Читать книгу Complexity Perspectives on Researching Language Learner and Teacher Psychology - Группа авторов - Страница 9

2

Reflections on Researching Dynamics in Language Learning Psychology

Peter D. MacIntyre, Sarah Mercer and Tammy Gregersen

Introduction

Humans are highly complex beings. Language and communication are complex systems. Combining psychology and language learning with the multifaceted communication process increases the complexity substantially. Add then to the mix the idea that all of these things are constantly changing over time and with even small differences in context, the challenge facing researchers in this field rounds into form. How does a research project capture such complexity and dynamism in a meaningful way without oversimplifying the lived reality? All three authors of this chapter have found taking a complexity-informed approach to research to be a most satisfying way of tackling this problem, but one that creates its share of challenges. In this chapter, we will reflect on our own experiences of grappling with the issues created by adopting a complexity perspective, focusing on timescales and dynamism. We will conclude this chapter with a series of recommendations, emerging from our experience, that we hope will assist other scholars wanting to research the dynamics of learner and/or teacher psychologies.

A Gap between Theorizing and Empirical Research

Within applied linguistics, theorizing complexity is well ahead of its empirical investigations (MacIntyre et al., 2017). Perhaps the reason for the imbalance between describing complex dynamic systems theory (CDST) and using it for empirical research is that CDST is itself meta-theoretical in nature, meaning that it is not a theory of language learning per se but an approach to creating theory (Larsen-Freeman, 2017); research does not test CDST directly. The ideas underlying CDST have been developed in several of the natural sciences, and now the social sciences and humanities including developmental psychology (Thelen & Smith, 1994) and communication (Fogel, 2006) have followed suit. In second language acquisition (SLA), CDST has attracted a collection of researchers interested in testing its applicability to various processes, many of whom are writing chapters in this volume and who have used it both as a theoretical lens as well as a basis for empirical design and analysis.

SLA provides a rich context for understanding dynamics with its focus on language development, change and stability. However, it is not just language competences that are in a state of flux, but a range of related factors during these processes that are constantly interacting with learner and teacher psychologies. This makes the psychology of language learning an especially fertile ground for dynamic studies to take root. To date, research has tested the question of the applicability of CDST to familiar topics in this field such as willingness to communicate, motivation, the self, agency, anxiety, enjoyment, teacher efficacy and strategies; we discuss specific research examples. The studies we describe below unanimously conclude that the approach is well suited and the phenomena under study meet the key conditions for CDST. The work so far has indicated how a CDST perspective can lead to novel developments in conceptualization, methodology and data analysis/presentation; the possibilities for future research excite the imagination. The question now is how to proceed with research that will further our appreciation of the scope and potential of this theoretical framework to extend our understandings of the field of language learning and teaching.

Benefits of a CDST Perspective on the Psychology of Language Learning

Our interest in CDST is based in large part on what it offers as a way of thinking about the psychology of language learning (PLL) and communication and how CDST can contribute to new ways of studying the processes involved. The continuous interactions among the myriad of inter-personal and intra-personal processes are intricate, nuanced, contextualized and ever-changing. The focus on complex interactions requires a new set of tools for building theory and research.

Before we consider the conceptual devices in the CDST toolbox that create the context for the approach, we must first establish what is being specified as a system. Defining a system is a matter of perspective; a lens that the researcher puts on a complex reality so that a system comes into focus as a functioning whole. The system remains open but its boundaries can be described. Discussions of CDST sometimes can become difficult to interpret if the system is not specified and this is an essential first step. Defining a system for study is a matter of emphasis, focussing on the most relevant processes. Not everything functions as a complex dynamic system and certain criteria must be fulfilled before a system can be defined (Mercer, 2016). To set the stage for the rest of the chapter and illustrate some of those characteristics, we highlight attributes of CDST that differ from other research approaches and make it especially relevant to PLL.

Timescales

Perhaps the most fundamental contribution of CDST is thinking explicitly about time and how it affects the processes under study. CDST forces a researcher to consider how a process unfolds over a chosen period of time. For example, in prior research, anxiety has typically been conceptualized as a long-term, trait-like quality of a learner, but in recent research it has been studied as rapidly changing from moment-to-moment, interacting with many of the features of communication events (MacIntyre, 2017). Both shorter and longer timescales have been used in prior research but they require very different approaches and answer different types of research questions. The novel contribution of CDST is to explicitly consider multiple, interacting processes occurring within the selected timescale. For example, the trend toward increasingly communicative forms of language teaching has been taking place over many years (Canale & Swain, 1980; Savignon, 2000); nested within such a broad trend will be patterns of talking within specific classrooms (King, 2013), and nested within a specific day in the classroom will be fluctuations in communication for teachers and learners as various activities take place (Mystkowska-Wiertelak & Pawlak, 2017). To probe further the conceptual context, de Bot (2014) recommends thinking about processes one timescale level up and one timescale level down in addition to the timescale of the process under study. For example, in one of the studies described below, we studied anxiety changes over the course of a 20-minute presentation. We found it helpful to think about the sub-processes (cognitive, emotional, physiological) that affect anxiety on a per-second timescale as well as how anxiety arousal during a classroom presentation is itself part of longer-term processes, such as learning during a semester-long course and how anxiety changes in the process of becoming a language teacher. Timescales are nested within each other, as seconds are nested in minutes, minutes within hours, hours within days, and so on. It may be that different dynamics are visible on different timescales, or possibly that patterns repeat across different timescales (see the discussion of fractals below).

