Document not found! Please try again

A Dynamic Ensemble for Second Language ... - Wiley Online Library

35 downloads 0 Views 196KB Size Report
Jubail Industrial College. The English Language Institute. Jubail Industrial City, 31961. P.O. Box 10099. Kingdom of Saudi Arabia. Email: [email protected].
A Dynamic Ensemble for Second Language Research: Putting Complexity Theory Into Practice PHIL HIVER International Graduate School of English 17 Yangjae-daero 81-gil Gangdong-gu Seoul 05408 South Korea Email: [email protected]

ALI H. AL-HOORIE University of Nottingham School of English University Park Nottingham NG7 2RD United Kingdom Jubail Industrial College The English Language Institute Jubail Industrial City, 31961 P.O. Box 10099 Kingdom of Saudi Arabia Email: [email protected]

In this article, we introduce a template of methodological considerations, termed “the dynamic ensemble,” for scholars doing or evaluating empirical second language development (SLD) research within a complexity/dynamic systems theory (CDST) framework. Given that CDST principles have yielded significant insight into SLD and have become central to the concerns of applied linguists in many domains, we propose the need for a practical blueprint to ensure compatibility between its theoretical tenets and empirical SLD research designs. Building on “complexity thought modeling” (Larsen–Freeman & Cameron, 2008a), we present a practical catalog of 9 considerations intended to inform research design at multiple stages. We contextualize the 9 considerations of the dynamic ensemble by discussing how these have been framed and addressed within one previous CDST study. Finally, we address the issue of what practical implementation of this dynamic ensemble would entail and introduce several case-based methods for building off of the considerations in our dynamic ensemble. We hope that this user guide will help orient researchers interested in working within a complexity framework and spur continued methodological discussion in the field. Keywords: complexity theory; dynamic ensemble; methodology; research design; transdisciplinarity

NEARLY TWO DECADES HAVE PASSED SINCE Larsen–Freeman (1997) first proposed that applied linguistics issues could profit by being viewed explicitly in complexity terms, and while complexity theory1 (CDST) may not be the dominant paradigm in second language research,

The Modern Language Journal, 100, 4, (2016) DOI: 10.1111/modl.12347 0026-7902/16/741–756 $1.50/0  C 2016 The Modern Language Journal

it has gained considerable currency since then (Larsen–Freeman, 2017). More importantly, however, complexity has persisted not only because it is a useful metaphor, but because it is an empirical reality (Morin, 2008), and one that has yielded significant insight into second language development (SLD). Since Larsen–Freeman and Cameron (2008a) first offered a novel perspective on long-standing questions to which traditional paradigms failed to offer satisfying answers, CDST has virtually exploded into domains as diverse as English as a lingua franca (Baird, Baker,

742 & Kitazawa, 2014), sociolinguistics (Blommaert, 2014), multilingualism (de Bot, 2012; Jessner, 2008), educational linguistics (Hult, 2010), second language (L2) pedagogy (Mercer, 2013), and conversation analysis (Seedhouse, 2010). The fact that understandings from complexity now appear central to the concerns of most applied linguists signals that CDST is here to stay. Applied linguistics is an applied social science and therefore any proposal to borrow CDST as a framework for research must have unapologetic practical utility. However, apart from literally a handful of exceptions (e.g., Byrne & Ragin, 2009; Dörnyei, 2014; Verspoor, de Bot, & Lowie, 2011), scholarly works connecting CDST to social inquiry stop short of the level of practical application that would allow scholars to ensure compatibility between empirical research designs and the theoretical tenets of complexity. In the current article we address this issue head-on by formulating an explicit operational user guide to the CDST considerations necessary both for L2 researchers designing a study and for consumers of research evaluating these studies. All research represents a story of sorts, and most traditional L2 research reports contain features corresponding with welldefined conventions—what we might call the “story grammar” of research—that help in understanding the structure and purpose of that research. Missing from the discussion of CDST in SLD is a focus on the methodological choices scholars must make. In a sense, then, our field is still developing the story grammar for talking about complexity research, a task we agree is essential for CDST to live up to its full potential for SLD research (MacIntyre, Dörnyei, & Henry, 2015). Thus, our primary goal in this article is to build a preliminary template of CDST considerations that are important in research design, and to make these issues from CDST as transparent and pragmatic as possible. We do not provide an exhaustive treatment of the ideas or terms underlying CDST as we feel that existing overviews published in our field have much to recommend them (e.g., Dörnyei, MacIntyre, & Henry, 2015; Larsen–Freeman & Cameron, 2008a, 2008b; van Geert, 2008). Instead, our overarching aim in this article is to outline a blueprint of methodological propositions that will allow researchers to incorporate core insights from CDST. To this end we present an initial map of the considerations for scholars who are doing, or considering doing, complexity-informed empirical research. As a preliminary step, let us scope outward to highlight the CDST ethos with regard to SLD research.

The Modern Language Journal 100 (2016) THE CONTRIBUTION OF COMPLEXITY TO APPLIED LINGUISTICS Scholars examining CDST’s contribution to knowledge-making in other disciplines have underscored its function as a frame of reference (Byrne, 2011), a conceptual toolbox (Walby, 2007), a habit of thought (Kuhn, 2008), a transdisciplinary discourse (Klein, 2004), even a worldview (Cilliers, 2005). Most also emphasize that complexity has yet to be articulated in such a way that it could be termed a theory per se (Overton, 2007). Accordingly, our usage corresponds with what Larsen–Freeman (2013, 2015b) has termed a meta-theory—a set of coherent principles of reality (i.e., ontological ideas) and principles of knowing (i.e., epistemological ideas) that, for applied linguists, underpin and contextualize object theories (i.e., of language and language development) consistent with these principles (de Bot et al., 2013). The complexity meta-theory groups together a set of well-known relational principles (Overton, 2013), namely that certain phenomena involve multiple parts interacting together through dynamic, nonlinear processes that lead to striking emergent patterns over time. As a metatheory, CDST represents a set of powerful intellectual tools and concepts “capable of informing all theories” (Morin, 1992, p. 371). These conceptual tools serve as a rigorous aid to thinking and theorizing, as well as conducting and evaluating research about the human and social world. The contemporary reorientation toward using complex-systemic understandings as the foundation for human and social inquiry suggests that CDST is now a part of mainstream intellectual culture (Norman, 2011). In retrospect, however, the earliest explicit attempts to form connections between phenomena of interest to applied linguists and the theoretical principles of complexity (e.g., Ellis, 1998; Larsen–Freeman, 1997) were a radical departure from what was, at the time, the norm—though some of that sentiment had already begun to emerge in the work of other scholars who did not explicitly associate themselves with CDST (e.g., Dörnyei & Malderez, 1997; MacWhinney, 1998; Meara, 1997; van Lier, 1988). Nearly two decades on, the current level of uptake of CDST principles suggests the existence of a coherent new normal that has begun to spread dynamically throughout mainstream applied linguistics (Larsen–Freeman, 2012a, 2015a, 2015b). Clearly, no individual framework is the singular solution to the challenges of understanding the complexities of SLD (Ortega, 2012; The

