ArticlePDF Available

Abstract

A quarter of a century has passed since complex dynamic systems theory was proposed as an alternative paradigm to rethink and reexamine some of the main questions and phenomena in applied linguistics and language learning. In this article, we report a scoping review of the heterogenous body of research adopting this framework. We analyzed 158 reports satisfying our inclusion criteria (89 journal articles and 69 dissertations) for methodological characteristics and substantive contributions. We first highlight methodological trends in the report pool using a framework for dynamic method integration at the levels of study aim, unit of analysis, and choice of method. We then survey the main substantive contribution this body of research has made to the field. Finally, examination of study quality in these reports revealed a number of potential areas of improvement. We synthesize these insights in what we call the "nine tenets" of complex dynamic systems theory research, which we hope will help enhance the methodological rigor and the substantive contribution of future research.
Research Article
COMPLEX DYNAMIC SYSTEMS THEORY IN LANGUAGE
LEARNING
A SCOPING REVIEW OF 25 YEARS OF RESEARCH
Phil Hiver *
Florida State University, USA
Ali H. Al-Hoorie
Royal Commission for Jubail and Yanbu, Saudi Arabia
Reid Evans
University of Massachusetts Medical School, USA
Abstract
A quarter of a century has passed since complex dynamic systems theory was proposed as an
alternative paradigm to rethink and reexamine some of the main questions and phenomena in applied
linguistics and languagelearning. Inthis article, we report a scoping review of the heterogenous body
of research adopting this framework. We analyzed 158 reports satisfying our inclusion criteria
(89 journal articles and 69 dissertations) for methodological characteristics and substantive contri-
butions. We rst highlight methodological trends in the report pool using a framework for dynamic
method integration at the levels of study aim, unit of analysis, and choice of method. We then survey
the main substantive contribution this body of research has made to the eld. Finally, examination of
study quality in these reports revealed a number of potential areas of improvement. We synthesize
these insights in what we call the nine tenetsof complex dynamic systems theory research, which we
hope will help enhance the methodological rigor and the substantive contribution of future research.
INTRODUCTION
All theories, if they are to avoid becoming passing academic fads or bandwagons, must
contribute something of substance that is new and worthwhilesomething that pushes
© The Author(s), 2021. Published by Cambridge University Press. This is an Open Access article, distributed
under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0),
which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
We would like to thank Alyssa Vuono, Janice Wu, and Hyejin An for their assistance with coding the report pool.
* Correspondence concerning this article should be addressed to Florida State University, School of Teacher
Education, College of Education, 1114 W. Call St., G128 Stone Building, Tallahassee, FL, 32306. E-mail:
phiver@fsu.edu
This article has been updated since its original publication. See https://doi.org/10.1017/S0272263122000262.
Studies in Second Language Acquisition,44(2022), 913941
doi:10.1017/S0272263121000553
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
the eld forward. Nearly three decades have passed since complex dynamic systems
theory (CDST) was rst introduced into the eld of language learning (Larsen-Freeman,
1994), and since then CDST perspectives and approaches have permeated many areas of
applied linguistics research. The uptake of CDST in applied linguistics research has
continued to accelerate, pushing further and faster even than in related elds such as
education and theoretical linguistics (Koopmans, 2020; Kretzschmar, 2015). As recent
work (e.g., Larsen-Freeman, 2017) synthesizing current strands of applied linguistics that
have been informed by CDST shows, CDST has made important contributions to
language development/acquisition (Lowie et al., 2010; Verspoor et al., 2008), language
attrition (Schmid et al., 2013), language ecology (Cowley, 2011; Kramsch & Whiteside,
2008), language evolution (Ke & Holland, 2006; Mufwene et al., 2017), language policy
and planning (Bastardas-Boada, 2013; Larsen-Freeman, 2018), language pedagogy (Han,
2020; Levine, 2020), bilingualism and multilingualism (Herdina & Jessner, 2002),
sociolinguistics (Blommaert, 2014), educational linguistics (Hult, 2010), and communi-
cation studies (Massip-Bonet et al., 2019), among other areas of applied linguistics.
Considering this mainstream interest in CDST, it seems that it is not just appropriate but
also necessary to assess this body of empirical work and evaluate the strength of its
contribution to the eld. Systematic and scoping reviews are uniquely positioned to afford
a new vantage point on an area of research, and assessing the nature and quality of
previous work has the potential to shape the future of research and practice (Alexander,
2020). Scoping reviews in particular are relevant when an area of research has not yet been
extensively reviewed or when it is of a complex or heterogeneous nature (Pham et al.,
2014)arguably the case with CDST research. Scoping reviews share a number of
procedural characteristics with systematic reviews, but where these two approaches to
synthesis diverge is in their purposes and aims. The purpose of a systematic review is to
identify the best available research on a specic question or a precise topic of research, and
this often leads to answers of the appropriateness or effectiveness of some practice (Munn
et al., 2018). Scoping reviews, however, look at what a eld has done and how. Their aim
is to examine how research is conducted in a certain eld and provide an overview of the
types of available evidence from that research (Arksey & OMalley, 2005). As a result,
scoping reviews generally evaluate patterns of knowledge and research methods from a
greater range of study designs (Levac et al., 2010).
In the present scoping review and methodological synthesis of 25 years of CDST
research, we had several objectives. In light of the growing methodological guidance
available, our primary aim was to look back at the methodological characteristics of all
previous empirical CDST studies in the eld to note trends and tendencies in designs and
analytical choices. By dening the shape of existing research designs, the eld can take
stock and chart a path forward. In addition to methodological characteristics, we were also
interested in the substantive contributions this sizeable body of CDST research has made
to the eld, and what evidence it has provided for the language learning research
enterprise. Given the readily apparent heterogeneity of research topics under the rubric
of CDST research in language learning, we wondered what conclusions this empirical
work allows us to draw and whether such a review could speak directly to broader issues
and shared concerns in the eld. Finally, we were interested in the rigor of this body of
empirical work. Although CDST research has made many advances, we intended to
explore whether this orchestrated search of the literature would reveal potential areas for
914 Phil Hiver, Ali H. Al-Hoorie, and Reid Evans
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
enhancing the quality of this body of research. We, thus, sought to identify future
directions for CDST research that will help it continue to push the eld forward with
more coherent evidence and sharper insights.
Study quality has become central to many subdomains of SLD research (Gass et al.,
2021). For instance, many syntheses have demonstrated that design tendencies related to
measurement and sampling in the eld leave much to be desired (Brown et al., 2018;
Nicklin & Plonsky, 2020; Vitta & Al-Hoorie, 2021). Others have highlighted the need for
greater transparency in checking and reporting assumptions (Hu & Plonsky, 2021), and
increased rigor in data analytical strategies and reporting results (Al-Hoorie & Vitta,
2019; Larson-Hall & Plonsky, 2015; Marsden et al., 2018; Paquot & Plonsky, 2017;
Plonsky, 2013,2014). In the context of synthetic work such as this, study quality can refer
to quality of the implementation of the methods or to the quality of inferences made from
the methods (see also Gass et al., 2021), and as commonly observed, sound implemen-
tation of methods is orthogonal to whether those methods support a given inference. To
our knowledge, this is the rst methodologically oriented review of CDST research in
language learning (but see Larsen-Freeman, 2017 for a detailed substantive synthesis).