Openness

The openness of a system refers to the notion that the system is subject to sources of influence that perhaps were not contemplated when planning a research design. Essentially, openness to unexpected influences increases substantially the difficulty of predicting the most relevant processes or systems to study, especially if considered at the individual level where personal psychological idiosyncrasies abound. In the research examples below, we note that allowing for unexpected influences on a system can help to bring clarity to the dynamics involved, highlighting what is typical and what might be unusual. To provide specific examples, we will highlight how density of measurement and including a qualitative component to research designs allows for the description of unexpected factors.

Predictability, stability and variability

A particular strength of CDST is the focus on variability and change over time. Each state of the system under study is taken to be a modification of the system’s previous state, and the trajectory of change in a system can be highly sensitive to its initial conditions. Given that we cannot know everything about a system and the interactions within it, there will always be some level of unpredictability about the kinds of potential changes that may occur. However, unpredictable does not mean ‘anything goes’ because there is not an infinite number of possible outcomes or states of the system. Some states are more likely to arise than others (see attractor states below). It is important to note the range of outcomes is influenced by initial conditions and the system dynamics meaning that the potential states of a system cannot be absolutely anything at all, but also cannot be precisely defined or predicted exactly in advance. The concept of dynamic stability reflects the potential for a system to remain in a relatively steady state for some period of time. It means there can be minor changes, fluctuations and variations but the overall state of the system remains relatively stable, but not fixed or static. In the research examples below, we show the importance of defining the timescale to focus on variability or stability, as processes may be volatile over short periods of time but relatively stable over longer ones.

Attractor and repeller states

The terms attractor and repeller states come from chemistry where they are not loaded with the connotations that occur when they are applied to human beings. In the vernacular, the term attractor connotes valuable and desirable; the term repeller connotes pushing something away. However, this is not the definition of those terms in CDST and their misuse will be problematic unless researchers are clear that attractor states are those to which a system tends to be drawn whether or not the state is thought to be pleasant and welcome – a long-running feud between rival teachers is an attractor state because it is relatively stable, even if it is also a nasty experience. Similarly, a repeller state may be pleasant or unpleasant, but by definition a system will not remain in that state for very long. The terms merely refer to the preferred state of the system, not whether that preferred state is positive or negative in valence. In the research examples below we take individual difference factors that have been studied using other methods, such as willingness to communicate and language anxiety, to be instances of attractor or repeller states of a system.

Emergence, self-organization and soft assembly

The idea of emergence suggests the state of the system can be considered more than the sum of its multiple, continuously interacting parts. Self-organization suggests that systems are not operating according to a predefined blueprint and do not reflect the inevitable unfolding of a plan. Rather, systems organize themselves and have an intrinsic tendency to display coherent patterns. Soft assembly refers to the idea that the interacting parts of a system can be shared and re-configured into coherent patterns, as systems organize themselves. A learner might describe ‘feeling motivated’ or ‘feeling anxious,’ two emergent states that often share features such as engagement with the learning process, involvement of the self, relationships with other people, a role for the teacher, emotional arousal, salience of learning goals and so on. Yet motivation and anxiety are experienced by individuals as qualitatively different states; generally speaking, motivation is pleasant and anxiety is not, motivation favours approach but anxiety suggests avoidance. The concept of emergence has been influential in connecting the dynamics of process to established research on familiar topics previously studied from other perspectives. In the examples below, we offer suggestions about how emergent states (being willing to communicate, experiencing anxiety, feeling motivation) make sense when considered as coherent, organized states dynamically assembled through the interactions of associated systems.

Fractals

Fractalization refers to the characteristic of systems to display self-similar patterns across levels. This means that the behaviour of a system on one timescale or level can potentially predict similar behaviour on different levels. Similarly, if patterns are found across timescales or levels, this can serve as evidence for a complex dynamic system. This means similar attractor/repeller states and similar types of dynamics can be manifested across system levels. For example, Mercer (2015) found that the self system of her learners exhibited comparable kinds of dynamics across different timescales of minutes, hours, weeks and months. The findings implied patterning in the types of system dynamics and system states when examined on different timescales. Fractals reveal that there can be surprising patterns and regularities in what may at first appear as random, chaotic systems, which, of course, can have important implications for research. In the research examples below, we are starting to see that some of the processes that occur on longer timescales also appear on shorter ones (see in particular descriptions of research into willingness to communicate and the self).

Research Questions from Standard and Complexity Approaches

Understanding these characteristics of a complex dynamic system has implications for how we formulate research questions, and even what is the focus of research. To illustrate the nature of research from a CDST perspective, we will outline typical characteristics and forms of quantitative and qualitative studies and show how they can be reconfigured and reconceptualized to better reflect CDST characteristics.

Quantitative compared with CDST approaches

The difference between a CDST approach to research and a traditional quantitative approach is substantial. CDST approaches can use quantitative data, and a number of data analytic techniques have been initiated (see Hiver & Al-Hoorie, 2020). But quantitative data analysis must be guided by research questions. Not only do the research questions themselves change, but the nature of the answers change as well. To clarify, Table 2.1 offers three examples of research questions that have been studied from a quantitative perspective, and a dynamic re-phrasing of questions in the same domain.

Table 2.1 Quantitatively-oriented research questions compared with CDST variations
Standard questions	CDST re-phrasing
Does language anxiety correlate with course grades?	What happens as anxiety rises and falls during a test?
On average, are extraverts more willing to communicate than introverts?	How does introversion/extraversion combine with other factors to contribute to creating a learner’s willingness to communicate?
Does motivation predict effort invested in learning?	What happens to effort as avoidance motivation rises?

