctuations in Tasks 2 and 7. This finding indicates that the higher proficiency learners performed and developed better than the lower proficiency learners did in the PM functions’ use and learning. Hence, this also demonstrates the existence of dynamic development in the longitudinal process as there reappears even no significant difference in the seventh task data between Tasks 6 and 8.

As far as the one-way ANOVA statistical results are concerned, we have not discerned how the statistical differences manifested themselves in the development of PM function types between the subgroups. Therefore, to uncover the differences between the subgroups in the development of PM function types, we then turn to deploy a figure that can visually present a diachronic development.

Figure 6 indicates the dynamic development of PM function types and the effect of aligning with contexts. The higher proficiency learners outperformed the lower proficiency learners in the PM function types, especially in the later periods, indicating that the former acquired or expanded PM function types earlier than the latter in the development process of interlanguage PMs.

6.1The dynamic development in the frequencies and functions of interlanguage PMs

Regarding RQ 1, our results indicate that the learners’ frequencies and functions of the PMs developed dynamically in contextualized learning. Such dynamic development manifests itself in what follows. Respecting the single tasks, relatively low are the frequencies of the three PMs in Tasks 1 and 2, which are even lower (while the ones in Tasks 4, 5, and 8 can be higher) than those in the cross-sectional or immediate effect studies, as indicated and reviewed in the previous sections (cf. Fung and Carter 2007Fung, Loretta, and Ronald Carter 2007 “Discourse Markers and Spoken English: Native and Learner Use in Pedagogic Settings.” Applied Linguistics 28(3):410–439. ; Ament et al. 2018Ament, Jennifer, Carmen Pérez-Vidal, and Júlia Barón Parés 2018 “The Effects of English-medium Instruction on the Use of Textual and Interpersonal Pragmatic Markers.” Pragmatics 28(4):517–545. ). Indeed, such results collected at a single time point from learners’ interlanguage can be far lower or higher than the normalized ones of the adult NSs in CANCODE educational sub-corpus. The point is that there emerges a problem of unreliability in comparing a-single-time-point-collected NNSs’ data with NSs’ data. Although the latter were also collected at a single time point (not longitudinal data), the NSs’ language is a fully-fledged mother language of adults, relatively stable and grounded upon large collections of texts and contexts. Conversely, NNSs’ interlanguage is a developing language that is dynamic and context-dependent, as Larsen-Freeman (2017a)Larsen-Freeman, Diane 2017a “Complexity and ELF” In The Routledge Handbook of English as a Lingua Franca, edited by Jennifer Jenkins, Will Baker, and Martin Dewey, Chapter 4 in Part 1. London: Routledge. states. Thus, cross-sectional or immediate effect analysis cannot reveal L2 dynamic development, and the data of this type can be contingent and unreliable due to PMs’ task/text-specific nature.

Moreover, the unreliability also lies in the fact that different contexts shape different PMs’ utilizations. Given this factor, had the results or frequencies been from a single conversation in a single context, the findings of cross-sectional or immediate effect analyses in the literature would have been contingent and unreliable. Therefore, from a time series perspective, the total frequency percentages of the PMs (yeah, well) in our study that are closer and comparable to the use frequencies of NSs (except you know) are comparatively more convincing based on data collected from many texts and contexts. This finding also reflects the dynamic development of the EFL learners’ PMs demonstrated through rise and fall or regression and progression.

Longitudinally, reasons for the dynamic changes and fluctuation may be variegated but are strongly related to two aspects. On the one hand, it is evident that the peak value in Figure 2 does not appear at the end of the trajectory but in the middle (i.e., the fourth time), and the valley value is not in the first but in the second time. Thus, a rational explanation for this is that the co-occurrence of both cotextually and contextually relevant variables causes the disparities in PM meaning production. Task 2 illustrates that, apart from the severe topic, i.e., complaining about work (cf. Task 4 has a light topic and a close/equal role relationship suitable for the use of the interpersonal function of yeah), the other factor that constrained the PM use is the setting. The driver was driving while complaining about his work, whereas his colleague, realizing the driving and raining situation, listened most of the time with little eye contact and occasionally gave rapid responses to achieve conversation continuity. Such situational context may reduce the number of turn-takings and the turn-initial markers or require corresponding conversation exchanges, which supports Gruber and Muntigl’s (2005)Gruber, Helmut, and Peter Muntigl 2005 “Generic and Rhetorical Structures of Texts: Two Sides of the Same Coin?” Folia Linguistica 39 (1/2): 75–113. view on the relation of situations to generic structures.