We are also not aware of any methodological reviews or syntheses of CDST research even
in the wider social sciences or educational research literature. Thus, a critical appraisal of
study quality can help to shed light on the transparency of this research, the relevance of
the research targets and questions under investigation, and the appropriateness of methods
of data analysis and presentation.
Of course, critical appraisal of research methods is not the pursuit of some form of
elusive and idealized methodological perfection. Evaluating the methods adopted by a
body of research serves a much more nuanced and meaningful purpose: to assess whether
that body of work is evidentially adequate(Petticrew & Roberts, 2006). When consid-
ering methodological aspects and study quality, we followed recommendations to exam-
ine broader and more general methodological issues rst as these can inform later reviews
that assess more ne-grained aspects of study quality (Siddaway et al., 2019). In this
scoping review we aimed to survey the methods employed by CDST researchers broadly,
looking at generic characteristics such as research objectives, design and methodological
orientation, sampling characteristics, data elicitation measures, and analytical strategies.
We turn now to outlining the topic, scope, and rationale for the present review.
WHAT IS CDST RESEARCH?
CDST is a meta-theory that provides an ontological position (i.e., principles of reality) for
understanding language, language use, and language development in complex and
dynamic terms (Hulstijn, 2020). It also captures epistemological ideas (i.e., principles
of knowing) that aid scientic thinking and theorizing. In the eld of language develop-
ment, CDST underpins and contextualizes object theories consistent with these principles
(Larsen-Freeman & Cameron, 2008), and these object theories address proximate ques-
tions about processes and outcomes of development. With regard to language, CDST
proposes that language is a complex adaptive system, exhibiting both stability and
dynamic change (Ellis & Larsen-Freeman, 2009). Language use is an iterative process
of coadaptation in which language users adapt to the context and other interlocutors to
realize the semiotic potential of language (Han, 2019). Language development is a
Complex Dynamic Systems Theory in Language Learning 915
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
nonlinear, emergent process that draws on local-to-global processes of construction and
global-to-local processes of constraint (de Bot, 2008). Whereas object theories
(i.e., theories of language, language use, and language development/learning) are provi-
sional, and their predictions must constantly be falsied and evaluated against observa-
tions of new evidence, the CDST meta-theory is broader in scope and relates to notions of
what phenomena, questions, and aspects of inquiry should be investigated and why they
merit research (Hulstijn, 2020; Overton, 2007).
In an applied eld like ours, the entry point to CDST research is likely to be
methodological and phenomenological, rather than at the more abstract level of theory
(Larsen-Freeman, 2016b). That is, studies may set out to investigate constructs or
questions pertaining to complex connections and dynamic processes of change, but are
likely less concerned with disentangling the ontology and epistemology that underlies that
mode of thinking (see also Ushioda, 2021). Research informed by CDST is different from
other, more conventional research in two main ways: the basic assumptions that underlie it
and the designs and methods that follow from those assumptions (Verspoor et al., 2011,
p. 123). All research methods and paradigms
1
have a number of inherent assumptions,
some of which are unstated or implicit in the techniques of data elicitation and analysis.
CDST research takes a systems view as its point of departure (see e.g., Larsen-Freeman,
2015). CDST posits that the reality of the human and social world is one in which, rst,
everything counts and everything is connected (i.e., the relational principle) and second,
everything changes (i.e., the adaptive principle) (Overton & Lerner, 2014). CDST
research reconceptualizes the core of language, language use, and language development
as systems or systemic phenomena grounded in a context-dependent and dynamic view of
development. This reorientation challenges many of the elds existing assumptions and
suggests new approaches to inquiry (Hiver et al., 2021).
There are multiple ways of approaching a topical area in our eld. Primarily, the study
of complex systems entails a focus on processes of change, and one way of doing so is
through dynamics-dominant research using time-intensive methods (see also Van Orden
et al., 2003 for a related framing). The question of how complex systems adapt to their
environment to maintain their functioning over time is in fact relevant to nearly every part
of applied linguistics (Larsen-Freeman & Cameron, 2008). Complex macrobehaviors,
dynamic microinteractions within a system, and the emergence of new patterns of
behavior are all of great interest (Ellis & Larsen-Freeman, 2009). Dynamics-dominant
research includes a focus on relational dynamics, trajectories of change and development,
self-organized processes, and emergent outcomes. Of course, because complex systems
also have constituent parts that together make up the system, another basic approach is
interaction-dominant research using relation-intensive methods. These designs describe
systemsparts and their interactions, providing a focus on the complex underlying
structure of interdependent relations (Hilpert & Marchand, 2018).
Especially important for our purposes, meta-theories such as CDST function as the
necessary intellectual blueprint for conducting and evaluating research (Overton, 2015).
For instance, Hiver and Al-Hoorie (2016) suggested that the core objectives of CDST
research in applied linguistics should be to
(a) represent and understand specic complex systems at various scales of description; (b) identify
and understand dynamic patterns of change, emergent system outcomes and behavior in the
916 Phil Hiver, Ali H. Al-Hoorie, and Reid Evans
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
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 inuencing the behavior of systems.
(p. 752)
These broad objectives may serve as guiding parameters for study design as well as a way
to gauge the overall contribution of a study or body of work.
There are other criteria to use when designing and evaluating CDST research in applied
linguistics (Larsen-Freeman & Cameron, 2008; Verspoor et al., 2011). A useful point of
entry are the operational considerations such as deciding what to case as a complex
system, the boundaries of this unit of analysis, and the level of resolution and timescale(s)
at which to analyze that system. Contextual considerations delineate the spatiotemporal
frame of reference for the system and environmental features that are empirically salient to
the system and its development. Macrosystem considerations account for dynamic out-
comes or states in which a system has stabilized and help to pursue a temporal under-
standing of adaptive change and trajectories of development. Microstructure
considerations dene the makeup of a complex system, describing the functional whole,
its constituents, and their relationships and interactions. Together these considerations
provide a window into interpreting system behavior and inducing change in a complex
system (Hiver & Al-Hoorie, 2016).
Very recent methodological advances have emerged in the eld that strive to do justice
to the complex, nonlinear learner development data. The main goal of these CDST-
inspired studies is to develop multifactorial, nonlinear, and probabilistic models that are a
better t for such complex and dynamic language learner data than those currently
available. For instance, a number of recent studies informed by CDST (e.g., Kliesch &
Pfenninger, 2021; Murakami, 2016,2020; Pfenninger, 2020; Verspoor et al., 2021) use
generalized additive (mixed) modeling (GAMM) to (a) tease apart spatially distributed
between- and within-learner variation, (b) disentangle mechanisms that have differing
inherent time-courses (e.g., what aspects have the strongest impact on ongoing L2 writing
development and over what timescale?), and (c) examine a systems interconnected
structure as well as its dynamic behavior (e.g., what interactions occur between various
cognitive and noncognitive ID variables across time). This approach examines variability
as an informative data point in its own right (Verspoor & de Bot, 2021) and includes
variability in its algorithms; it is, thus, ideal for analyzing nonlinear change over time in
iterated learning experiments.