Within a particular study, questions such as those in the left-hand column of Table 2.1 will have a definite answer. Using statistical tests, we can say that the correlation between anxiety and course grades is −0.36 (Teimouri et al., 2019), extraversion correlates with willingness to communicate (WTC) at r = 0.39 (MacIntyre & Charos, 1996) and the correlation between the Ideal L2 Self (motivation) and intended effort is 0.61 (Al-Hoorie, 2018). Across quantitative studies, different values will emerge. It is understood that correlations will vary from one study to another, as will group means and so on, but often the statistical results are similar and discrepancies accounted for by differences in sampling or methods used from one study to the next. The answers produced most often are phrased in concrete terms, using statistical values to create a sense of confidence, and when theory is added to the mix, a sense of genuine research progress can arise. Many researchers and research consumers find this approach satisfying.

The CDST toolbox produces a different approach by starting with a process-oriented account of the phenomena. Instead of finding the value of a correlation between motivation and intended effort or language course grades for example, the CDST account might describe what learners who feel motivated are thinking and feeling, what is changing in their cognitive and emotion systems, what they are doing behaviourally, what is happening with their language production and nonverbal communication, and so on. In this case, the motivation system is under scrutiny as an emergent, self-organizing state that is soft assembled from a learner’s cognition, emotions, behaviours, social context, interactions with peers, teachers, culture, other interlocutors and more. The specific details of the motivation process will differ from one person to the next – perhaps one person clearly imagines a desired future of smooth L2 communication getting closer and closer, but another ties their motivation directly to enjoying the present classroom context and the relationships therein. The research focus shifts from discovering generalizable patterns and estimating group averages to a focus on the specific perspectives of individual experiences and processes.

An additional implication of changing the phrasing of research questions is that it alters the types of analytic techniques that can be applied. Questions such as those in Table 2.1 often are answered with statistical analysis such as correlation, regression, structural equation modelling and analysis of variance, among others. Broadly speaking, there are two ways of constructing research questions - group comparisons and correlation. In group comparisons, we might ask whether differences between the means (averages) for two or more groups have arisen by chance or whether a difference between group means is statistically significant and meaningful. In studies based on correlation the objective is to describe the strength of a tendency for scores to rise and fall together; stronger correlation leads to better prediction. In both cases, variability or deviations (from the mean or from the predicted values in the case of correlations) usually cannot be explained and are used for statistical purposes as estimates of random error. Large sample sizes are valued because they provide more stable estimates of means, correlations and random error, allowing for more confidence in the results.

A dynamic focus follows different rules. Methods to answer CDST questions are not yet commonplace and there is need to further develop and disseminate methodology. In writing CDST projects for publication, we are learning that a focus on dynamics upends many of the accepted criteria for the evaluation of research. Quantitative research projects in the psychology of language learning are most often done with a cross sectional design where a researcher uses a pre-determined set of instruments to measure the concepts being studied while collecting data in as large a sample as possible. In a cross-sectional research design, each person is tested once. The kind of data relevant to a dynamic approach are different. van Dijk et al. (2011: 62) proposed three key criteria for research methods to address CDST questions:

…if we really want to know how an individual (or group) develops over time we need data that is dense (i.e. collected at many regular measurement points), longitudinal (i.e. collected over a longer period of time), and individual (i.e. for one person at a time and not averaged out).

The differences between approaches can upset the expectations of readers. One of the implications of shifting the focus so radically is that the gatekeepers in the field – the reviewers and editors of grant applications, journal articles, dissertation committees and so on – may be accustomed to applying a set of principles drawn from one methodological approach that are inappropriate in the other. For example, in a typical statistical approach, a large sample size is highly valued but in a dynamic approach, a large sample size may be overwhelming because of the density of data at the individual level.

The comparison of quantitative research methods with dynamic ones shows that the perspectives are quite different – not irreconcilable – but different. To provide a more complete picture, the next section contrasts qualitative and dynamic methods.

Qualitative compared with CDST approaches

Qualitative studies are often designed to describe the phenomena under study and to explain what is happening from the participants’ perspective, and sometimes in their voice. Compared to quantitative studies, qualitative studies usually make fewer a priori assumptions about the data and instead carefully interpret the information from interviews, focus groups, textual analysis and other multimodal sources. Qualitative investigations typically allow for the openness of the systems discussed above, are often narrative in style, tend to be closely tied to context, and are often concerned with specificity and uniqueness by examining smaller samples in depth. In terms of describing the dynamics and openness of a given situation, qualitative methods have distinct advantages and require generally less substantial adjustments than a quantitative approach from a CDST perspective.

However, qualitative data are not inherently complex or dynamic. Unless a researcher is careful to include it in the design of the study, identifying the timescale under discussion can sometimes be difficult. The type of data utilized will depend on the specific research projects and questions. Many tools can be designed to focus on stability or variability such as narratives, journals or interviews. Typically, qualitative data is based on self-report data and this implies a host of methodological concerns such as impression management, memory problems, degree of awareness about the issues under investigation and ability to articulate responses. To capture the dynamics created by interacting systems, data can be collected in an ongoing fashion or retrospectively. However, in both cases, the processes that contribute to the dynamics of change may or may not be known by the respondents who typically are asked to account for events as they understand them. Ongoing approaches to data collection are possibly the most promising for investigating dynamics, enabling a research perspective on before, during and after the process under investigation. Capturing the ongoing dynamics as close to the moment they happen allows for a high density of data collection points as well as a reduction in the problems associated with memory and distance to the actual events in action. In contrast, retrospective research risks gaps in relevant data as well as a lower degree of data density.