Phil Hiver and Ali H. Al-Hoorie Douglas Fir Group, 2016). However, too often applied linguists let their paradigms define problems (Hulstijn et al., 2014). Nevertheless, by emphasizing dynamic change, interconnectedness, and multicausality, and by discouraging determinism, reductionism, and precise (rather than probabilistic) prediction, this new complexity agenda has provided a more useful perspective for looking at existing problems and opened the door to reconfiguring the field’s program of knowledge and making better sense of SLD phenomena (Larsen–Freeman, 2015a). As with other theoretical frameworks (e.g., critical or sociocultural theories), complexity has not offered ready-made research templates, nor should it be expected to. The real and more exciting contribution the complexity perspective has made is not purely in the realm of methods of instrumentation and analysis, but instead in ontological and epistemological considerations of how we think about the world, considerations that are linked with the issue of how we engage in scientific inquiry (Ellis & Larsen–Freeman, 2006; Ortega, 2013). The inescapable fact is that many researchers have recognized that CDST principles are indispensable for consolidating existing understanding and providing new empirical answers to long-standing questions—even in domains that did not interface with complexity in their original conceptualization or empirical validation. Examples in which CDST has become an integral part of empirical research include L2 anxiety (Gregersen, MacIntyre, & Meza, 2014), learner language (Larsen–Freeman, 2006, 2010; Lowie & Verspoor, 2015), lexical development (Ellis & Larsen–Freeman, 2009; Verspoor, Lowie, & van Dijk, 2008), L2 motivation (Dörnyei et al., 2015), L2 writing (Baba & Nitta, 2014; Verspoor, Schmid, & Xu, 2012), self-concept (Henry, 2015; Mercer, 2014), and willingness to communicate (MacIntyre & Legatto, 2011). This evidence indicates that acceptance of CDST principles is beginning to reach a critical mass in our field. THE DYNAMIC ENSEMBLE Our main objective in this article is to provide practical suggestions for how empirical L2 research designs can incorporate conceptual tools from CDST. To this end, we have constructed a blueprint termed the dynamic ensemble (Table 1). This dynamic ensemble functions as a practical catalog of complexity considerations, all of which should inform the planning and design of SLD research. It builds on “complexity thought modeling” (Larsen–Freeman & Cameron, 2008a,

743 p. 41)—originally introduced as an approach to exploring a research question from a CDST perspective—and is compatible with various existing methods as we explain in detail below. What we propose is a systematic expansion of it, and we imagine this as a user guide with questions that can be consulted at multiple junctures in the research process to inform the choice of research problems, development of hypotheses, sampling of participants, types of data collected, and analysis and interpretation. A growing number of empirical studies using a CDST perspective can be found. From them we have selected one study (Spoelman & Verspoor, 2010) that effectively illustrates many of the considerations we raise. While we acknowledge that multiple sources exist to exemplify these considerations, using different studies for each of the points would be burdensome for readers as it would require a high level of familiarity with each study. Thus, we have chosen to embed the discussion of this single article within our text, using it to contextualize the nine considerations of the dynamic ensemble. We introduce considerations from our blueprint before explaining how they have been addressed by Spoelman and Verspoor, then also recommend some further improvements in light of our considerations. Certain considerations may be more prominent than others simply because their study was not designed to align systematically with the dynamic ensemble. We should note that by choosing this one study, we are not elevating it to model status. Rather, Spoelman and Verspoor’s study provides a richness of evidence framed from a CDST perspective that facilitates consideration of a number of the aspects we wish to highlight—as well as their interrelationships. We begin here with a brief summary of this article. In their study, Spoelman and Verspoor (2010) investigated the development of accuracy rates and complexity measures in the learner language of a Dutch adult learner of Finnish. Using data (i.e., 54 writing samples) produced over the course of 3 years, these researchers analyzed the learner’s language around (a) distinct sources of complexity, for which developmental patterns were investigated at the word, noun-phrase (NP), and sentence levels; and (b) overall accuracy, for which they calculated the development of case error rates relative to the overall number of cases used (the Finnish case system comprises 15 cases). They examined the variability in the learner’s language by first plotting the accuracy and complexity score range for each measurement occasion (i.e., 1 to 54) in a moving min–max graph plot;

744

The Modern Language Journal 100 (2016)

TABLE 1 The Dynamic Ensemble Operational Considerations

Systems

Level of Granularity

Contextual Considerations

Context

Systemic Networks

Macro-System Considerations

Dynamic Processes

Emergent Outcomes

Micro-Structure Considerations

Components

Interactions

Parameters

What is the complex system under investigation? What gives this case phenomenological validity? Who are the agents in the system? On what timescale(s) will the system outcome(s) or behavior(s) be examined? What type(s) and what level(s) of data are required to study the system? What are the contextual factors that are part of the environmental frame of reference for the system, its dynamic actions, and its patterned outcomes? How are these contextual factors formalized into system parameters that influence behavior? How does the system adapt to the context it is embedded in, and vice versa? To which other systems (i.e., nodes) does this system link? What is the nature of these networked relationships? What processes ensue in coordination with other systems? When and how should these links be highlighted explicitly and investigated? What general principles of change exist for this system? What specific mechanisms of change are present in the system? What trajectory has the system followed, and how did it get to where it is? What causal signature dynamics (e.g., self-organization) produced the system outcomes, and why? What salient dynamic outcome configurations (i.e., attractor states) emerge for this system, and why? What are the characteristics of these patterns of stability for the system in the state landscape? What variability exists around these patterns of stability? What are the parts that make up the system under investigation? Which are the most prominent components of the system in a given process of change, or for an emergent outcome, and why? What types of relationships exist between system components, and what are their characteristics? How do these exchanges manifest and affect system behavior? How do these relationships change over time? What are the constraints and specifications that mediate the changes and interactions possible within a system, and how do they determine the system’s behavior? What are the critical dimensions or values of a system (e.g., the motors of change) which, when they fluctuate, may result in a change in outcome?

then they calculated raw and detrended correlations between variables to compare the statistical similarity of the time-series measurements; finally, they tested for the probability of interaction between these variables using Monte Carlo simulation. Their data demonstrated that the nonlinear

variability seen in learner language provides valuable insight into L2 developmental phenomena, highlighting that the interaction within a linguistic domain (e.g., between word, NP, and sentencelevel complexity) illustrates how developmental systems are often in competition for resources.

Phil Hiver and Ali H. Al-Hoorie Operational Considerations Systems. In the complex social world that is our research stage, what do we take as the basic unit of analysis? A unit that reflects this complex reality is a complex system.2 We agree that “there is nothing metaphysical about complex systems” (Cilliers, 2000, p. 31), and we propose that the most tractable approach would be to restrict the term system to something that has phenomenological validity or concrete existence. Although they are less definitive about this than with regard to other considerations, Spoelman and Verspoor (2010) do provide some indication that the system under investigation is their learner’s writing development in L2 Finnish, which is phenomenologically valid in light of the learner minoring in Finnish for 3 years and having to produce written homework assignments on academic topics throughout this period of study. This view, that systems are real entities that reflect the operation of actual causal mechanisms, is consistent with the notion in social complexity of a case as a complex system (Ragin & Becker, 1992; Uprichard, 2013). Casing is the act of specifying the phenomenological boundaries of a unit of analysis for investigation, although boundary does not imply closure. In light of the growing consensus that the learning process cannot be separated from the learner (Larsen–Freeman, 2012a), we would suggest that in the human and social domains a necessary additional dimension of a system is an agent (de Bot et al., 2013). Distinct from broader meanings in CDST, by agent here we mean people, or collections of people, capable of exercising independent choices or intentional actions that contribute causally to any behavior of the system (Al–Hoorie, 2015). In their study, Spoelman and Verspoor (2010) position their learner, a young adult native speaker of Dutch with no previous knowledge of the Finnish language, as the agent of the system. Apart from an individual, any of the following could also be cased as a complex system: a group, a social movement, a language classroom, a community of professionals, or an institution. The individuals participating in each of these systems are the agents, with the smallest system possible having only a single agent. To extend existing definitions (de Bot & Larsen–Freeman, 2011, p. 9; Larsen–Freeman & Cameron, 2008a, pp. 36–38) with these criteria, for SLD research purposes a complex system (a) has concrete phenomenological validity, (b) is composed of multiple connected and interacting parts, including an agent (or agents), (c) is open to adaptive feedback and dynamic,