AN INTEGRATIVE FRAMEWORK FOR CDST RESEARCH
Many applied linguists have recognized that the issues they are tackling are fundamentally
complex, broad, and systemic (Han, 2020; see Larsen-Freeman, 2017, for a conceptual
review). With CDST methods, the debate around the merits of qualitative versus quan-
titative research has been superseded by concern for the merits of individual versus group
level (i.e., high- or low-N) designs and analyses, and the timescale or number of occasions
(i.e., high- or low-T) appropriate for these designs. CDST encourages design decisions at
several distinct levelsaim, unit of analysis, and method (Figure 1)that reorient
research toward processes of learning and development rather than exclusively focusing
Complex Dynamic Systems Theory in Language Learning 917
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
on the product of learning (see also Larsen-Freeman, 2020). An integrative framework
that combines these elements of research assumptions and design choices can be used to
evaluate the contribution of CDST research.
Starting with the aim, an integrative design might be exploratory or may attempt to test
certain understandings or expectations, including observationally and (quasi-)
experimentally. Although the complex social world does not lend itself to universals that
can be applied across all settings and populations, it is nevertheless possible to form
probabilistic predictions by comparison to other similar systems, under similar conditions
and contexts, with similar outcomes (Hiver & Al-Hoorie, 2020b). Consequently, when
using CDST research tools in applied linguistics, there is no reason to shy away from
making predictions and then subjecting these predictions to empirical test. As the double-
headed arrow shows (Figure 1), integrative CDST designs should take both of these aims
into account. Adopting a dual exploratoryfalsicatory approach can radically reorient
researchers and their aims, making them actively seek negative, disconrming results
rather than exclusively celebrating positive ones and experiencing disappointment when
encountering inevitable negative ndings.
A second level where a study can be designed in an integrative way is the unit of
analysis, which has to do with whether the level of granularity in a CDST study is at
the individual or the group level. Here some have contrasted an idiographic, person-
centered, individual-level approach with a nomothetic, variable-centered, group-level
approach (see e.g., Lowie & Verspoor, 2019). The former is focused on nding what is
unique in each individual, while the latter looks for generalizations that apply across
many individuals. This unit also applies to timescales and processes of change in
which the nomothetic approach emphasizes general proles of interindividual vari-
abilityoften using cross-sectional dataand the overall mean trajectory of all cases,
whereas the idiographic approach emphasizes intraindividual variabilityoften
longitudinallyand the unique developmental trajectories of each individual
(Verspoor et al., 2011).
Group-based research
2
remains popular in applied linguistics research, though
individual-based designs may allow researchers to more readily operationalize the
assumptions of CDST (Lowie, 2017) because this type of research holds a close lens to
development and change without averaging away individual idiosyncrasies (Molenaar &
Campbell, 2009). The utility of both individual and group based designs also squares with
Molenaars(2015) thinking on the appropriate level of granularity in such researchnot
requiring an exclusive focus on the individual case, but instead centering the objective to
build more adequate models that take into account individual factors without giving up the
search for general patterns and tendencies: analyses of intra-individual variation does not
preclude valid generalization across subjects. In this way nomothetic knowledge about
idiographic processes can be obtained(Molenaar, 2015, p. 37). Individual-based
research designs allow meticulous analyses of single cases while group-based results
uncover broader tendencies that can show how these results vary in the population. If a
group is the system chosen as the unit of analysis for research, or if it is any higher-level
system than an individual, then it may be that group-level data are more relevant for that
particular study. An integrative design at the unit of analysis would attempt to draw from
both the individual-level and group-level of analysis which are complementary from a
CDST perspective.
918 Phil Hiver, Ali H. Al-Hoorie, and Reid Evans
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
The nal choice is the method. Integrative CDST designs can draw from both
qualitative and quantitative methods to advance knowledge in a particular area of applied
linguistics. CDST research encourages mixing quantitative or qualitative methods to
investigate broad questions of interest (Hiver et al., 2021; Larsen-Freeman & Cameron,
2008). Whether quantitative, qualitative, or some integrated combination, CDST methods
deal primarily with longitudinal data if they operate using the adaptive principle, but may
also apply to cross-sectional data if concerned with the relational principle. Longitudinal
data and designs are usually more CDST compatible because these focus on the outcomes
or patterns that are reached at different points in time as well as the mechanisms that
explain how an outcome is reached. Additionally, it is nearly impossible to study change
and development (the adaptive principle) without also accounting for context and
interconnectedness (the relational principle).
THE PRESENT STUDY
As 25 years has passed since CDST was introduced to the eld, it is time to look back at
this body of research and systematically review it. As mentioned in the preceding text, we
approached this scoping review project with two parallel objectivesone descriptive and
one substantive. These correspond with our research questions. Given this body of
research spanning the 25 years from 1994 to 2019, we asked the following research
questions:
RQ1. What are the methodological characteristics of CDST studies in the eld (including partic-
ipants, contexts, timescales, and analytic strategy)?
RQ2. What are the substantive contributions of these CDST studies to the eld?
RQ3. What areas for improving CDST study quality are apparent?
METHOD
INITIAL SEARCH
We conducted a search for studies spanning the 25-year period of interest (19942019).
We chose this period because 1994 marks the date of the very rst contribution on the
topic of complexity theory/dynamic systems theory in the elda conference paper
delivered by Larsen-Freeman (1994) at the Second Language Research Forum. Our scope
covered peer-reviewed articles, book chapters, conference papers and proceedings, and
FIGURE 1. Integrative framework for CDST research designs proposed by Hiver and Al-Hoorie (2020b).
Complex Dynamic Systems Theory in Language Learning 919
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
doctoral dissertations. We conducted our search in databases relevant to our eld
(i.e., ERIC, MLA, ProQuest, and PsycINFO) using the search terms shown in Figure 2.
As we describe, we also looked beyond the results of the database searches at this stage to
ensure that important and pertinent research reports were not overlooked. Figure 3 shows
this entire process.
As Alexander (2020) proposes, when constructing a report pool, a robust search
procedure must justify the specic delimitations instituted with consideration of the
potential consequences of those decisions. With this in mind, we rst specied where
FIGURE 2. Database search terms.
FIGURE 3. PRISMA ow chart illustrating identication of studies through database search.
Source: Page et al. (2021).
920 Phil Hiver, Ali H. Al-Hoorie, and Reid Evans
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
search terms should appear (i.e., in one or both the abstract or main text) to avoid the false
negatives likely to arise from either more generic or polysemous use of the term
complexity (e.g., used to denote a measure of language production) exclusively in titles
and keywords. This restriction also enhances the replicability of our approach. This search
returned a total of 2,341 hits from the combined database. We then supplemented this pool
by a Google Scholar search and an ancestry search as redundancy checks.
To mitigate selection and publication biases, we also set out to intentionally incorporate
so-called gray literature (Rothstein & Hopewell, 2009) in our report pool. This includes
nontraditional research documents that are found outside of typical publishing venues
such as organizational reports, working papers, and conference proceedings. Finally, we
put out a call to solicit unpublished work, edited volumes not cataloged by the search
engines, or preprints we might have missed in our search. We then examined this total
report pool against the inclusion criteria described in the next section.
INCLUSION CRITERIA
To be eligible for inclusion in this scoping review, the report had to satisfy the following
criteria:
1. It must involve an empirical design (whether quantitative, qualitative, or mixed method).
Methodological and conceptual articles
3
were excluded.
2. It must explicitly identify itself as operating within, or informed by, CDST or its terminological
antecedents.