One of the advantages of qualitative data is that it enables a holistic perspective on phenomena and processes and can potentially reveal more of the complexity of the system than an approach which reduces or fragments variables and processes. However, there is a danger that researchers might use the conceptual framework of CDST as a meta-theory to seek to explain data without the appropriate design to support it. As CDST becomes more widely known as a research approach, there is a risk that it becomes merely a nod to the methodological fashion of the moment. We should caution against dressing a traditional study in CDST clothing because it is something new or different. A gratuitous mention of CDST is not appropriate or even relevant unless a CDST perspective has been applied throughout the design of the study, data collection and analysis process.

Research Examples of CDST in Language Learning

To research dynamics from a CDST perspective, the central concern is not whether the data are quantitative versus qualitative in nature, but rather how the data are collected and analysed. Many cross-sectional and longitudinal designs to date can claim to have looked at change over time but typically they are not examining the dynamics in action and the actual processes of change which is the focus of a CDST perspective. Note also that longitudinal data in the form of test-retest designs where data collection occurs on only two occasions would not usually lead to the density of data required to assess dynamic changes as they happen. Furthermore, qualitative data do not automatically meet van Dijk et al.’s (2011) criteria for CDST data unless that data explicitly targets interacting processes. Data that best address the concerns of CDST are dense, longitudinal and individual. Here the term ‘longitudinal’ implies specifying a timescale for study which might be a long time period (measured by change over the years, the typical sense of ‘longitudinal’) or the timescale can be relatively brief measured by change over a few minutes which also can be considered longitudinal, especially if change is rapidly occurring.

The types of data required for dynamic studies are best obtained from methods designed to be dynamic from the outset rather than retro-fitting a study that used one-time measurement or pre-post designs. We have undertaken several such studies and found them to be quite challenging. We made our share of mistakes along the way. In doing so, we learned some lessons that can be shared in the hopes of allowing future researchers a smoother experience. In this final section, we will share some of those challenges, problematize them and describe the solutions we have worked out. The topics below include some of the most widely studied topics in the psychology of language learning, reinforcing the notion that the goal is to look at familiar processes in a different way to create new understandings and challenge existent conceptualizations.

Willingness to communicate

WTC was originally defined and studied in the native language communication literature, reflecting a stable predisposition to approach or avoid communication. The original WTC scale presented a set of 12 situations where a person might choose to initiate conversation with friends, acquaintances or strangers in a public speaking situation, large meeting, small group, or a dyad. In the L2 literature, additional measures of WTC were developed to focus also on tendencies to communicate inside versus outside classrooms, different skill areas and in different educational contexts (MacIntyre & Ayers-Glassey, in press).

The shift to a more dynamic perspective began most notably with MacIntyre and Legatto (2011) who examined WTC from a dynamic perspective using what has become known as the idiodynamic method. The challenge of adopting the dynamic approach was to measure WTC on a near continuous basis, in real time. Given that it is not possible to have research participants both speak and rate their WTC simultaneously, the solution was to record the speech on video for immediate playback. A second challenge was to provide numerical ratings of WTC that showed meaningful fluctuations over time. To capture those ratings required new software which could play back the participants’ video and synchronize it with the WTC ratings made by the participant every second. The video serves to stimulate recall and immediate playback mitigates memory biases that affect reporting of previous affective states. This procedure allows researchers to gather a series of numerical ratings, linked in time with the video playback and to capture fluctuations in WTC longitudinally but over a short period of time (less than 5 minutes). The method also requires an interview with the person making the ratings to explain their rationale for increases and decreases in WTC scores, and allows for interviews with interlocutors and other observers. The interviews help to account for the unseen thought and emotion processes that contribute to communication.

Data analysis in this study examined both tendencies at the group level and analysis within persons, and what the authors called analysis in both horizontal and vertical directions. Results at the group (horizontal, across persons) level showed, unsurprisingly, that some tasks produced more WTC and longer speaking times than others. Results of the vertical analysis (within an individual) showed that there were meaningful exceptions to those trends. Further, results emphasized the intricate, continuous interactions among WTC, anxiety, vocabulary retrieval from memory, prior experience, self-presentation and nonverbal communication. In presenting the pattern of WTC fluctuations the authors emphasized that WTC was an attractor state built in part as an evolution from the previous state of the system; once a task was initiated, the respondent tended to continue even if there was some difficulty encountered and anxiety was aroused.

Motivation

Much of the literature on motivation has been studied from the theoretical standpoint of either the integrative motive (Gardner, 2010) or the L2 self system (Dörnyei, 2005). In both cases, motivation is considered over a long timescale of months and years. The integrative motive refers to a long-term process of taking on valued characteristics of another group. Similarly, the L2 self system reflects relatively long term processes including accumulated L2 experience, a sense of obligation (ought-to self), and the ideal future self. The most common research methods involve questionnaire-based measures of motivation.