745 nonlinear change in behavior, (d) is part of the context that is part of it, and (e) exhibits emergent outcomes. Spoelman and Verspoor’s (2010) choice of a system, indeed, satisfies these criteria. Caution must be exercised, however, not to take for granted that the unit under investigation is a complex system (Larsen–Freeman & Cameron, 2008b); it may actually be a simple or closed system, or not a system at all. Viewed from our current definition, therefore, constellations (e.g., goals, interest) and abstract phenomena (e.g., L2 proficiency, L2 motivation) differ from systems because they do not produce an outcome by themselves, and must first be located within an agent who experiences and acts on them. Level of Granularity. In designing a study, deliberately deciding what to investigate as a complex system, and the timescale(s) (i.e., the temporal window or duration at which a process is to be studied) and level(s) (e.g., a whole-system level, a micro-components level) at which to investigate and analyze that system will impact the type of questions appropriate for exploration, the types of evidence that can be collected, and ultimately the theoretical and empirical advances made (Eve, Horsfall, & Lee, 1997). In Spoelman and Verspoor’s (2010) study, for example, the timescale on which SLD is examined spans a period of 3 years, with samples of learner language data collected at 54 intervals from written homework tasks. This type of data, collected longitudinally and analyzed in a multivariate and dynamic way, allowed the authors to explore intra-individual developmental patterns in the learner’s accuracy and complexity. The higher the dimensionality of a system (e.g., the universe) the more data is necessary for a valid representation of it, and because system changes and stabilities occur continuously across time and levels of activity, this sort of modeling would approach the complex reality of the system proper (Byrne, 2011). The goal in CDST research will rarely be to represent the entire complex system in question. However, settling on precise levels of detail in data collection and analysis, as Spoelman and Verspoor (2010) do, for instance, in limiting their analysis to three related measures of complexity (i.e., at the word, NP, and sentence level), can contribute to understanding complex phenomena without knowing the entire hierarchy of nested levels and timescales. One strategy for settling on the appropriate level of granularity, recommended by Lemke (2000) and de Bot (2015), is to investigate dynamic phenomena at a timescale of primary

746 interest along with two adjacent timescales and generating complementary data at these corresponding timescales. Another pragmatic strategy for telescoping the analytical focus or perspective breadth of a study is through simultaneously exploring only the most conspicuous segments, aspects, or interactions of a system (e.g., complexivists studying history may deal with either whole epochs, individual biographies, or historic events; Kiel & Elliot, 1996). The goal here is to systematically produce coarser or more finely grained research designs as appropriate, rather than partitioning a system purely for the sake of convenience. Greater clarity and precision in choosing an appropriate level of granularity often narrows the focus of attention, producing relatively arbitrary boundaries in data collection and analysis (Ulrich, 2001). In their study, Spoelman and Verspoor (2010) acknowledge that SLD can take many shapes, given that complexity and accuracy both manifest themselves in multiple domains of language and implicate various underlying linguistic, cognitive, and psycholinguistic correlates. Thus, in pragmatically deciding the level of granularity to adopt in a study, researchers will rarely be able to claim comprehensiveness of all the considerations that bear on a phenomenon (Cilliers, 2001). Nevertheless, when applied judiciously and offset by critical transparency, these strategies present the opportunity to produce structural definition in data and still capture the complex causal dynamics of a system without idealizing away its essential aspects. Contextual Considerations Context. Grounding a system in a context is crucial to understanding its behavior and outcomes, as context is an integral part of any system under investigation (Ushioda, 2015). Context encompasses the background situational features in place before interactions occur among system components, which can either limit or facilitate certain outcomes. These conditions will include features that are directly observable (i.e., that can be recorded or measured) or that are otherwise empirically relevant (i.e., salient in the dataset) to the makeup and location of the system, and which co-adapt with it. Spoelman and Verspoor (2010) highlight surface-level contextual aspects of the system: The agent of the system was an undergraduate theoretical linguistics major learning Finnish at a Dutch university over a period of 3 years. Finnish happens to have one of the more elaborate case systems. Their study deliberately examined this learner’s SLD at

The Modern Language Journal 100 (2016) the earliest stages of proficiency, and the written homework task conditions involved available reference materials and no time pressure. Thus, context first functions as a way of bracketing the system within an environment and giving ecological coherence to that system, its actions, and its states (Byrne & Ragin, 2009). However, outcomes and change not only emerge in context, they are also mediated and adapted by contextual factors (Radford, 2008). Spoelman and Verspoor (2010) do not elaborate on how the contextual factors they highlight might mediate, influence, or adapt to dynamic mechanisms of change in the learner’s SLD. As we have mentioned, a system is inseparable from the context that is part of it (Mercer, 2016); unlike with closed systems, contextual factors are a major determinant of complex system behavior and outcomes, which may be “formalized into the system parameters” (Larsen– Freeman & Cameron, 2008a, p. 68). The notion that dynamic mechanisms of change, in interaction with the context, produce a causal force by which outcomes appear, is not new in our field (de Bot et al., 2013), and intensely concentrating attention on dynamic mechanisms for change and stability should not lead researchers to treat context as an optional add-on for explaining system development. Just as knowledge often cannot be made sense of fully if it remains separate from a wider schematic context, these system dynamics are only made meaningfully coherent when framed in a social setting (Vallacher, van Geert, & Nowak, 2015). It follows, of course, that a researcher should aim to obtain intimate knowledge of the system and its context. One of the key elements of CDST research is to determine “the range of transferable application of any processual and causal knowledge” (Byrne, 2011, p. 155), and because context is a key causal factor for any dynamic change in a system, this is best done by referencing the context in which the system under investigation is embedded. Systemic Networks. If systems are the basic structural building blocks of the complex social world and context is pragmatically constrained to include the social, cognitive, and psychological aspects that form the immediate environment of a system’s ecology, then a network is the architectural superstructure in which these systems are embedded (Kadushin, 2012). This multi-node hub is composed of interconnected systems, their relationships, and processes which together form the foundational web of the complex social world. It may help to think of this web as a CDST

747

Phil Hiver and Ali H. Al-Hoorie equivalent to a nomological net, which functions as a specification of the phenomenological concepts or theoretical constructs of interest in a study, their observable manifestations, and the linkages between them. While they do set out to examine complexity and accuracy organically and longitudinally, with their main focus on dynamic variability, Spoelman and Verspoor (2010) do not comment in detail on the networked relationship between systems. Among other networked systems, these scholars could have considered the ways in which the L2 learner’s Finnish writing development linked and coordinated with aspects of her underlying knowledge base and its cognitive representation, with the ongoing L2 classroom instruction she was receiving, with her formation of a multilingual identity, or with her ongoing academic performance in linguistics (her major). Because any single system will be just one of many nodes embedded within this dynamic network of interwoven systems, progressively unraveling which systems are networked and the precise dimensions in which they reciprocate will often form a sophisticated agenda for a given strand of CDST research. Considering interconnectedness also relates to the conceptual abstractions opened up by the analysis of systems. Here, we acknowledge the possibility for systems to be conceptualized as theoretical, as some studies have done (Henry, 2015). However, our own experience illustrates that relying on conceptual and abstract distinctions to construe a complex reality is problematic for practical reasons. Time and again, editors and reviewers have asked what we are framing as the system under investigation, and how we propose to resolve the issues of boundary specification or agency. On the other hand, regarding our proposal for an agentic, phenomenologically real conception of systems, we are not suggesting that cases be treated as contiguous or networked systems simply because they were sampled together. A relevant set of cases can only be thought of as adjacent or networked if those systems are bounded together in some phenomenologically real way (Carolan, 2014). Interconnectedness, with each system taking all other systems as its global environment, is an important message of CDST research; nevertheless, this consideration should still be tempered by prudence with regard to how wide a net is cast and how deep within the network structure researchers go (Mercer, 2015). On reflection, it is Spoelman and Verspoor’s (2010) choice of granularity level—which was an appropriate timescale and level of data—that imposes some limits on their ability to contem-

plate and illustrate networks between their system and others. Despite this seeming trade-off, we would suggest that networks should be considered throughout research design, data analysis, and interpretation of results.