3. It must be related to language learning. Reports on either nonlanguage education or theoretical
linguistics were excluded.
4. It must be in English.
5. It must be available before August 2019.
Here we must add several caveats about our inclusion criteria. First, work in the eld over
the past two decades has emerged from two related theoretical frameworkscomplexity
theory (CT) and dynamic systems theory (DST) (see e.g., Larsen-Freeman, 2007). As
many readers and scholars in this domain will suspect, it is unlikely that those working
with complexity theoryin SLD were doing different work than those working with
dynamic systems theory.Consensus simply had not yet been reached on terminology.
CDST is a more recent amalgam
4
that reects the self-organization of nomenclature.
While it has become the elds theoretical umbrella term of choice, it is an emergent entity
with both new and existing properties of CT and of DST (e.g., Larsen-Freeman, 2017).
Because we locate much of our own work within this paradigm, we were hyperaware of
this terminological diversity and were explicit about looking for these terminological
antecedents in the report pool.
Second, unlike some synthetic work in our eld guided by specic questions (e.g.,
how effective is form focused instruction?; see Kang et al., 2018), here it is a theoretical
framework that drives our inclusion criteria. As a result, an element of self-selection is
inherent when ltering out all studies that did not self-identify as being CDST research.
The procedural challenge in creating such a report pool is, of course, that the decision of
which studies to include or exclude markedly inuences the outcome of the review. For
instance, we are aware of several empirical studies routinely cited by CDST scholars as
exemplars of this approach that are nevertheless not framed by the original authors as
Complex Dynamic Systems Theory in Language Learning 921
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
CDST research, or that never mention being informed by CDST (see e.g., Eskildsen,
2009). However, a scoping review with search terms and inclusion criteria like ours could
not subjectively include such studies on a case-by-case basis as report pool construction
would become arbitrary and lack reproducibility. We cast a wide net with this inclusion
criterion and sampled self-labeled CDST studies without preltering how robust this self-
labeling was or whether studies focused exclusively on CDST. Because of this, the report
pool included a heterogeneous array of topics and themes. We, therefore, acknowledge
this limitation, and are cautious in interpreting this report pool as a awless representation
of CDST research on second language learning.
CODING
Applying our inclusion criteria rst to the title-abstract-keyword of all unltered reports,
we obtained a total of 488 reports (see Figure 4 for a yearly breakdown). No proceedings,
conference papers, edited book chapters, or unpublished work met all our inclusion
criteria, primarily due to lack of explicit detail as to how they were informed by CDST.
Journal articles in this pool were primarily, though not exclusively, from SSCI and
SCOPUS indexed journals. While these journals have been observed to present high-
quality research, which the eld trusts as both robust and consistent (Andringa & God-
froid, 2020), restricting reports to such journals may present a representativeness limita-
tion. For this reason, we did not undertake any further ltering of journal articles.
Presumably because CDST is perceived as a comparatively novel theoretical orientation
that has high potential for application in empirical work, there were many dissertations in
our pool. For the sake of a comprehensive sample and parsimonious analysis, here we
combined dissertations with journal articlesthough we acknowledge that dissertations
often do not tend to follow conventional journal preferences, are broader in scope, and
often include innovative ideas, but also undergo a somewhat different review process than
peer-reviewed journal articles. This pool of studies was then manually inspected against
the inclusion criteria by all three authors and discussed until 100% agreement was
0
10
20
30
40
50
60
70
80
FIGURE 4. Studies ltered from initial search (k=488) by year of publication.
922 Phil Hiver, Ali H. Al-Hoorie, and Reid Evans
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
reached. As a result, 158 reports were retained in the nal pool (89 journal articles and
69 dissertations).
These 158 reports were then coded individually using a descriptive categorization
scheme (see Supplementary Material) that included detailed markers such as study design
and length as well as more substantive descriptors such as empirical contribution and
study limitations. Each researcher coded a third of the nal pool. To validate these
judgments a second researcher along with a team of two trained coders independently
coded 30% of all reports. The observed interrater agreement (83.6%) across coding
categories was above the conventional 80% threshold (McHugh, 2012) and the observed
kappa (κ=.67, p< .001) approached conventional agreement standards. While kappa is a
conservative estimate of interrater agreement, especially as possible categories increase
(Brutus et al., 2010), we consider the reliability of our coding to be acceptable, but we
acknowledge that future researchers may improve upon it.
RESULTS
METHODOLOGICAL CHARACTERISTICS
Starting with the characteristics of participants found in CDST research, varied sample
sizes and participant age groups were included. Figure 5 shows that 14 studies included an
Nof 1, and that in this pool there were fewer studies as sample size increased. When
combined with several other design characteristics, this highlights the increasing impor-
tance of individual-based and idiographic research. Though a handful of studies included
larger samples, perhaps due to CDSTs interest in the individual learner sample sizes
tended to be modest. Roughly 40% of all studies featured a sample size of N10, and only
13 studies in the entire report pool included a sample of N> 100. The largest sample size in
the article pool was N=924 (Mdn =13.5, IQR =31), while the largest sample size in the
dissertation pool was N=1,723 (Mdn =16, IQR =28.5). Within this pool, studies with
younger participants were clearly the minority (Table 1), with 112 studies (70.8%)
0
2
4
6
8
10
12
14
16
0 102030405060708090100
Number of Studies
Sample Size
FIGURE 5. Sample size of studies in the report pool.
Complex Dynamic Systems Theory in Language Learning 923
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
sampling either university students or adults aged 18 or older. The rarest were studies with
participants aged seven years and younger (4 studies) followed by those with respondents
aged 712 (10 studies). Eight studies featured multiple, mixed age groups, while the age
of participants was unspecied in nine studies. While this may reect some of the elds
sampling tendencies in general these characteristics have remained unexplored in CDST
research to date.
Because CDST is a relational-contextual perspective in which spatiotemporal context
plays an integral role in making sense of empirical ndings, we expected adequate depth
of contextual detail to feature in the studies we reviewed. Table 2 shows that a wide range
of research contexts were represented in the study pool, with foreign and second language
learning contexts accounting for 132 studies (83.5%) of the total. Other research contexts
were only minimally present, including bilingual language contexts, heritage language
contexts, and a mix of several of these within the same study.
Various instructional settings were also part of this pool. In addition to the 79 studies
(50%) that took place in conventional instructed language settings, our pool showed that
only a handful CDST studies have been conducted in online learning, in immersion
environments, in study abroad contexts, or in language for specic purposes classrooms.
Only three studies investigated untutored, naturalistic language learning. Considering the
importance of context in CDST research, the number of studies that left unspecied either
the research context (14 studies; 8.8%) or the instructional setting (41 studies; 26%) was
largea point we turn to in our discussion.
Participants also represented various L1 backgrounds and target L2s (Table 2). We
categorized a total of 24 different L1s here based on their geographical origin for the sake
of parsimony (i.e., some studies featured multiple languages). Far fewer target languages
were featured. Among these, what stands out is the dominance of L2 English as a target
language, accounting for nearly 70% in the pool. Though we only included reports written
in English, this imbalance is perhaps to be expected given the global importance of L2
English. It also stands in contrast to the relatively low frequency of other languages that
are, arguably, equally widespread and important target languages. Spanish was the second
most represented L2 in our pool (10.1%), while some world languages were featured in
just a single study. Finally, eight studies did not specify the target language in question.