Likewise, the low frequencies of the PMs in the third and seventh times can result from the formal situations that correlate unequal power, low affective involvement, and infrequent contact (Eggins 2004Eggins, Suzanne 2004 An Introduction to Systemic Functional Linguistics. 2nd Edition. London: Continuum.). In other words, such situational variables may confine the utilization of certain PM functions, e.g., interpersonal functions. A glance back at the original film clips verified the finding in that NSs in these film conversations employed fewer PMs, either. In contrast with the PM frequencies in the fourth time and others, our findings have corroborated the afore-reviewed important argument (see also Taguchi 2011Taguchi, Naoko 2011 “Pragmatic Development as a Dynamic, Complex Process: General Patterns and Case Histories.” The Modern Language Journal 95(4): 605–627. ) that situational variables influence pragmatic development. Probably, the learners with their shared attention to the multidimensional representations, including the non/linguistic representations in the multimodal settings, are prone to approaching or aligning with NSs’ shared cognition, the norm of PM functions, in their shared contextualized activities. In line with Bazzanella (2006)Bazzanella, Carla 2006 “Discourse Markers in Italian: Towards a ‘Compositional’ Meaning.” In Approaches to Discourse Particles, edited by Kerstin Fisher, 449–464. Amsterdam: Elsevier., such contextualized activities, involving the co-occurrence of cotextual and contextual parameters, can activate the PMs’ meanings and promote their alignment with text and contexts. Thus, our time series analysis provides evidence for the general view (Atkinson 2010Atkinson, Dwight 2010 “Extended, Embodied Cognition and Second Language Acquisition.” Applied Linguistics 31(5): 599–622. ; Atkinson et al. 2007Atkinson, Dwight, Eton Churchill, Takako Nishino, and Hanako Okada 2007 “Alignment and Interaction in a Sociocognitive Approach to Second Language Acquisition.” The Modern Language Journal (91):169–188. ; Pickering and Garrod 2004Pickering, Martin J., and Simon Garrod 2004 “Toward a Mechanistic Psychology of Dialogue.” Behavioral and Brain Sciences 27(2): 169–26. ) on the incidence of alignment with situational context on learning.

On the other hand, the learners’ PM competence in the trajectory is on the rise, especially in the development of PM multifunctionality. In other words, in the first four tasks, the students mostly deployed the interpersonal function of the PM, yeah. After practices aligned with different situational contexts, the structural function type of yeah emerged out of the contextual affordances. As for the multi-functions of well, its structural and interpersonal functions developed better than the cognitive function. In the case of you know, its functions’ frequencies are different in that its cognitive function developed earlier than (and outperformed) its interpersonal function. The PMs differ from one another as they have different frequencies in use and different relations to utterances or context. Specifically, in the dynamic developmental process of Interlanguage PMs, the dominating frequency of the interpersonal function of yeah may be gradually leveled off by the use frequencies of the other function and functions of other PMs in the subsequent tasks. Despite some inconsistent findings by Ament et al. (2018)Ament, Jennifer, Carmen Pérez-Vidal, and Júlia Barón Parés 2018 “The Effects of English-medium Instruction on the Use of Textual and Interpersonal Pragmatic Markers.” Pragmatics 28(4):517–545. that NNS learners tend to use more structural or textual PMs and fewer interpersonal PMs than NSs, our findings show that the interpersonal functions of yeah and well except you know have more frequencies than other functions. However, what counts is that there is no comparability between as their results are subject to written L2 data with different PMs and contexts, especially their cross-sectional and immediate effect study method. Data collected at a time point within a specific context can be static and contingent, as mentioned above, and cannot reflect the dynamic development of interlanguage PM functions.