Macro-System Considerations We have suggested that in SLD research cases are the methodological equivalent of complex systems, and it is through choices of research questions and data collection procedures that researchers can focus more closely on systems’ dynamic processes of change or on emergent outcomes. Although in this blueprint the system dimensions of becoming (i.e., dynamic change) and being (i.e., emergent outcomes) can be seen as two sides of the same coin, we must also emphasize that L2 research to date has had a more conventional product or outcome focus. A particular added value of research from a CDST perspective comes from investigating how the process of SLD evolves over time. Dynamic Processes. The pivotal characteristic of complex systems is that of dynamic change and adaptation, which may be gradual or dramatic. Whereas emergent outcomes (see subsequent discussion) account for what a system is doing now and the state in which it has stabilized, adaptive change provides a temporal narrative for the process of how and why the system got here and where it may be going. Through moving window graphs, for instance, Spoelman and Verspoor (2010) provide an intuitive visual representation of the shape of the system’s growth as a nonlinear learning curve “filled with peaks and regressions, progress and backsliding” (p. 535). As they experience change, systems attempt to take advantage of it by adaptively restructuring the working parts and connections—using positive feedback to amplify change, or negative feedback that dampens it (Holland, 2012). Adaptations that result in a system spontaneously, but purposefully, varying its internal structure or its higher-order function is evidence of self-organized change (i.e., not explicitly engineered). Self-organization can, thus, be seen as a robust general process that leads to emergent outcomes (Gaustello & Liebovitch, 2009). While dynamic change is constant, careful tracing of a system’s self-organization offers one solution to fingerprinting the moving target that the signature dynamics represent. For example, Spoelman and Verspoor (2010) provided longitudinal evidence that self-organization in

748 complexity and accuracy progressed in discontinuous developmental jumps (i.e., stage transitions) combined with instances of isolated stepwise growth. Tracing the modes of change that a system undergoes is one way of testing causal inferences about trajectories of change (Bennett & Checkel, 2015). Specific mechanisms of change produce a particular time signal (i.e., pattern or trajectory of change over time) in the system, which is essential for understanding the causal complexities of system development or change. Spoelman and Verspoor’s (2010) results, for instance, indicate that variability in the vicinity of a developmental jump for accuracy was highest in the earliest stages, and that the three types of complexity (i.e., word, NP, sentence-level) interacted and competed for resources until the 46th text. Each and every system has a history that plays a critical role in its trajectory of development, its dynamics, and its process of becoming (Prigogine, 1980). Just as the emergent states for a system are not unlimited, the trajectories to those outcomes are more or less finite, although the dynamic behavior may include rich variations or facets that are diachronically asymmetrical (Elman, 2003). Emergent Outcomes. Within a CDST frame of reference, the outcomes of interest to SLD scholars are no longer those found in studies taking a standard product approach and asking questions such as “Does planning time predict an increase in CAF measures on a task?” or “Are L2 learners’ attentional resources correlated with their processing speed?” Instead, CDST is concerned with emergent outcomes, tied to the notion of attractor states, which represent pockets of dynamic equilibrium that a system stabilizes into (Hiver, 2015a), and their existence explains why the complexity inherent in emergence results in some stability. Spoelman and Verspoor’s (2010) findings illustrate this particularly well: From the 28th written text onward, acquisition of 12 of the 15 Finnish cases settled, indicating system stabilization; toward the end of the data sample, the distribution of complexity types (i.e., word, NP, sentence-level), by remaining “within a steady bandwidth” (p. 547) and minimizing their competition, also showed evidence of stabilizing into an attractor state. It may seem counterintuitive that emergent outcomes at the system level have no direct counterpart at the lower component level (Holland, 2012). However, the fact that higherorder patterns of dynamic equilibrium for a system are emergent allows for a more accurate and parsimonious explanation than is possible by aggregating the individual components and their interactions ( Jörg, 2011).

The Modern Language Journal 100 (2016) To reiterate an earlier point, the number of novel emergent outcomes observable in the social world is finite (De Wolf & Holvoet, 2005). A board game provides one illustrative analogy of this. The state of a game at any given time is the placement of pieces on the board, and gameplay consists of moving pieces around the board from configuration to configuration based on the rules of the game. Similarly, the state space is the X landscape (X signifying the phenomenon under study such as L2 development) on which all emergent outcomes or states for a system phenomenon can be found. By pinpointing other occasions of emergent outcomes in their data—the 11th text for accuracy, the 12th and 42nd texts for sentence and NP complexity respectively, and the 23rd text for the relationship between accuracy and complexity—Spoelman and Verspoor (2010) capture snapshots of system equilibrium in the development landscape. One method, retrodictive qualitative modeling (e.g., Dörnyei, 2014), exploits this notion of a possibility landscape by identifying salient outcome patterns, and investigating the unique signature dynamics (i.e., the robust causal mechanisms of the “gameplay” within a system) that preceded those outcomes. Only a finite number of possible attractor states exist for a given system, and while the law of unintended consequences may still apply, identifying these may reduce much of the unpredictability of complex systems’ functioning and allow researchers to make informed choices about how to interact with respective outcomes (Vallacher et al., 2015). Micro-Structure Considerations Components. The task of describing and explaining system behavior must take into account the makeup of that complex system. This presents a dilemma, as the convention may be to position variables as the basic unit of analysis, and attempt causal-analytic explanations from these. Social complexivists have raised concerns regarding this type of design, proposing that because the social world is not composed of variables, they do not merit being reified as the entire object of research (Byrne & Callaghan, 2014; Ragin, 1997). Spoelman and Verspoor (2010) are very deliberate in their multidimensional conceptualization of complexity and accuracy, and they acknowledge that their primary perspective of SLD is one that is integrative and more ecologically valid. This reflects reality: Most phenomena in the social realm are multi-determined and dynamic. Because no single input or force governs the development and behavior of a system, change in

Phil Hiver and Ali H. Al-Hoorie system behavior is not the net causal effect of a variable’s force on a system (Larsen–Freeman, 2015b). Given the integral nature of the considerations in this dynamic ensemble, we would suggest that isolating individual components for examination—regardless of the level of sophistication applied to their analysis—cannot give a true measure of their influence. One approach to dealing with system components is to acknowledge that variables constitute partial attributes of a system—the real integral unit in the social world (Byrne & Uprichard, 2012)—and that variables may in fact be more complex and dynamic than our measures portray them to be (Michell, 2008). This perspective would also require researchers to consider context and networks (see our previous comments). Then, working from the outside in, researchers would begin by casing the systems in context that do make up the real world and scrutinizing their emergent outcomes and dynamic behavior and only then move to the component or variable level. This is precisely what Spoelman and Verspoor (2010) accomplish prior to narrowing their focus to the developmental function of withinsubject variability. Complexity inquiry rules out the possibility of adequately understanding a complex system and its behavior by examining only one level or manifestation of it. Thus, rather than seeing the whole system and its parts as being in tension, frequent “level jumping” (Davis & Sumara, 2006, p. 26) in data collection and analysis may be necessary in SLD research. Interactions. System outcomes are not the result of sums of components, but of dynamic interactions. No matter how many components are in a system, if there is no potential for components to interact, there is no complexity (Vallacher et al., 2015). Ideas from game theory (Gintis, 2009) and nonlinear pedagogy (Chow et al., 2016) provide a way for SLD researchers to explore these internal dynamics from the perspective of their manifestations and their latent characteristics. The manifestation of interaction is the specific observable behavior between the components themselves, between multiple systems, and with the environment. The latent characteristics give interactions their causal and functional coherence, and include the aim, purpose or intention of the interaction; its directionality, intensity, frequency, and duration; its utility, and the rewards or costs that accrue from it. Spoelman and Verspoor (2010) characterize the interactions within their participant’s learner language (i.e., between accuracy and complexity,