Turning to study design characteristics, we looked at the general approach to study
design as well as the timescale of data collection in the reviewed studies (Table 3).
TABLE 1. Participant characteristics in CDST research
k%
Age
young children (< 7) 4 2.5%
primary school (712) 10 6.3%
secondary school 15 9.4%
university (18þ) 48 30.3%
other adult (25þ) 64 40.5%
multiple, mixed age groups 8 5%
unspecied 9 5.6%
Note:k=158.
924 Phil Hiver, Ali H. Al-Hoorie, and Reid Evans
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
Whereas over a third (59 studies) were cross-sectional, more than 53% of studies
(84 studies) were longitudinal in design. In relation to the eld more generally, this is a
substantially higher proportion (Al-Hoorie & Vitta, 2019). The overall approach to data
collection or data sampling was ambiguous in the remaining 15 studies. Examples of these
include analyses of usersasynchronous chat messages, video observations of classroom
interaction patterns, computer-assisted corpus analysis, and analysis of classroom
TABLE 2. Contextual characteristics in CDST research
k%
Research Context
foreign language 90 56.9%
second language 42 26.5%
bilingual 2 1.2%
heritage 1 0.6%
mixed 9 5.6%
unspecied 14 8.8%
Instructional Setting
generic instructed setting 79 50%
online/VLE 15 9.5%
immersion 10 6.3%
language for specic purposes 5 3.1%
uninstructed setting 3 1.9%
study abroad 2 1.2%
mixed 3 1.9%
unspecied 41 26%
Participant L1
European languages 95 60.1%
Asian languages 45 28.4%
Middle Eastern languages 10 6.3%
specied as mixed23 14.5%
unspecied 7 4.4%
Target L2
English 110 69.6%
Spanish 16 10.1%
French 9 5.6%
German 8 5.0%
Chinese (inclusive) 7 4.4%
Japanese 6 3.8%
Arabic 2 1.2%
Swedish 2 1.2%
Finnish 1 0.6%
Italian 1 0.6%
Persian 1 0.6%
Polish 1 0.6%
Portuguese 1 0.6%
Vietnamese 1 0.6%
unspecied 8 5.0%
Note:k=158. Numbers for participant L1s may not sum to the total number of studies due to some studies
including multiple samples. Numbers for target L2 may not sum to the total number of studies due some studies
including multiple target languages. VLE =virtual learning environment.
Complex Dynamic Systems Theory in Language Learning 925
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
pedagogical artifacts. With regard to study length, data elicitation took place most often
over a span of months (54 studies), followed by studies with a timespan of weeks
(33 studies), years (32 studies), hours (9 studies), and days (5 studies). Comparative data
from other reviews in the eld indicates that this proportion of studies with a time window
of months and years is markedly higher in CDST studies (Vitta & Al-Hoorie, 2020). Study
length in our report pool ranged from 90 minutes to 4 years. Note that these numbers do
not refer to the frequency of data elicitation but to the duration of the study. More often
than not, details regarding the frequency of data collection were not specied in these
studies, which made it difcult to determine, for instance, if studies with a timespan
measured in weeks elicited data from participants daily over this period, twice (at the start
and end of this period), or only once per participant over the course of the study.
With reference to dynamic method integration (Table 4), CDST research entails design
decisions at several distinct levels: study aim, unit of analysis, and choice of method. It is
perhaps notable that more than 80% of studies (130 studies) were exploratory and only
28 studies had a falsicatory aim, that is to test hypotheses empirically that are related
either to CDST principles (e.g., that intraindividual variation is informative about
development) or topically circumscribed predictions (e.g., that there are regularities in
trajectories of L2 development). No single study we reviewed combined both exploratory
and falsicatory aims, a nding that seems counter to the hybrid nature of a great deal of
research in the eld. However, by necessity we coded these notions (conrmatory
vs. exploratory) from the research objectives formulated by studies in the report pool
and from characteristics of their research designs, not by examining claims made by
authors that their data conrmedor supportedcertain conclusions after the fact.
The choice of unit of analysis was also straightforward for many studies in this pool.
The unit of analysis in 73 studies was the group, and in 70 studies it was the individual. Six
studies specied the unit of analysis as texts (i.e., learner language), and the unit of
analysis was unspecied in four studies. There were ve studies in this pool that included
both individual analyses and group analyses as explicit comparisons across levels. These
we classied as having more than one unit of analysis. While this is a very small subset of
studies, they illustrate the extent to which relying exclusively on group-level data and
TABLE 3. Timescales in CDST research
k%
Approach
longitudinal 84 53.1%
cross-sectional 59 37.3%
unspecied 15 9.5%
Study Length
hours 9 5.6%
days 5 3.1%
weeks 33 20.8%
months 54 34.1%
years 32 20.2%
unspecied 26 16.4%
Note:k=158.
926 Phil Hiver, Ali H. Al-Hoorie, and Reid Evans
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
insights may impoverish the elds understanding of various phenomena (see also Lowie
& Verspoor, 2019).
Table 4 further shows that choice of method was split across qualitative (74 studies),
quantitative (46 studies), and mixed methods (36 studies). Here we adopted an inclusive
denition of methodology related to the purpose, focus, design, procedures (e.g., means of
sampling, data collection, and analysis) of studies in the report pool. Two studies in the
total pool did not describe their methodological choices clearly. The large number of
purely qualitative studies may reect the general tendency for newcomers (e.g., graduate
students or scholars newly interested in CDST) attempting to apply methods for inves-
tigating interconnectedness and dynamic development to default to methods that capture
rich dense datasets(Ushioda, 2021, p. 252). This is borne out in our data, with roughly
80% of dissertations in our pool drawing heavily on qualitative designs. While our review
in no way suggests that exclusively qualitative methods are poorly suited to studying
complexity and dynamicity, we did nd particular limitations in the present pool of
studies, two of which relate to collecting data and adopting analyses that do not lend
themselves to either investigating connections in context or to dynamic change and
development. We leave discussion of these issues until later.
Closely related to the design decisions we reviewed in the preceding text are the choices
of data elicitation methods and data analytical strategies. In contrast with methodological
work suggesting that CDST research should both innovate with existing methods and
expand on these (e.g., Lowie, 2017; MacIntyre at al., 2017), we found that a range of
conventional and widely used techniques for data collection were present in reviewed
studies (Table 5). The technique most frequently adopted was interviews and focus groups
(68 studies; 43%). Other data elicitation methods included analysis of written samples of
learner language, oral language/interaction samples, and observations. Surveys, tests, and
pedagogic tasks were also commonly employed by CSDT researchers. Other data sources
used more sparsely included think-aloud protocols, stimulated recall, and eld notes.
TABLE 4. Elements of the dynamic method integration framework in CDST research
k%
Study Aim
exploratory 130 82.3%
falsicatory 28 17.7%
both 0 0%
Unit of Analysis
group 73 46.2%
individual 70 44.3%
texts 6 3.8%
more than one unit 5 3.1%
unspecied 4 2.5%
Methodology
qualitative 74 46.8%
quantitative 46 29.1%
mixed 36 22.8%
unspecied 2 1.2%
Note:k=158.