MacIntyre and Serroul (2015) studied motivation from a dynamic perspective, over a short time period. The first challenge in doing so was to define the specific dimensions of motivation on which to focus. The problem in practical terms was to define a quality of motivation that changes on a per-second basis. The solution was to go outside the language arena, back to motivational basics, and look at approach and avoidance motives because they are defined as changing from moment-to-moment (see Epstein & Fenz, 1965). The experimental task featured a structured set of eight questions, administered orally by a research assistant, to be answered in the L2, similar in format to an oral quiz. Results showed fluctuations in approach-avoidance motives that were tied closely to interactions among attributes of the specific tasks, ongoing vocabulary retrieval, emotion processes, self-related cognition and other processes. There was no evidence of influence from integrative motivation or the L2 self. This pattern was somewhat surprising given the relative lack of emphasis on tasks as the driver of motivation in the L2 literature, but in retrospect the results make sense given the nature of the situation and timescale under study. The issues raised by central concepts in both integrative motivation and the L2 self system seem to refer to processes occurring over longer timescales or which might not be applicable in the specific experimental situation created by the study. In this study, the change in methods and timescales precipitated revisiting theory to find a perspective that better suited a CDST study on a per-second timescale. This suggests not only practical changes and fresh insights when researching from this perspective but also potential benefits offered by revisiting theories through a new lens.

Anxiety

As is the case for WTC and motivation, language anxiety has been studied widely in SLA. The theoretical approach to anxiety research was originally a mix of concepts adapted from other areas (including trait anxiety and test anxiety) with uncertain applicability to language learning contexts. In the mid 1980s, however, a specialized approach created scales to measure typical levels of anxiety in language-related contexts including classrooms (Horwitz et al., 1986), language use in the community (Gardner, 1985) and anxiety tied to specific skills such as reading and writing (Horwitz, 2017).

Gregersen et al. (2014) studied anxiety from a dynamic perspective linking anxiety arousal with physiological changes (heart rate) in the context of delivering a classroom presentation by L2 learners. One challenge of this study was to coordinate the measurement of anxiety and heart rate in real time. Physiological measures such as heart rate are tied to specific measurement timescales (e.g. beats-per-minute). The recording equipment used, similar to the heart rate tracking band that a runner might wear around the chest, also produces missing data when the sensor briefly loses contact with the skin as a person moves around. One significant problem was that the per-second timescale of measurement produced fine-grained information but with many missing data points. The solution was to define segments of a much longer communication event, averaged over longer time periods, which could be graphed together. By averaging ratings within a segment of communication stable measures of both anxiety and heart rate were generated, removing some of the noise that affects data collection under real world conditions (i.e. outside a controlled laboratory setting).

A second challenge presented by this study was the process of data summary and interpretation. In addition to analysing the connection between heart rate and anxiety ratings, the upward and downward trends or slopes in the line graph of participants’ reactions were assessed to track the periods in which anxiety was rising and falling. The inspiration for this approach was Newton’s famous laws of motion which were taken as analogous to the ‘motion of emotion.’ In essence, the tendency to experience anxiety was taken to be subject to external influences that provide impetus to change the emotional trajectory of the participant, either upward toward more anxiety or downward toward lower anxiety. The slope of the resulting line of anxiety ratings reflects the strength of the resulting force of change. Connecting multiple sources of data gathered in real time allowed the researchers to address the challenge of coordinating different data sets on multiple timescales.

This study showed how dynamic perspectives can challenge assumptions from more traditional studies, such as the correlation studies that are prevalent in the literature. Certainly there is value in taking long term stability into consideration, as typically less anxious speakers prepared and performed differently than the typically anxious ones. However, the anxiety reaction of a usually less anxious speaker seemed to be different from those more accustomed to the experience. Traditional quantitative methods would have glossed over this unusual case, treating its variability from the norm as error.

Anxiety + enjoyment

Although anxiety has been well studied in SLA, other emotions have not been widely investigated. There is a recent series of studies on language enjoyment that provide a positive emotional counterpart to the large number of anxiety studies that exist. Many of the enjoyment studies use the Dewaele and MacIntyre (2014) Foreign Language Enjoyment scale (Dewaele & MacIntyre, 2016; Dewaele et al., 2016; Dewaele et al., 2017).

Boudreau et al. (2018) studied the dynamics of not one but two emotions, anxiety and enjoyment. The idiodynamic approach allows for studies of the coordinated actions of different emotions. The challenge was to conceptualize how the emotions might jointly operate. Although prior questionnaire-style research showed a significant negative correlation between them ranging between −0.24 and −0.34, the size of the correlation suggests anxiety and enjoyment are not mutually exclusive or opposite ends of the same continuum. From a CDST perspective, the challenge was to measure the emotions with two separate ratings. The solution was to employ the idiodynamic software twice, once to get ratings of anxiety and a second time to get ratings for enjoyment, in counter-balanced order. A second challenge was to show the relationship between anxiety and enjoyment during free-flowing communication. The solution was to create meaningful segments of time wherein a communicative event occurred, such as during a complete, fluent utterance on a single topic, and then examine the coordinated trends for anxiety and enjoyment. Doing so allowed a description, within segments, of the periods in time where anxiety rises and enjoyment falls (indicative of negative correlation), and the reverse tendency of falling anxiety and rising enjoyment (also indicative of a negative correlation). Further, there were occasions where anxiety and enjoyment both were increasing at the same time, a pattern not predicted by the negative correlation found in questionnaire research. Furthermore, results showed that the various correlational patterns were found even within the same person, suggesting that the relationship between anxiety and enjoyment can change from negative to positive based on what is happening in the communication context. The results of this study show the value of CDST concepts such as emergence, dynamic stability and soft assembly.

Teacher stress

The teacher has been a somewhat neglected topic in the psychology of language learning, with much more of the focus placed on learner psychology. Prior research suggests that language teaching can be a stressful occupation (Hiver & Dörnyei, 2015).