749 and between types of complexity) as simultaneously drawing on supportive and competitive relationships leading them to describe these components as “connected growers” (p. 548). This manifested, for example, in the form of overall significant competition between word and sentence complexity, but not between NP and word complexity. Interactions between components of a system and with the environment are indeed a system’s lifeblood as, without these relationships, systems would be unable to develop or behave dynamically, but to throw even more excitement into the mix, these interactions are themselves dynamic. Spoelman and Verspoor (2010) illustrate the existence of changing system interactions that are distributed across varying strengths at different periods of development. At the very early stages, for instance, accuracy and complexity appeared to compete for attentional resources, before tapering off for a longer period, followed by several more iterations of this up–down pattern. Because system change is contingent in large part on these interactions, they are essential to understanding a system’s self-organized processes and emergent outcomes (Overton, 2013). However, SLD researchers examining the characteristics of interactions, their manifestations, and how they change must remain tacitly aware that these are not singular explanatory causal mechanisms. Parameters. Parameters reveal yet another layer of the multidimensional conceptualization of causality necessary in CDST research. Each serves a complementary purpose. Order parameters (also known as constraint parameters) are the various contextual constraints and specifications that determine the changes and interactions possible within a system (Haken, 1997). Spoelman and Verspoor (2010) portray SLD as a resourcedependent process in which these resources (e.g., attention, aptitude, frequency of input) are inherently limited. The order parameter they allude to here is that the developmental load a system can sustain is restricted by these limited resources. Part of the reason why the complex dynamic social world does not exhibit infinite permutations is because order parameters reduce the degrees of freedom within which components are able to interact (Gaustello & Liebovitch, 2009). These guidelines for interaction among a system’s elements function as operating rules for interpreting system behavior. Once these rules are known, it becomes possible to make more robust observations and potentially influence movement toward a desired outcome (Morrison, 2012).

750 Control parameters (also known as driving parameters), on the other hand, are critical dimensions or values of a system (e.g., temperature, interest rates, stress, taxes) which, when they fluctuate, may result in a change in outcome (Haken, 2009). Control parameters are particularly useful for intentionally inducing change in a system, and there may be large sets of control parameters present for any complex system in SLD, each operating on different scales. On this point, little is said in Spoelman & Verspoor (2010), but it is not hard to imagine the notion of competition as one key control parameter in the system. Determining which control parameters a system is particularly sensitive to is a key task of CDST research as it may be the most productive way of finding the “motors of change” (Larsen–Freeman & Cameron, 2008a, p. 70) for intervention. Research may indicate, for instance, that a confluence of personal relevance, task requirements, and motivation act as control parameters for L2 task performance. In this way, system intervention may entail setting the conditions and shaping the path of emergent outcomes from a good enough design of the pertinent control parameters (Byrne & Uprichard, 2012). PUTTING COMPLEXITY INTO PRACTICE CDST research is, perhaps, at too early a stage in applied linguistics for us to conduct a state-ofthe art review of available methods. Nevertheless, one ongoing purpose of CDST research will be to develop a methodological repertoire adequate to the social phenomena we are concerned with in applied linguistics. Taking our cues from other social and human disciplines, where a wide range of methods for complexity research are already in use, here we address the issue of what practical implementation of the dynamic ensemble would entail. First and foremost, because it is a meta-theory, CDST does not dictate the use of unique methods of data elicitation and analysis, nor does it exclude existing research methods so long as they are fundamentally compatible with the principles of complexity (Byrne, 2011). CDST is grounded in the phenomenological reality of the social world and calls for approaches that emerge from the needs of inquiry (Morin, 2008), which we believe complements the recent pivot toward a more transdisciplinary, problem-focused orientation to research methodology (King & Mackey, 2016). CDST has undeniable methodological implications regarding, for instance, what may count as a more valid representation of causal mechanisms. On the grounds of incompatibility, this may rule

The Modern Language Journal 100 (2016) out certain techniques (e.g., pre/posttest designs, linear analyses) as they do not shed adequate light on the dynamic processes CDST research is interested in. In a sense, however, there are no such things as “methods of complexity” because CDST encourages innovation and diversification in understanding complex social phenomena (Manson, 2001). The pragmatism central to complexity research dictates that, instead of searching for problems to which to apply our tools, we should be looking for tools suitable for solving the problems we come up against—with the added caveat that fundamental criteria of sound research practices still hold (see e.g., Banaji & Crowder, 1989). There is no shortage of suitable research methods for CDST (e.g., narrative methods, multilevel modeling procedures, event history analysis, grounded theory, Bayesian analysis, cluster analytical methods, the experience sampling method, nonlinear time-series analysis). We would suggest, therefore, that CDST research might gain greater traction in applied linguistics if scholars recognized that CDST encourages repurposing existing methodological toolkits— both qualitative and quantitative—to ensure they are congruent with the complexity framework (van Geert, 2008). While a book-length treatment would be necessary to provide adequate detail about potential methods for SLD research, below we briefly introduce five case-based methods—widely used in the social sciences—which are some of the most accessible designs for building off of the nine considerations in our dynamic ensemble for research. We also suggest questions about SLD which could be addressed empirically using each method. These methods incorporate the CDST logic of causal explanation, generalization, and hypothesis confirmation in which system outcomes are the contingent products of multiple complex adaptive mechanisms and causal analysis must explain why the course of development ultimately led to the outcome in question rather than to alternative ones (Morrison, 2012). The fact that our complex social world presents a set of phenomenological outcomes and self-organized processes, which are found in recurring instances and guises, has real significance for this project (De Wolf & Holvoet, 2005). Qualitative Comparative Analysis Qualitative comparative analysis (QCA) is a set-theoretic method appropriate for either theory building or theory testing (Rihoux & Lobe, 2009). Better suited to the macro or meso level of

751

Phil Hiver and Ali H. Al-Hoorie granularity, QCA fits well into a multi-method design: Data is often both qualitative and quantitative, and the analysis proper is designed to arrive at a complex model of an emergent outcome being investigated. Similar to retrodictive qualitative modeling, QCA begins with the selection of one or more cases (i.e., complex systems) and definition of a particular outcome of interest before investigating the causal conditions that led those systems to that emergent outcome (Ragin, 2014). Contextual factors are central to QCA and are coded as fuzzy or crisp variables, as are the system components, interactions, and parameters which empirically factor in to the outcome of interest. However, with its focus on a relevant outcome for particular systems and a complex causal explanation of that outcome, QCA foregrounds emergent states and says little about systemic networks. QCA might be used to explore how dynamic learner factors interact with contextual factors and contribute to various outcomes and stages in the process of individual L2 development, and to uncover how cognitive and affective processes which unfold over the long term influence decisions and actions by L2 users on shorter timescales. Rihoux and Lobe (2009) provide an intuitive overview of QCA’s purposes and procedures, and we know of at least one study in our field that has used QCA (Hiver, 2015b). Process-Tracing Process-tracing is a within-case (i.e., single system) method used to explain complex causal mechanisms at a micro level of granularity. By analyzing “evidence on processes, sequences, and conjunctures of events” (Bennett & Checkel, 2015, p. 7), termed diagnostic evidence, processtracing attempts to identify how and why an intervening causal chain led to an emergent outcome. A versatile method, process-tracing may be quantitative, quasi-quantitative, or entirely qualitative in design. This method has parallels in historiography, and unlike QCA—which also models emergent outcomes—process-tracing relies on microscopic tracing of a dynamic trajectory and examines evidence for competing explanations through a sequence of inferential tests. Process-tracing can progress (a) backward into the system’s context and history, (b) forward into the dynamic mechanisms of change, (c) upward into the network of systems that anchor and interact with the system being examined, and (d) downward into the system’s components, interactions, and parameters to explain complex causal