Complex Dynamic Systems Theory in Language Learning 927
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
Thirteen studies featured other types of data elicitation tools such as samples of student
academic work, drawing tasks, or momentary sampling measures (e.g., the idiodynamic
approacha research template that collects data on time-dependent variation within a
single individual or unit). Notably, the majority of studies in the report pool, in both the
article and dissertation subsets, included multiple complementary data sources. Studies
that did so included at least two but often up to four data sources in combination, and were
distributed across nearly all years. This may reect a general tendency to approach data
collection in CDST research with a more is morementality: because everything counts,
everything is connected, and everything changes, study design may have followed the
premise that more data is more appropriate to examine such phenomena fully.
Turning to analysis techniques, qualitative coding and analysis methods appeared to be
those employed most often in the reviewed studies (64 studies; 40.5%), perhaps a logical
extension of the large number of studies that adopted qualitative data collection tech-
niques. This was nearly triple the frequency of the next largest category of analysis
techniques. Qualitative data analysis techniques here included content and discourse
analysis, ethnographic analysis, inductive thematic coding or grounded theory analysis,
and metaphor analysis. Twenty-four other studies (15.2%) adopted dynamic statistical
analysis such as using the coefcient of variation (2 studies), min-max graphs and moving
correlations (5), recurrence quantication analysis and Monte Carlo simulation (3),
growth curve modeling (1), time-series analysis (5), generalized additive mixed-effects
models (1), state space plots and grids (1), fractal analysis (1), or trend analysis (3) and
timeplots (2). When examining other data analytical strategies, we found that eight studies
relied on descriptive statistics alone (not including studies reporting effect sizes) and a
TABLE 5. Analytical strategies in CDST research
k%
Data Collection
interview/focus group 68 43%
written language sample 44 27.8%
survey/questionnaire 41 26%
lesson observation 35 22.1%
tests 25 15.8%
oral language/interaction sample 23 14.5%
pedagogic task 17 10.8%
eld notes/memos 12 7.6%
stimulated recall 4 2.5%
think-aloud protocols 2 1.2%
other 13 8.2%
Analysis Technique
qualitative coding/analysis 64 40.5%
conventional inferential statistics 27 17.1%
dynamic statistical analyses 24 15.2%
descriptive statistics 8 5%
advanced multivariate statistics 4 2.5%
unspecied 74 46.8%
Note:k=158. Numbers for data collection strategy and analysis technique may not sum to the number of studies
due to many studies including multiple types of data and multiple analyses.
928 Phil Hiver, Ali H. Al-Hoorie, and Reid Evans
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
further 27 studies adopted conventional inferential statistical analyses. These included
analyses such as t-tests, canonical correlations, analyses of variance (ANOVA), and linear
regression analysis. A handful of other advanced multivariate statistical analyses were
used (four studies), including factor analysis and principal components analysis, cluster
analysis, and latent variable modeling (i.e., SEM). We also found a large number of
instances (74 studies; 46.8%) in which the data analysis technique was either unclear or
unspeciedexamples of this include unintuitive descriptions such as we analyzed our
data in Excelor the data were coded manually.The nding that a large proportion of
studies
5
did not fully establish methodological integrity for the reader is one we return to
in the following text when reecting critically on our other research questions.
SUBSTANTIVE CONTRIBUTIONS
In addition to methodological characteristics of these studies, we were also interested in
determining what substantive contributions this pool of studies has made to the eld.
Because we cast a wide net and sampled self-labeled CDST studies without preltering
how robust this self-labeling was or whether studies focused exclusively on CDST, the
report pool included a heterogeneous array of topics and themes (e.g., learnerspercep-
tions toward classroom tasks, how digital games mediate language use, language attrition
in rst generation immigrants, and the development of authorial voice and rhetorical
knowledge in L2 writing, etc.). Across all these we looked at contributions in two broad
areas: rst, empirical contributions and, second, practical contributions (i.e., related to
both research and pedagogy) to the eld.
Table 6 shows that empirical contributions were demonstrated in a variety of areas.
Two of the most noticeable contributions were that studies reported evidence supporting
the claim that the phenomena or constructs under study were indeed complex and
dynamic: Thirty-one studies (19.6%) corroborated the existence of dynamic regularities
in development, and another 29 studies (18.3%) provided evidence of system intercon-
nectedness and interaction between elements being studied. Other notable contributions
included evidence of the inuence of context in development, of the nonlinearity of
development or the presence of nonlinear predictors, of emergent outcomes and patterns,
and of system adaptation or self-organization in response to inputs or to contextual
affordances. Among other contributions were studies that provided evidence of inter-
and intraindividual variability, as well as studies illustrating the methodological value of
applying CDST tools to advance understanding in the eld and the compatibility of CDST
with previous research drawing on other diverse paradigms. A small number of also
established evidence of sensitivity to initial conditions and of equinalitythe notion that
a given state or outcome can be reached through multiple pathways. Here we intentionally
focused our coding on these categories because many of these contributions are distin-
guishing features of CDST that other theories do not account for or even investigate.
We were, of course, interested in what practical contributions CDST studies have made
to the eld. Such contributions are the subject of recent work (e.g., Levine, 2020) and,
because they are sought after by many, perceptions that such applications are not readily
accessible may act as a curb on wider uptake of CDST in the eld (Dewaele, 2019).
Table 7 shows that practical contributions in reviewed studies were not few in number.
Contributions ranged widely from studies offering direct pedagogical insights (34 studies)
Complex Dynamic Systems Theory in Language Learning 929
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
and explicit discussion of a fuller, more multidimensional understanding of the phenom-
ena under investigation (24 studies), to the explanatory power of contextual factors in
developmental over and above other explanans (23 studies), and conrmation of the
particularities of individuals and intraindividual variation (13 studies). Another contri-
bution was the emergence of new previously undiscovered or unapplied criteria for
existing issues (10 studies)for example, using notions of system adaptation from CDST
in understanding the development and maintenance of multicompetence, and drawing
insights from both CDST and evidence regarding maturational constraints in relation to
L1 attrition during L2 acquisition. Other practical contributions related closely to appli-
cations for research across these heterogenous topics. This includes studies that applied a
novel perspective that helped uncover new insights into the phenomena under investiga-
tion (23 studies), studies that shifted attention to new aspects of existing phenomena
(13 studies), or those that showed the limitations of existing perspectives (9 studies). Still
others made contributions by integrating multiple complementary data sources (17 stud-
ies), developing new conceptual tools for the topics being studied (10 studies), tapping
TABLE 6. Empirical contributions to second language development
k%Sample studies
Evidence of dynamic regularities in development 31 19.6% (Baten & Hakansson, 2015; Lenzing, 2015;
Nelson, 2011; Pham, 2011)
Evidence of system interconnectedness/
interaction
29 18.4% (Jessner et al., 2015; Plat et al., 2018; Serani,
2017)
Methodological innovation/advances 22 13.9% (Caspi & Lowie, 2013; Huang, 2010;
Murakami, 2014; Poupore, 2018)
Evidence of nonlinear predictors/growth 21 13.3% (Churchill, 2007; de Leeuw et al., 2013;
Spoelman & Verspoor, 2010)
Evidence of the role of context in development 20 12.7% (Fogal, 2015; Hepford, 2017; Kostoulas,
2015; Pomerantz & Bell, 2007)
Evidence of emergent patterns 19 12.0% (Baba & Nitta, 2014; King, 2013; Larsen-
Freeman, 2006; Van Horn, 2017)
Compatibility of CDST with other paradigms 18 11.4% (Feryok, 2010; Göpferich, 2013;
Viswanathan, 2016)
Evidence of interindividual variability 14 8.8% (Fukuda, 2014; Kopečková et al., 2016;
Larsen-Freeman, 2006)
Evidence of system adaptation/self-organization 13 8.2% (Bragg, 2018; Larsen-Freeman, 2006; Reigel,
2008; Roehr-Brackin, 2014)
Evidence of intraindividual variability 12 7.6% (Kim et al., 2017; Lowie et al., 2017; Opitz,
2013; Polat & Kim, 2013)
Evidence of nested systems 10 6.3% (Ebert et al., 2014; Liu & Chao, 2018; Mercer,
2011; Pham, 2016)
Evidence of sensitivity to initial conditions 9 5.6% (Finch, 2010; Perales Escudero, 2011;
Waninge et al., 2014)
Evidence of equinality 9 5.6% (Murakami, 2014; Smyk, 2012; Souza, 2013)
Complementarity of group and individual levels 3 1.9% (Lowie & Verspoor, 2019; Xiao, 2016)
Evidence of fractal nature of development 1 0.6% (Evans, 2019)
Evidence of multicausality 1 0.6% (Morel, 2019)
Note:k=158. Number of theoretical/empirical contributions may not sum to the total number of studies due to
some studies including multiple contributions. Full references to the studies cited in this table are listed in the
online supplementary material.