A study currently underway is examining teacher stress using the experience sampling method (Talbot et al., 2019). Experience sampling has been a preferred method of studying flow experiences in daily life (Csikszentmihalyi, 2014). The method provides research participants with a device (e.g. a pager or smartphone) that beeps at various points in time during the day. After receiving the beep, respondents answer questions about the activity they were performing at the moment. The method allows researchers to collect data in real time, as immediately as possible. Our study examined sources of teacher stress and uplifts, which are moments of positivity. We sought a diverse sample of teachers who were willing to install a new app on their phone. The app administered established, multi-item questionnaires on topics such as personality, wellbeing and stress (assessed on one occasion, not dynamically) along with multiple daily notifications (8 per day) to assess current stress levels dynamically through a series of structured responses that quickly described the context in which the beep arrived (e.g. I am in school, I am at home, etc.) as well as self-ratings of stress or uplifts.

The study examined 47 teachers, producing data points numbering in the tens of thousands. Summarizing that data is a challenge. On the one hand, individual level data analysis allows tracking of stress ratings over the time frame of the study. We can contextualize those fluctuations using information from questionnaires (teacher personality, wellbeing, life stressors) and demographic information. Yet this is not (yet) a description of a complex dynamic system. We have identified three teachers who scored high and three who scored low in wellbeing for case study analysis that triangulates information about the teacher and changes in stress over time. The challenge we face as researchers is to integrate the data to show the interactions among factors that contribute to teachers’ stress as part of the system of teachers’ emotional reactions to events, and not simply that there are highs and lows in the dynamics in stress ratings over time. We will look closely at the changes in stress ratings over time, asking questions such as ‘how quickly do those ratings change?’, ‘what is the teacher doing when she/he is making the ratings?’ and ‘how might additional information such as employment status or family context be factored into the changes in stress reaction?’ Preliminary results show the complex patterns of stress reactions in which a teacher’s home life, relationships, obligations, leadership activities and so on are interacting with both teaching and non-­teaching activities to create stress on some occasions and uplifts on other occasions. Our plan is to identify the signature dynamics related to stress for the teachers who have high wellbeing scores and those with low wellbeing scores to see if there are discernible differences between the teachers. After the case studies are complete, the rest of the data set can be interrogated using the case study information to focus analysis. The data present challenges because a large number of data points need to be summarized without losing the complex interactions they show.

Self

The self in its various guises has been an increasing focus of research in SLA (Mercer & Williams, 2014). As a notion, it has been fragmented into various constructs designed to capture various aspects of self and typically reflecting different research perspectives. For example, self-­efficacy is very tightly defined and typically measured through questionnaires, whereas identity is more broadly defined and connected to specific social roles and contexts, and is typically the focus of qualitative work.

Mercer (2015) has defined the self as a complex dynamic system. She has studied how the self functions on different levels of perception and across different timescales. The focus has been on how the different facets and aspects of self interact to create an overall emergent sense of self. In her 2015 study, Mercer focused on the dynamics of the self on different timescales. To do this, she collected data with four volunteer students on the timeframe of seconds/minutes using idiodynamic software, across minutes/hours within class using a survey, across days/weeks using journals and across weeks/months using interviews. The data were analysed on each level for dynamics and these were compared. She found that there were similar patterns of dynamics and ‘if… then’ signature dynamics (Mischel, 2004) across timescales. This refers to patterns where IF this happens, THEN this is likely to happen. These findings raise interesting questions about the role of fractalization in the dynamics of systems and thus the possibility of predicting dynamic patterns across timescales. The challenge for researchers is how to best integrate dynamic research across different timescales.

In summary, this brief overview of specific research reveals how studies designed and implemented from a CDST perspective are beginning to show some of the possibilities afforded by the approach. WTC, motivation, anxiety, enjoyment, stress and the self have been studied extensively from traditional quantitative and qualitative perspectives. Yet, there is something new about the CDST account that foregrounds interactions among concepts and provides novel insights into the processes underlying familiar concepts.

Insights for Researching Language Learning Psychology from a CDST Perspective

The above illustrative studies and their research methods have presented a number of challenges. Real time data are messy, dense and can be difficult to summarize without losing the nature of the dynamics. Issues that arose were addressed sometimes by blending data sets and forms of analysis at both the group and individual level, with emphasis on the latter. We have also experienced challenges when writing about CDST and the need to use language that can be conceived as jargonistic to those unfamiliar with the notions and concepts underlying CDST as well as the need to explain extensively the theory before getting to the study itself.

Yet, despite the challenges, the benefit of a CDST perspective in the above studies was that it allowed us to offer a different interpretation of what was happening than traditional methods would allow. We feel there are five main types of insights such work can afford:

(1)Exploring different timescales/levels of granularity: In a complex dynamic system, there are many ways of defining a system and it depends where a researcher chooses to set the boundaries and focus. A guiding principle for deciding what is being defined as a system is to consider what functions as a whole (Larsen-Freeman & Cameron, 2008). This can then be considered in terms of levels of granularity and the related timescales of dynamism. By starting with the explicit recognition of the timescale under study, we are able to set limits on the processes under consideration, providing focus for the studies. The idiodynamic studies reported above specified a brief timescale measured in seconds and minutes, the ESM study deals with a timescale of days in a week. But as the notion of fractals suggests, these specific timescales are nested within other timescales, as the study of the self demonstrates. Patterns observed on one timescale can sometimes be found at others. The challenge becomes how to combine research examining systems at various levels of granularity and across timescales.