processes for a given outcome (Beach & Pedersen, 2013). Within instructed L2 settings, process tracing might be used to analyze how learning events involving teachers and learners add up to coherent wholes of activity over periods ranging from minutes and hours to days and months, and to trace how multilingual learners construct a sense of identity through learning and use of their additional languages. In the introduction to their edited volume, Bennett and Checkel (2015) provide a comprehensive list of techniques and best practices in process tracing; however, we are not aware of any study in the L2 field that has applied this method. Concept Mapping Concept mapping is a theoretically grounded diagrammatic method used for complex-systemic problem solving, that is, investigating a system’s emergent states and dynamic processes of change, and in turn doing something about them. Using thematic clusters, concept mapping produces a spatio-temporal representation of a system and the structural links between clusters. It thus draws on Kurt Lewin’s often-cited maxim that “there is nothing so practical as a good theory” (Kane & Trochim, 2007), and is particularly suited for building large-scale, concrete models of systems. More often qualitative than not, concept mapping is firmly focused on the system level of complex phenomena. Though its primary objective is to serve as the basis for action, it incorporates fewer considerations from context or systemic networks than other methods discussed here. With an aggregate visual diagram of a system and its mechanisms as the point of departure, concept mapping aims to strategically implement innovations that optimize system functioning in order to solve real-world problems (Moon et al., 2011). Given the nonlinear relationship between instruction and L2 learning, concept mapping might be used to examine the precise roles that adaptive behaviors in L2 task performance play in the self-organized development of a learner’s complexity, accuracy, and fluency, and to study the motors of change in developing dynamic, fluid, and socially situated language competences. In their accessible guide, Kane and Trochim (2007) demonstrate the uses of concept mapping which range from organizing work flows and solving problems, to developing models of processes and synthesizing knowledge. We are unaware of any study in the SLD field designed using concept mapping.

752 Social Network Methods Social network methods, similar to concept mapping, use visual matrices (e.g., information maps, sociograms) and add their own sophisticated computational analyses that draw on graph theory. These methods are equally well-suited to exploratory research designs or to hypothesis testing (Kadushin, 2012). Unlike concept mapping, however, social network methods shift their focus more broadly to modeling and analyzing the environment (i.e., systemic networks) that systems exist in and the relational patterns and implications this interwoven network has on systems (Mercer, 2015). Although social network methods’ primary emphasis is on the graphic architecture of the systemic networks, multiple levels of analysis are possible, and these methods do place importance on what this structure can reveal about the behavior and functioning of the systems within it (Carolan, 2014). For these reasons, systemic networks can also be used to model dynamic processes and to shed light on the micro level considerations we have included in our dynamic ensemble. We know of at least one recent study in our field (i.e., Gallagher & Robins, 2015) employing these methods. Social network methods might also be used to examine broad questions of how the changing priorities, populations, and problems of L2 contexts influence the larger educational system’s agendas, policies, and practices. Agent-Based Modeling Agent-based modeling is a method for building, from the ground up, working models of complex systems and for simulating their dynamic processes and emergent outcomes (Castellani & Hafferty, 2009). This type of modeling is commonly used to investigate the empirically optimal solutions to system behavior and outcomes, especially when actual manipulation of a system or its agents for this purpose presents practical challenges (e.g., due to issues of access or scale). Agent-based modeling is a process often used for solution finding in complex scenarios. Among other steps, it involves formulating a model that is a transformation of an empirical situation, specifying quantitative values and indicators that connect a model and the target system (i.e., parameterization), then assessing, calibrating and scaling a model based on data collection (e.g., using observations, surveys, interviews, or other data elicitation tasks). By combining heterogeneous sources and levels of data, its priority is on discovering ways to represent the interacting

The Modern Language Journal 100 (2016) components and experimentally display the emergent properties and dynamic reactions of systems (Siegfried, 2014). Agent-based modeling might be used to study how delayed effects and instances of regression contribute to an L2 learner’s trajectory of development, and to model how conceptual information is transformed, filtered, re-organized, and added to throughout the L2 development process. In their review, Macy and Willer (2002) provide a clear introduction of the assumptions and principles of model design. To our knowledge no L2 research has yet used this method. SUMMARY AND CONCLUSIONS In this article, we began by highlighting the need for a more explicit consideration of how the theoretical framework of complexity can be adopted in L2 empirical research designs. We outlined the conceptual tools of complexity in order to clear up some of the apprehensions surrounding CDST research. The central purpose of this article was to present a blueprint of consensus-forming considerations for empirical CDST research—a model we termed the dynamic ensemble. Rather than an intimidating list of desiderata our proposed categorization should be seen as just one of many possible ways to structure the field. Using an exemplary study we have illustrated the ways in which these nine considerations might inform empirical research in practice. We ended this article by considering what practical implementation of the dynamic ensemble would entail, and introduced several accessible research methods for building off of the nine considerations in this template. Using this dynamic ensemble, the core objectives of CDST research in applied linguistics are to (a) represent and understand specific complex systems at various scales of description, (b) identify and understand the dynamic patterns of change, emergent system outcomes, and behavior in the environment, (c) trace, understand, and, where possible, model the complex mechanisms and processes by which these patterns arise, and (d) capture, understand, and apply the relevant parameters for influencing the behavior of the systems. Perhaps thus far, CDST has not easily lent itself to telling compelling research stories, but many SLD scholars will undoubtedly recognize its untapped potential to provide more accurate answers and solutions. Because there are only a certain number of patterns that one will see or come up against, we firmly believe that the varied complex dynamic phenomena observable in the

Phil Hiver and Ali H. Al-Hoorie social and human world are well within the reach of scientific inquiry. There may in fact be research stories about SLD that can only be told properly using complexity. If current levels of engagement with CDST in L2 research are any indication, complexity will continue to grow as a significant force in the field for some time to come. We are certain that the best research and the most insightful findings in SLD are still ahead of us, and welcome greater engagement with CDST as we attempt to move research to a new level of complexity, rigor, and usefulness.

NOTES 1 Contemporary branches of the three parent fields general systems theory, cybernetics, and dynamic(al) systems theory are subsumed under the umbrella term complexity theory (alternatively complexity science). However, as one reviewer recommended, SLD scholars have increasingly adopted the abbreviation “CDST” to maximize the inclusivity of mutually intelligible and complementary foci (e.g., emergentism, dynamic systems theory, chaos theory). 2 We adopt this term because complex systems are by definition dynamic, whereas dynamic systems are not inherently complex. The term complex dynamic system, technically, is redundant. Complex adaptive systems, on the other hand, are the particular class of complex systems that learn adaptively.

REFERENCES Al–Hoorie, A. H. (2015). Human agency: Does the beach ball have free will? In Z. Dörnyei, P. MacIntyre, & A. Henry (Eds.), Motivational dynamics in language learning (pp. 55–72). Bristol, UK: Multilingual Matters. Baba, K., & Nitta, R. (2014). Phase transitions in development of writing fluency from a complex dynamic systems perspective. Language Learning, 64, 1–35. Baird, R., Baker, W., & Kitazawa, M. (2014). The complexity of ELF. Journal of English as a Lingua Franca, 3, 171–196. Banaji, M., & Crowder, R. (1989). The bankruptcy of everyday memory. American Psychologist, 44, 1185– 1193. Beach, D., & Pedersen, R. B. (2013). Process tracing methods: Foundations and guidelines. Ann Arbor, MI: The University of Michigan Press. Bennett, A., & Checkel, J. (2015). Process tracing: From metaphor to analytic tool. Cambridge: Cambridge University Press. Blommaert, J. (2014). From mobility to complexity in sociolinguistic theory and method. Tilburg Papers