In sum, the interlanguage PM functions’ adjustments or changes are also part of dynamic development, as can be seen obviously in Figure 6. Such dynamic changes and development manifest themselves not only in the mapping of situations with conversational structures via different PM functions but also in the meaning-making resulting from the multidimensional simultaneity of language creation in context. Namely, PMs correlate larger situational and conversational patterns or structures; on the other hand, PMs per se are relatively dynamic in terms of their incompatible multi-functions or diachronic and synchronic meanings, emerging from the dynamic of timescales in context (Kramsch and Whiteside 2008Kramsch, Claire, and Anne Whiteside 2008 “Language Ecology in Multilingual Settings. Towards a Theory of Symbolic Competence.” Applied Linguistics 29(4):645–671. ), which may better explain the conflicting complex classifications of PM functions in the literature. Thus, our findings support Larsen-freeman’s (1997Larsen-Freeman, Diane 1997 “Chaos/Complexity Science and Second Language Acquisition.” Applied Linguistics 18(2):141–165. , 2006Larsen-Freeman, Diane 2006 “The Emergence of Complexity, Fluency, and Accuracy in the Oral and Written Production of Five Chinese Learners of English.” Applied Linguistics 27(4):590–619. ) view on CT or dynamic development in EFL learning.

6.2Influence of types of role relationships on the use of interlanguage PMs

Regarding RQ 2, our results indicate that the types of role relationships had effects on the employment of the PMs. Based on this study’s design, the interlocutors of the equal role relationships in Tasks 2, 4, 5, and 8 used the PMs more frequently than those of unequal role relationships in Tasks 1, 3, 6, and 7. The reason for this may be that the language used in the equal relationship is more casual and intimate, and opinions are expressed more directly, which invokes more choices of the functions of the PMs, in particular, interpersonal functions. This finding is consistent with that of Eggins (2004)Eggins, Suzanne 2004 An Introduction to Systemic Functional Linguistics. 2nd Edition. London: Continuum. that informal situations correlate equal power, high emotional involvement, and frequent contact in natural conversations, or vice versa. Such interpersonal interactions require learners’ collaborative or synergetic learning in alignment with the film-situated role-plays. The alignment theory holds that there is a synergetic influence on learning among the mind, body, and the world (Atkinson 2010Atkinson, Dwight 2010 “Extended, Embodied Cognition and Second Language Acquisition.” Applied Linguistics 31(5): 599–622. ). In enacting the equal role relationships, the dyadic role-playing depends upon reciprocal conversations realized by question and answer and high affective involvement manifested chiefly by PMs’ interpersonal functions and/or facial expressions. One may obtain assistance from the other when confronted with difficulties in sustaining their ongoing conversation. Moreover, the dyads should be willing to cooperate in playing the roles. Should one be unwilling to be friendly to ‘a good friend’ in a role-play situation, the other will find it hard to act or speak intimately via the PMs’ interpersonal functions.

Given the different effects of different types of role relationships on the students’ language expression, the diversity of situation, particularly affective involvement, is ecologically crucial to the PMs’ learning and development in EFL classroom environments. Nevertheless, this is something neglected in the literature. Owing to the different results caused by the differences in role relationships and sampling, cross-sectional design, case tracking research and even pre-and-post-test studies without a control group or delayed post-test may have their respective limitations betraying the role of contexts and the dynamic development. The reason is that a single test based on a single text/context at a specific point of time generates utilization of text /context-specific PMs and functions. Such utilizations or frequencies of PMs and functions are locally situated within a single text, which cannot reflect PMs’ complex system (norm of PMs in use) or the dynamic changes and development of Interlanguage PMs. Our time series study reveals the interlanguage PMs’ dynamic development and dependence upon or relevance to the context through alignments with the two types of role relationships.