930 Phil Hiver, Ali H. Al-Hoorie, and Reid Evans
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
into greater phenomenological reality in the issues under investigation (8 studies), and
achieving superior ecological validity (10 studies).
STUDY QUALITY
Our third and nal research question relates to methodological rigor and what areas, if
any, were apparent for improving CDST research going forward. To this end, we
examined apparent limitations of study design (Table 8) in our review pool. Note that
these were design limitations we explicitly coded as such and not those listed by authors as
limitations of their studies.
Some of the most prevalent design issues we identied were studies relying on data or
analyses that were seemingly inappropriate for investigating change and development
(41 studies; 26%), and studies relying on data or analyses that were poorly suited to
investigating connections in context (22 studies; 14%). For instance, it is not hard to
appreciate why studies drawing on a single round of interviews or cross-sectional test data
at one or two time points would struggle to shed light on such issues. This result was also
incongruent with the strong evidence in this pool that phenomena of interest or constructs
under study were complex and dynamic (see Table 6). We return to this unanticipated
nding further in the text that follows and reect on the extent to which these studies were
indeed informed by CDST in their design.
TABLE 7. Practical contributions to the field of language learning
k%Sample studies
Pedagogical insights 34 21.5% (Fukuda, 2014; Mercer, 2011; Yashima et al., 2018)
Multidimensional understanding of
issues
24 15.2% (Baba & Nitta, 2014; de Leeuw et al., 2013; Hepford,
2017; Nelson, 2011)
Importance of context 23 14.5% (Braga, 2013; Thompson, 2017; Waninge et al., 2014)
Applying a novel perspective to topic
under investigation
23 14.5% (Jessner et al., 2015; Polat & Kim, 2013; Xiao, 2016)
Integrating multiple complimentary data
sources
17 10.8% (Kikuchi, 2017; Lasagabaster, 2017; Pham, 2016)
Attention to unexplored aspects of
existing phenomena
13 8.2% (Churchill, 2007; Lowie et al., 2016)
Learner differences/variability as source
of information
13 8.2% (Lowie & Verspoor, 2019; Rosmawati, 2014;
Spoelman & Verspoor, 2010)
Emergence of new insights on existing
topic
10 6.3% (Larsen-Freeman, 2006; Serani; 2017; Wang, 2016)
Increased ecological validity 10 6.3% (Bragg, 2018; Cameron & Deignan, 2006; Scholz,
2017)
Developing new conceptual tools 10 6.3% (King, 2013; Link, 2015; Plat et al., 2018)
Evidence of the limitations of existing
perspectives
9 5.6% (Baten & Hakansson, 2015; Kounatidis, 2016; Larsen-
Freeman, 2006)
Capturing phenomenological reality 8 5.0% (de Bot, 2015; Onitsuka, 2018; Zheng, 2016)
Demonstrating feasibility of a
transdisciplinary approach
1 0.6% (Göpferich, 2013)
Note:k=158. Number of practical contributions may not sum to the total number of studies due to multiple
contributions coded in some studies and none in others. Full references to the studies cited in this table are listed
in the online supplementary material.
Complex Dynamic Systems Theory in Language Learning 931
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
Other design limitations we observed included the limited scope of data many studies
drew conclusions from (31 studies; 19.6%) and sample selection bias (21 studies; 13.3%),
evident, for example, in studies with no sampling frame or a nonpurposive sample.
Evident here too was the limited transferability or generalizability of a handful of
accompanying conclusions to similar samples or contexts (15 studies; 9.5%), due to
inattention to external validity. It is rarely generalizability in its conventional sense that
CDST scholars are chasing (Hiver & Al-Hoorie, 2020b; Larsen-Freeman, 2017). How-
ever, especially when considering the lack of detail in specifying contextual factors
(Tables 2,3, and 8) and data analysis techniques adopted (Tables 5 and 7) that was
apparent in some studies, this nding was not entirely unanticipated.
Several other limitations in study design highlighted through our coding include the
presence of some ambiguity in the application of CDST concepts and terminology
(15 studies; 9.5%). This may be partly due to our inclusion criteria which selected for
studies self-labeled as CDST. In several studies, for example, readers are presented with
direct claims about the importance of CDST for the research but based on the questions
explored in the study and the design and methods used; it was unclear how CDST had
informed the study. In several other studies that were terminology heavy, it was unclear in
lay terms what the systembeing discussed by the researchers was, what precisely made
it adaptive,”“self-organizing,or nonlinearin nature, or what patterns were
emergent.This limitation links to another, regarding the exclusive metaphorical
application of CDST applying only its terms or concepts (12 studies; 7.6%)these were
distributed nearly equally across report type and year of publication. Larsen-Freeman and
Cameron (2008) propose that CDST is a necessary metaphor that can push the eld
towards radical theoretical change(p. 11) but they are equally clear that CDST is much
more than metaphor when it is literalized into eld-specic theory, research, and
practice(p. 15). We agree, and discuss below how future applications of CDST might
extend beyond its value as metaphor.
Other less frequent study limitations we observed included underspecied participant
information and analytical techniques (6 and 10 studies respectively), the ecological
fallacyassuming that relationships observed for groups apply equally for individuals
TABLE 8. Study limitations in CDST research
k%
data/analyses incongruent with investigating dynamic change 41 26.0%
limited scope of data 31 19.6%
data/analyses incongruent with investigating connections in context 22 14%
sample selection bias 21 13.3%
ndings with limited transferability/applicability 15 9.5%
ambiguity in applying CDST concepts 15 9.5%
contextual factors underspecied 14 8.8%
exclusive metaphorical application of CDST 12 7.6%
analytical strategy underspecied 10 6.3%
ecological fallacy 8 5.0%
participant information underspecied 6 3.8%
statistical assumptions violated 4 2.5%
abstractness of the analytical strategy 3 1.9%
932 Phil Hiver, Ali H. Al-Hoorie, and Reid Evans
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
and vice versa (8 studies) (Lowie & Verspoor, 2019), and violation of basic statistical
assumptions (4 studies) (see e.g., Al-Hoorie & Vitta, 2019). Taken together these
limitations point to some clear implications regarding areas for improving CDST study
design going forward.