(2)Investigating various forms of dynamism and stability: The focus of this chapter has been on how systems change or remain stable over time. Dynamism does not have to mean change from one state to another; it can mean fluctuations within a relatively stable state. Dynamic stability can be observed as changes in processes on lower timescales may not have noticeable effects on higher levels of granularity and across longer timescales. Systems may show homeostasis. The challenge is how to view different degrees and types of dynamism and stability and how these may interact across levels and timescales.

(3)Focusing on ongoing processes: When examining dynamics, we have stressed that single data collection points may establish a snapshot of the state of a system for the participants in the research, but the research design does not necessarily reveal processes in action. Research must be designed from the start to examine the actual processes of change as they happen and this necessitates specific research designs with dense data collected at appropriate points in time.

(4)Examining uniqueness AND commonalities: A CDST perspective does not imply that scholars ignore patterns across systems or parts of systems, there is much to be learned from examining how findings may be similar across individuals and units of analysis. However, CDST does foreground strongly the benefits to be gained from looking at outliers, the unusual, the unique and the unexpected. These reflect the core characteristics of a complex dynamic system.

(5)Taking holistic perspectives and conceiving of open systems: The psychology of language learning has defined, differentiated and measured a number of specific concepts such as motivation, anxiety and WTC. However, as we observe them in operation in real time, many theoretical distinctions melt away. A CDST approach requires putting concepts such as motivation, anxiety and WTC in motion, describing how they move together, and what constitutes a meaningful system. The notion of open systems provides a lesson in contrasting the group level of data analysis with individual level. Traditional quantitative methods require defining the concept in the study in advance and measuring them with appropriate instruments. This has the effect of closing the potential influences on the system to only those defined in advance. By allowing for open systems, and gathering qualitative data repeatedly to assess how the processes interact over time, we are able to identify influences from the perspective of a language speaker or teacher that would have been missed otherwise. This allows for a richer description than otherwise would be possible.

Conclusion

In this chapter, we have reflected on our understandings of what researching language learning psychology from a CDST perspective implies. We have considered how the kinds of questions, use of theoretical frameworks and constructs as well as methodologies and forms of analysis must adjust to accommodate the specific characteristics of a complex dynamic system and its commensurate meta-theory. We have focused in particular on the dynamics of such systems and looked at several illustrations of research focused on dynamics, how such studies need to be conducted, the challenges posed by such work and the fresh insights they can generate. We hope to have inspired others who may wish to research from this perspective and examine dynamics specifically. We have shown how challenges can be creatively met and how the insights gained from such work can push forward our understandings to challenge our established and conventional views of constructs and their interrelationships. A CDST perspective is still in its relative infancy in the field and its merit for language learning psychology will be tested out over time. As a theory, it must offer something new and worthwhile, making a substantial contribution of growth to the field; otherwise, it becomes merely another short-lived academic fashion or fad. We feel this perspective offers rich potential, especially for the field of language learning psychology. As we have argued in this chapter and as we believe our work to date has shown, CDST is an innovative lens through which to reflect on the complex and dynamic nature of human psychology and its collective as well as individual characteristics. We feel exciting research times lie ahead and the challenges inherent in such work will be met when scholars work together with a spirit of innovation, criticality and openness to new ways of thinking and researching.

References

Al-Hoorie, A. (2018) The L2 motivational self system: A meta analysis. Studies in Second Language Learning and Teaching 8, 721–754.

Boudreau, C., MacIntyre, P.D. and Dewaele, J.-M. (2018) Enjoyment and anxiety in second language communication: An idiodynamic approach. Studies in Second Language Learning and Teaching 8 (1), 149–170.

Canale, M. and Swain, M. (1980) Theoretical bases of communicative approaches to second language teaching and testing. Applied Linguistics 1, 1–47. doi:10.1093/applin/1.1.1

Csikszentmihalyi, M. (2014) Validity and Reliability of the Experience-Sampling Method. New York: Springer.

de Bot, K. (2014) Rates of change: Timescales in second language development. In Z. Dörnyei, P.D. MacIntyre and A. Henry (eds) Motivational Dynamics in Language Learning (pp. 29–37). Bristol: Multilingual Matters.

Dewaele, J-M. and MacIntyre, P.D. (2014) The two faces of Janus? Anxiety and enjoyment in the foreign language classroom. Studies in Second Language Learning and Teaching 4 (2), 237–274.

Dewaele, J.-M. and MacIntyre, P.D. (2016) The predictive power of multicultural personality traits, learner and teacher variables on Foreign Language Enjoyment and Anxiety in classrooms. Presented at the International Association for Language and Social Psychology, Bangkok, Thailand.

Dewaele, J.-M., MacIntyre, P., Boudreau, C. and Dewaele, L. (2016) Do girls have all the fun? Anxiety and enjoyment in the foreign language classroom. Theory and Practice of Second Language Acquisition 2, 41–63.

Dewaele, J.-M., Witney, J., Saito, K. and Dewaele, L. (2017) Foreign language enjoyment and anxiety: The effect of teacher and learner variables. Language Teaching Research, 1–22.

Dörnyei, Z. (2005) The Psychology of the Language Learner: Individual Differences in Second Language Acquisition. London: Routledge.

Epstein, S. and Fenz, W.D. (1965) Steepness of approach and avoidance gradient in humans as a function of experience: Theory and experiment. Journal of Experimental Psychology 70 (1), 1–12.