Despite EFL learners’ difficulty reaching NSs’ breadth and depth of markers (Fung and Carter 2007Fung, Loretta, and Ronald Carter 2007 “Discourse Markers and Spoken English: Native and Learner Use in Pedagogic Settings.” Applied Linguistics 28(3):410–439. ), our learners’ PMs, i.e., yeah and well, except you know, are comparable to NSs’. Among them, the learners used well slightly more frequently than NSs (cf. Table 3). Such results undoubtedly owe to the two types of role relationships as equal role relationships in the four situational contexts have been proven to facilitate PM production. In this sense, a task design for data collection is supposed to have balanced roles of un/equal relationships. To sum up, the alignment with the movie business situations with different role relationships and turn-taking scaffolding in the follow-up prompt content tasks is facilitative for matching the PMs’ different functions with their corresponding contexts in conversations. This finding supports the general view on PMs’ central function as expressing the relation of an utterance to the preceding utterance or to the context (Brinton 1996Brinton, Laurel J. 1996 Pragmatic Markers in English: Grammaticalization and Discourse Functions. Berlin: Mouton de Gruyter. ).

6.3Subgroups’ differences in the frequencies and functions of interlanguage PMs

Regarding RQ 3, our results show that the higher proficiency learners are superior to the lower proficiency learners concerning the frequency and function types of the PMs. In the film-situated alignment, the learners showed their ability to interact interpersonally and align with the film-based texts and contexts. The dyadic alignment in PMs emerged between the higher and the lower proficiency learners. It is evident that during dyadic interactions, the learners may repeat each other’s PMs as delineated in the alignment theory; especially, the lower proficiency learners may imitate the higher proficiency learners’ PMs used in their dyadic role-plays as the dyads were not fixed but randomly paired. Longitudinally, however, both subgroups have improved their PM competence. Such improvements occurred in accomplishing the turn-oriented tasks aligned with the NSs’ film-situated language, especially PMs. Hence, our findings support Pickering and Garrod’s (2004)Pickering, Martin J., and Simon Garrod 2004 “Toward a Mechanistic Psychology of Dialogue.” Behavioral and Brain Sciences 27(2): 169–26. view that the situation models correlate with linguistic representations.

The main explanation for the difference between the subgroups is that the higher proficiency learners manifested their more vital learning ability in cognitive processing in alignment with the situations, especially the relationship between PM functions and coherence or organization in discourse. Therefore, they utilized more PMs and function types by the end of data collection. This finding is consistent with that of Ament et al. (2018)Ament, Jennifer, Carmen Pérez-Vidal, and Júlia Barón Parés 2018 “The Effects of English-medium Instruction on the Use of Textual and Interpersonal Pragmatic Markers.” Pragmatics 28(4):517–545. . Their cross-sectional study results show a significant increase in the overall frequency and variety of types of PMs used from Year 2 to Year 3. Despite the discrepancies in the population and learning contexts, their cross-sectional design-based finding and our longitudinal one corroborate each other in the sense that there exist different learning or developmental stages in the process of learning interlanguage PMs. It seems to suggest that some interlanguage PM functions may be acquired late as the learners were already in Year 3 upon the emergence of the significant differences in the later stage of the project, reaching an upper-intermediate proficiency level. The higher proficiency learners deployed more effectively the PM multi-functions in relating to context and structural priming, and their developmental process of the PMs took precedence over that of the lower proficiency learners. The latter may have difficulty organizing syntax and expressing the content to such an extent that they are less likely to master the PMs’ various functions.

Furthermore, owing to the increasing differences in the PM functions’ use frequencies between the two subgroups, a change emerged from quantity to quality, i.e., the significant differences in the last tasks. This finding also verifies that the learners’ pragmatic competence of PMs had a dynamic development in the use frequencies of PMs or the categories of PM function types. Such dynamic development mainly resulted from the learners’ learning abilities, especially listening and speaking competencies. Indeed, the lower proficiency learners’ listening difficulty in understanding the film conversations may lead to their difficulties in speaking, including the use of PMs and recognition of PM functions, in that listening and speaking naturally relate to and align with each other. Holistically, the significant difference reflects the process of the dynamic development of pragmatic competence and the dynamic changes of different proficiency learners’ PMs, though the subgroups came from the same population or the same class. This result then verifies the view (see Larsen-Freeman 1997Larsen-Freeman, Diane 1997 “Chaos/Complexity Science and Second Language Acquisition.” Applied Linguistics 18(2):141–165. ) that individual differences exist in the dynamic development of L2 learning competence.