753 in Culture Studies, 103. Accessed 26 April 2016 at http://www.tilburguniversity.edu/upload/5ff19e 97-9abc-45d0-8773-d2d8b0a9b0f8_TPCS_103_ Blommaert.pdf. Byrne, D. (2011). Applying social science. Bristol, UK: Policy Press. Byrne, D., & Callaghan, G. (2014). Complexity theory and the social sciences: The state of the art. New York: Routledge/Taylor & Francis. Byrne, D., & Ragin, C. (Eds.). (2009). The SAGE handbook of case-based methods. Thousand Oaks, CA: SAGE. Byrne, D., & Uprichard, E. (2012). Useful complex causality. In E. Kincaid (Ed.), The Oxford handbook of philosophy of social science (pp. 109–129). Oxford: Oxford University Press. Carolan, B. (2014). Social network analysis and education: Theory, methods and applications. Thousand Oaks, CA: SAGE. Castellani, B., & Hafferty, F. (2009). Sociology and complexity science: A new field of inquiry. Berlin: Springer. Chow, J. Y., Davids, K., Button, C., & Renshaw, I. (2016). Nonlinear pedagogy in skill acquisition: An introduction. New York: Routledge/Taylor & Francis. Cilliers, P. (2000). Knowledge, complexity, and understanding. Emergence, 2(4), 7–13. Cilliers, P. (2001). Boundaries, hierarchies, and networks in complex systems. International Journal of Innovation Management, 5, 135–147. Cilliers, P. (2005). Complexity, deconstruction and relativism. Theory, Culture and Society, 22, 255–267. Davis, B., & Sumara, D. (2006). Complexity and education: Inquiries into learning, teaching and research. Mahwah, NJ: Lawrence Erlbaum. de Bot, K. (2012). Rethinking multilingual processing: From a static to dynamic approach. In J. C. Amaro, S. Flynn, & J. Rothman (Eds.), Third language acquisition in adulthood (pp. 70–94). Philadelphia/Amsterdam: John Benjamins. de Bot, K. (2015). Rates of change: Timescales in second language development. In Z. Dörnyei, P. MacIntyre, & A. Henry (Eds.), Motivational dynamics in language learning (pp. 29–37). Bristol, UK: Multilingual Matters. de Bot, K., & Larsen–Freeman, D. (2011). Researching second language development from a dynamic systems theory perspective. In M. Verspoor, K. de Bot, & W. Lowie (Eds.), A dynamic approach to second language development (pp. 5–23). Philadelphia/Amsterdam: John Benjamins. de Bot, K., Lowie, W., Thorne, S., & Verspoor, M. (2013). Dynamic systems theory as a comprehensive theory of second language development. In M. del, P. G. Mayo, M. J. G. Mangado, & M. M. Adrián (Eds.), Contemporary approaches to second language acquisition (pp. 199–220). Philadelphia/Amsterdam: John Benjamins. De Wolf, T., & Holvoet, T. (2005). Emergence versus self-organisation: Different concepts but promising when combined. In S. Brueckner, G. Di Marzo Serugendo, A. Karageorgos, & R. Nagpal (Eds.),

754 Engineering self-organising systems: Methodologies and applications (pp. 1–15). Berlin: Springer. Dörnyei, Z. (2014). Researching complex dynamic systems: ‘Retrodictive qualitative modelling’ in the language classroom. Language Teaching, 47, 80–91. Dörnyei, Z., MacIntyre, P., & Henry, A. (Eds.). (2015). Motivational dynamics in language learning. Bristol, UK: Multilingual Matters. Dörnyei, Z., & Malderez, A. (1997). Group dynamics and foreign language teaching. System, 25, 65–81. Ellis, N. C. (1998). Emergentism, connectionism and language learning. Language Learning. 48, 631– 664. Ellis, N. C., & Larsen–Freeman, D. (2006). Language emergence: Implications for applied linguistics. Introduction to the special issue. Applied Linguistics, 27, 558–589. Ellis, N. C., & Larsen–Freeman, D. (2009). Constructing a second language: Analyses and computational simulations of the emergence of linguistic constructions from usage. Language Learning, 59(s1), 90–120. Elman, J. (2003). It’s about time. Developmental Science, 6, 430–433. Eve, R., Horsfall, S., & Lee, M. (Eds.). (1997). Chaos, complexity, and sociology: Myths, models, and theories. Thousand Oaks, CA: SAGE. Gallagher, H. C., & Robins, G. (2015). Network statistical models for language learning contexts: Exponential random graph models and willingness to communicate. Language Learning, 65, 929– 962. Gaustello, S., & Liebovitch, L. (2009). Introduction to nonlinear dynamics and complexity. In S. Gaustello, M. Koopmans, & D. Pincus (Eds.), Chaos and complexity in psychology (pp. 1–40). Cambridge: Cambridge University Press. Gintis, H. (2009). The bounds of reason: Game theory and the unification of the behavioral sciences. Princeton, NJ: Princeton University Press. Gregersen, T., MacIntyre, P., & Meza, M. (2014). The motion of emotion: Idiodynamic case studies of learners’ foreign language anxiety. Modern Language Journal, 98, 574–588. Haken, H. (1997). Visions of synergetics. Journal of the Franklin Institute, 334B, 759–792. Haken, H. (2009). Synergetics: Basic concepts. In R. Meyers (Ed.), Encyclopedia of complexity and systems science (pp. 8926–8946). New York: Springer. Henry, A. (2015). The dynamics of possible selves. In Z. Dörnyei, P. MacIntyre, & A. Henry (Eds.), Motivational dynamics in language learning (pp. 83–94). Bristol, UK: Multilingual Matters. Hiver, P. (2015a). Attractor states. In Z. Dörnyei, P. MacIntyre, & A. Henry (Eds.), Motivational dynamics in language learning (pp. 20–28). Bristol, UK: Multilingual Matters. Hiver, P. (2015b). Once burned, twice shy: The dynamic development of system immunity in language teachers. In Z. Dörnyei, P. MacIntyre, & A. Henry (Eds.), Motivational dynamics in language

The Modern Language Journal 100 (2016) learning (pp. 214–237). Bristol, UK: Multilingual Matters. Holland, J. H. (2012). Signals and boundaries: Building blocks for complex adaptive systems. Cambridge, MA: MIT Press. Hulstijn, J. H., Young, R. F., Ortega, L., Bigelow, M., DeKeyser, R., Ellis, N. C., Lantolf, J. P., Mackey, A., & Talmy, S. (2014). Bridging the gap. Studies in Second Language Acquisition, 36, 361–421. Hult, F. (2010). The complexity turn in educational linguistics. Language, Culture and Curriculum, 23, 173– 177. Jessner, U. (2008). A DST model of multilingualism and the role of metalinguistic awareness. Modern Language Journal, 92, 270–283. Jörg, T. (2011). New thinking in complexity for the social sciences and humanities: A generative, transdisciplinary approach. Dordrecht, the Netherlands: Springer. Kadushin, C. (2012). Understanding social networks: Theories, concepts, and findings. Oxford: Oxford University Press. Kane, M., & Trochim, W. (2007). Concept mapping for planning and evaluation. Thousand Oaks, CA: SAGE. Kiel, L. D., & Elliot, E. (Eds.). (1996). Chaos theory in the social sciences: Foundations and applications. Ann Arbor, MI: The University of Michigan Press. King, K., & Mackey, A. (2016). Research methodology in second language studies: Trends, concerns, and new directions. Modern Language Journal, 100, Supplement 2016, 209–227. Klein, J. (2004). Interdisciplinarity and complexity: An evolving relationship. Emergence. Complexity & Organization, 6(1/2), 2–10. Kuhn, L. (2008). Complexity and educational research: A critical reflection. In M. Mason (Ed.), Complexity theory and the philosophy of education (pp. 169–180). Chichester, UK: Wiley-Blackwell. Larsen–Freeman, D. (1997). Chaos/complexity science and second language acquisition. Applied Linguistics, 18, 141–165. Larsen–Freeman, D. (2006). The emergence of complexity, fluency, and accuracy in the oral and written production of five Chinese learners of English. Applied Linguistics, 27, 590–619. Larsen–Freeman, D. (2010). Not so fast: A discussion of L2 morpheme processing and acquisition. Language Learning, 60, 221–230. Larsen–Freeman, D. (2012a). Complex, dynamic systems: A new transdisciplinary theme for applied linguistics? Language Teaching, 45, 202–214. Larsen–Freeman, D. (2012b). The emancipation of the language learner. Studies in Second Language Learning and Teaching, 2, 297–309. Larsen–Freeman, D. (2013). Complexity theory: A new way to think. Revista Brasileira de Linguistica Aplicada, 13, 369–373. Larsen–Freeman, D. (2015a). Saying what we mean: Making a case for “language acquisition” to become “language development.” Language Teaching, 48, 491–505.