Fogel, A. (2006) Dynamic systems research on interindividual communication: The transformation of meaning-making. Journal of Developmental Processes 1, 7–30. See http://www.psych.utah.edu/people/people/fogel/publications/Fogel-Dynamic%20Systems%20Research.pdf

Gardner, R.C. (1985) Social Psychology and Second Language Learning: The Role of Attitudes and Motivation. London: Edward Arnold Publishers.

Gardner, R.C. (2010) Motivation and Second Language Acquisition: The Socio-educational Model. New York: Peter Lang.

Gregersen, T., MacIntyre, P.D. and Meza, M.D. (2014) The motion of emotion: Idiodynamic case studies of learners’ foreign language anxiety. Modern Language Journal 98 (2), 574–588.

Hiver, P. and Al-Hoorie, A.H. (2020) Research Methods for Complexity Theory in Applied Linguistics. Bristol: Multilingual Matters.

Hiver, P. and Dörnyei, Z. (2015) Language teacher immunity: A double-edged sword. Applied Linguistics, 1–20. https://doi.org/10.1093/applin/amv034

Horwitz, E.K. (2017) On the misreading of Horwitz, Horwitz, and Cope (1986) and the need to balance anxiety research and the experiences of anxious language learners. In C. Gkonou, M. Daubney and J.-M. Dewaele (eds) New Insights into Language Anxiety: Theory, Research and Educational Implications (pp. 31–47). Bristol: Multilingual Matters.

Horwitz, E.K., Horwitz, M. and Cope, J. (1986) Foreign language classroom anxiety. Modern Language Journal 70, 125–132. doi: https://doi.org/10.1111/j.1540-4781.1986.tb05256.x

King, J. (2013) Silence in the second language classrooms of Japanese universities. Applied Linguistics 34, 325–343.

Larsen-Freeman, D. and Cameron, L. (2008) Complex systems and applied linguistics. The Modern Language Journal 92 (4), 644–645.

Larsen-Freeman, D. (2017) Complexity theory: The lessons continue. In L. Ortega and Z. Han (eds) Complexity Theory and Language Development: In Celebration of Diane Larsen-Freeman. Amsterdam: John Benjamins.

MacIntyre, P.D. (2017) An overview of language anxiety research and trends in its development. In C. Gkonou, M. Daubney and J.-M. Dewaele (eds) New Insights into Language Anxiety: Theory, Research and Educational Implications (pp. 11–30). Bristol: Multilingual Matters.

MacIntyre, P.D. and Ayers-Glassey, S. (in press) Measuring willingness to communicate. In P. Winke and T. Brunfaut (eds) The Routledge Handbook of Second Language Acquisition and Language Testing. Abingdon: Routledge.

MacIntyre, P.D. and Charos, C. (1996) Personality, attitudes, and affect as predictors of second language communication. Journal of Language and Social Psychology 15, 3–26.

MacIntyre, P.D. and Legatto, J.J. (2011) A dynamic system approach to willingness to communicate: Developing an idiodynamic method to capture rapidly changing affect. Applied Linguistics 32, 149–171.

MacIntyre, P.D., MacKay, E., Ross, J. and Abel, E. (2017) The emerging need for methods appropriate to study dynamic systems. In L. Ortega and Z. Han (eds) Complexity Theory and Language Development: In Celebration of Diane Larsen-Freeman (pp. 97–122). Amsterdam: John Benjamins.

MacIntyre, P.D. and Serroul, A. (2015) Motivation on a per-second timescale: Examining approach-avoidance motivation during L2 task performance. In Z. Dörnyei, P.D. MacIntyre and A. Henry (eds) Motivational Dynamics in Language Learning (pp. 95–108). Bristol: Multilingual Matters.

Mercer, S. (2015) The dynamics of the self in SLA: A multilevel approach. In Z. Dörnyei, P. MacIntyre and A. Henry (eds) Motivational Dynamics in Language Learning (pp. 139–163). Bristol: Multilingual Matters.

Mercer, S. (2016) Complexity and language teaching. In G. Hall (ed.) The Routledge Handbook of English Language Teaching (pp. 473–485). Abingdon: Routledge.

Mercer, S. and Williams, M. (2014) Multiple Perspectives on the Self in SLA. Bristol: Multilingual Matters.

Mischel, W. (2004) Toward an integrative science of the person. Annual Review of Psychology 55 (1), 1–22.

Mystkowska-Wiertelak, A. and Pawlak, M. (2017) Willingness to Communicate in Instructed Second Language Acquisition. Combining a Macro and Micro-Perspective. Bristol: Multilingual Matters.

Savignon, Sandra J. (2000) Communicative language teaching. In M. Byram (ed.) Routledge Encyclopedia of Language Teaching and Learning (pp. 125–129). London: Routledge.

Talbot, K., MacIntyre, P., Gregersen, T., Mercer, S., Ross, J. and Banga C.A. (2019) Language teacher stressors and uplifts, wellbeing, and personality Traits: An International ESM study. Presented at the annual conference of the American Association of Applied Linguistics, Atlanta GA, 2019.

Teimouri, Y., Goetze, J. and Plonsky, L. (2019) Second language anxiety and achievement: a meta-analysis. Studies in Second Language Acquisition, 1–25. doi:10.1017/S0272263118000311

Thelen, E. and Smith, L.B. (1994) A Dynamic Systems Approach to the Development of Cognition and Action. Cambridge, MA: MIT Press.

van Dijk, M., Verspoor, M. and Lowie, W. (2011) Variability and DST. In M. Verspoor, K. de Bot and W. Lowie (eds) A Dynamic Approach to Second Language Development: Methods and Techniques (pp. 55–84). Amsterdam: John Benjamins Publishers.