Phil Hiver and Ali H. Al-Hoorie Larsen–Freeman, D. (2015b). Ten “lessons” from complex dynamic systems theory: What is on offer. In Z. Dörnyei, P. MacIntyre, & A. Henry (Eds.), Motivational dynamics in language learning (pp. 11–19). Bristol, UK: Multilingual Matters. Larsen–Freeman, D. (2017). Complexity theory: The lessons continue. In L. Ortega & Z. Han (Eds.), Complexity theory and language development: In celebration of Diane Larsen–Freeman. Philadephia/Amsterdam: John Benjamins. Larsen–Freeman, D., & Cameron, L. (2008a). Complex systems and applied linguistics. Oxford: Oxford University Press. Larsen–Freeman, D., & Cameron, L. (2008b). Research methodology on language development from a complex systems perspective. Modern Language Journal, 92, 200–213. Lemke, J. (2000). Across the scales of time: Artifacts, activities, and meanings in ecosocial systems. Mind, Culture, and Activity, 7, 273–290. Lowie, W., & Verspoor, M. (2015). Variability and variation in second language acquisition orders: A dynamic reevaluation. Language Learning, 65, 63–88. MacIntyre, P., & Legatto, 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., Dörnyei, Z., & Henry, A. (2015). Hot enough to be cool: The promise of dynamic systems research. In Z. Dörnyei, P. MacIntyre, & A. Henry (Eds.), Motivational dynamics in language learning (pp. 419–429). Bristol, UK: Multilingual Matters. MacWhinney, B. (1998). Models of the emergence of language. Annual Review of Psychology, 49, 199–227. Macy, M., & Willer, R. (2002). From factors to actors: Computational sociology and agent-based modeling. Annual Review of Sociology, 28, 143–166. Meara, P. (1997). Towards a new approach to modelling vocabulary acquisition. In N. Schmitt & M. McCarthy (Eds.), Vocabulary: Description, acquisition and pedagogy (pp. 109–121). Cambridge: Cambridge University Press. Manson, S. (2001). Simplifying complexity: A review of complexity theory. Geoforum, 32, 405–414. Mercer, S. (2013). Towards a complexity-informed pedagogy for language learning. Revista Brasileira de Linguistica Aplicada, 13, 375–398. Mercer, S. (2014). The self from a complexity perspective. In S. Mercer & M. Williams (Eds.), Multiple perspectives on the self in SLA (pp. 160–176). Bristol, UK: Multilingual Matters. Mercer, S. (2015). Social network analysis and complex dynamic systems. In Z. Dörnyei, P. MacIntyre, & A. Henry (Eds.), Motivational dynamics in language learning (pp. 73–82). Bristol, UK: Multilingual Matters. Mercer, S. (2016). The contexts within me: L2 self as a complex dynamic system. In J. King (Ed.), The dynamic interplay between context and the language learner (pp. 11–28). Basingstoke, UK: Palgrave: Macmillan.

755 Michell, J. (2008). Is psychometrics pathological science? Measurement, 6, 7–24. Moon, B., Hoffman, R., Novak, J., & Cañas, A. (Eds.). (2011). Applied concept mapping: Capturing, analyzing, and organizing knowledge. Boca Raton, FL: CRC Press. Morin, E. (1992). From the concept of system to the paradigm of complexity. Journal of Social and Evolutionary Systems, 15, 371–385. Morin, E. (2008). On complexity. Cresskill, NJ: Hampton Press. Morrison, K. (2012). Searching for causality in the wrong places. International Journal of Social Research Methodology, 15, 15–30. Norman, D. (2011). Living with complexity. Cambridge, MA: The MIT Press. Ortega, L. (2012). Epistemological diversity and moral ends of research in instructed SLA. Language Teaching Research, 16, 206–226. Ortega, L. (2013). SLA for the 21st century: Disciplinary progress, transdisciplinary relevance, and the bi/multilingual turn. Language Learning, 63, 1–24. Overton, W. (2007). A coherent metatheory for dynamic systems: Relational organicism-contextualism. Human Development, 50, 154–159. Overton, W. (2013). A new paradigm for developmental science: Relationism and relationaldevelopmental systems. Applied Developmental Science, 17, 94–107. Prigogine, I. (1980). From being to becoming: Time and complexity in the physical sciences. New York: W. H. Freeman. Radford, M. (2008). Prediction, control and the challenge to complexity. Oxford Review of Education, 34, 505–520. Ragin, C. (1997). Turning the tables: How case-oriented methods challenge variable-oriented methods. Comparative Social Research, 16, 27–42. Ragin, C. (2014). The comparative method: Moving beyond qualitative and quantitative strategies (2nd ed.). Oakland, CA: University of California Press. Ragin, C., & Becker, H. (1992). What is a case? Exploring the foundations of social inquiry. Cambridge: Cambridge University Press. Rihoux, B., & Lobe, B. (2009). The case for qualitative comparative analysis (QCA): Adding leverage for thick cross-case comparison. In D. Byrne & C. Ragin (Eds.), The SAGE handbook of casebased methods (pp. 222–242). Thousand Oaks, CA: SAGE. Seedhouse, P. (2010). Locusts, snowflakes and recasts: Complexity theory and spoken interaction. Classroom Discourse, 1, 4–24. Siegfried, R. (2014). Modeling and simulation of complex systems: A framework for efficient agent-based modeling and simulation. Berlin: Springer. Spoelman, M., & Verspoor, M. (2010). Dynamic patterns in development of accuracy and complexity: A longitudinal case study in the acquisition of Finnish. Applied Linguistics, 31, 532–553.

756 The Douglas Fir Group. (2016). A transdisciplinary framework for SLA in a multilingual world. Modern Language Journal, 100, Supplement 2016, 19–47. Ulrich, W. (2001). The quest for competence in systemic research and practice. Systems Research and Behavioral Science, 18, 3–28. Uprichard, E. (2013). Sampling: Bridging probability and nonprobability designs. International Journal of Social Research Methodology, 16, 1–11. Ushioda, E. (2015). Context and complex dynamic systems theory. In Z. Dörnyei, P. MacIntyre, & A. Henry (Eds.), Motivational dynamics in language learning (pp. 47–54). Bristol, UK: Multilingual Matters. Vallacher, R., van Geert, P., & Nowak, A. (2015). The intrinsic dynamics of psychological process. Current Directions in Psychological Science, 24, 58–64.

The Modern Language Journal 100 (2016) van Geert, P. (2008). The dynamic systems approach in the study of L1 and L2 acquisition: An introduction. Modern Language Journal, 92, 179–199. van Lier, L. (1988). The classroom and the language learner. London: Longman. Verspoor, M., de Bot, K., & Lowie, W. (Eds.). (2011). A dynamic approach to second language development. Philadelphia/Amsterdam: John Benjamins. Verspoor, M., Lowie, W., & van Dijk, M. (2008). Variability in second language development from a dynamic systems perspective. Modern Language Journal, 92, 214–231. Verspoor, M., Schmid, M., & Xu, X. (2012). A dynamic usage based perspective on L2 writing. Journal of Second Language Writing, 21, 239–263. Walby, S. (2007). Complexity theory, systems theory and multiple intersecting social inequalities. Philosophy of the Social Sciences, 37, 449–470.

Suggest Documents