DISCUSSION
This scoping review looked rst at the methodological characteristics of CDST research,
at the contributions this body of research has made to the eld, and nally at CDST study
quality. Our review pointed to clear trends in how the eld has investigated complex and
dynamic phenomena of interest andbased on this body of researchwhat shared
concerns and issues in the eld we now think of differently.
First, this body of work clearly supports the claims that have been made in the
theoretical literature that language, language use, and language development/learning
are complex and dynamicthese are all notions, our review suggests, that are now
undisputed. The two most prominent contributions that studies in our review made are
in fact related to the existence of dynamic regularities in development and the complex,
interconnected, and interactive nature of the topics and constructs under investigation
(Table 6). As mentioned earlier, scholars have previously highlighted several core
objectives of CDST research in applied linguistics (Hiver & Al-Hoorie, 2016). It is clear
from our review that the eld has made particularly strong advances relating to the rst
two of these objectives (i.e., describing various complex systems and identifying various
patterns of dynamic change in context), and has begun work on the third objective
(i.e., modeling complex mechanisms and dynamic patterns), butdespite more than
50% of studies collecting data from an instructed L2 settinghas left the remaining
objective largely aside (i.e., understanding how to intervene or inuence systems
behavior). Applied linguists arguably aim to go further than mere description and enact
certain forms of complex praxis in social contexts (Al-Hoorie et al., 2021; Larsen-
Freeman, 2016a). Application of a elds scienticndings and insights is one of the
most important modes of social science research. By consequence, with more than two
and a half decades of thinking and research on the matter, continued work with descriptive
ndings limited to insights such as phenomenon X is complex in its make-upor
process Y is nonlinear in its developmentis unlikely to push the eld forward in a
substantive way at this stage as such claims are now already established.
The contribution of CDST work going forward will be to offer more robust explanatory
conclusions at increasingly relevant timescales and levels of resolution. Given the shift of
perspective that accompanies a familiarity with CDST, there is a need for greater work on
systemic interventions (Byrne & Callaghan, 2014). While the eld has been quick to
amass evidence that many phenomena are relational, nonmechanistic, and indeterminate
in their development, as an applied eld we have yet to do the necessary work to
understand whether and how to intervene in and inuence the complex dynamic realities
of the phenomena under investigation. Here, by intervene in and inuence systems we
mean intentionally generate positive change that is complex, situated, iterative, time
scaled, and reciprocal in nature (see e.g., Steenbeek & van Geert, 2015; van Geert &
Steenbeek, 2014, for similar arguments). Criteria are also needed for developing and
evaluating these systemic interventions that are sensitive to features of context-
Complex Dynamic Systems Theory in Language Learning 933
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
dependence, multiplicity, and interactions. Complex interventions will be those designed
to respond adaptively to a number of relational components in context, when various
levels of the system (e.g., individual, group, or organizational levels) are targeted by the
intervention, managing a number of anticipated and surprising behaviors manifested by
those involved, and leading to variability in outcomes.
6
As our nding that more than 82%
of studies had an exploratory and descriptive aim suggests, we have much to do to think in
CDST terms about deliberate intervention and to develop research tools for this (Osberg &
Biesta, 2010).
Second, in both empirical and applied terms, the important role of context in under-
standing development is clearly apparent. It has almost become a truism for studies to
conclude that the spatiotemporal context plays an integral role in affecting development.
The fact that outcomes and change not only emerge in context, they are also mediated and
adapted by contextual factors(Hiver & Al-Hoorie, 2016, p. 746) is an integral part of
necessary design considerations for CDST research. This conclusion, however, must also
be juxtaposed with the somewhat surprising number of studies reviewed in which the
research setting or the instructional context were either underspecied or unspecied (see
Table 2). This is especially bewildering given the large number of dissertations in the pool
that purport to draw on ecological frameworks (e.g., informed by the work of van Lier,
2004) that presuppose detailed descriptions of context. It is important to be able to develop
evidentiary accountsand explanations that go beyond the uniqueinstance (Byrne & Ragin,
2009), and one way of doing so is to use contextual information to specify the range of
applicability of developmental mechanisms, without essentializing context. Context, which
itself changes, is much more than background variables and should be understood as more
than a constellation of such macrofactors. Going forward, instead of token, perfunctory
mentions that context is inuential, CDST research must articulate explicitly what contex-
tual factors are being taken into account and how context informs study design. This way,
information about the role of particular contextual factors in particular causal mechanisms
will come to be incorporated more clearly and more concretely in evidentiary accounts and
explanations in the eld (see also Kaplan et al., 2020).
Third, as a research community too, the eld has developed new ways of operating that
are accompanied by and that require a different framing(Larsen-Freeman, 2020,
p. 202). The methodological characteristics of this body of CDST research have certainly
made the case that idiographic research is not only valid, but also necessary and important.
It has taken some time for this notion to gain traction, yet judging by the nearly 10% of
studies in the pool with an Nof 1, and a full 45% of studiesregardless of sample size
adopting the individual as the unit of analysis, this is an understanding that has gained
wider acceptance. There is also signicant value in the elds growing recognition of the
importance of innovating with new modes of data elicitation and dynamic analytical
strategies, whether case-based or variable-based (Table 5). Expanding the methodological
repertoire beyond conventional methods and developing expertise in new designs and
analytical techniques are key initiatives that the eld should continue to pursue (see Hiver
et al., 2021; MacIntyre et al., 2017). One indication of the importance of this relates to our
nding (Table 8) that many studies reviewed relied on data or analyses that were
seemingly inappropriate for investigating change and development or were poorly suited
to investigating connections in context. Our report pool contained studies claiming
evidence for dynamic development that did not draw on data with a temporal aspect in
934 Phil Hiver, Ali H. Al-Hoorie, and Reid Evans
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
a way that would allow for such an interpretation. Other reports argued for evidence of
intraindividual variability while looking at data in an insufciently individual way. Form
must follow function: the choice to adopt certain methods of data elicitation and analysis
should be driven by the aim(s), unit(s) of analysis, and the outcome(s) or process(es) under
investigation.
Other ndings also indicate the need for increased transparency and rigor in methodo-
logical designs and in reporting relevant choicesissues also articulated in other subdo-
mains of the eld (see e.g., Hu & Plonsky, 2021; Marsden et al., 2018; Paquot & Plonsky,
2017). For instance, the large number of CDST studies in which the general approach to data
collection and the length of study was unspecied, or the data analysis technique unclear, is
cause for concern. This nding may also be linked to the large number of studies in which
CDST concepts were applied ambiguously, in an exclusively metaphorical way, or due to
their exploratory nature. CDST is not merely a useful set of metaphors for conceptualizing
second language development phenomena: complexity is an empirical reality. As such,
CDST research must move beyond the exclusively metaphorical application that describes
ndings with a language borrowed from CDST (Hiver & Al-Hoorie, 2020b). Metaphors
may be adequate if we wish to conceptualize phenomena (Larsen-Freeman & Cameron,
2008); however, the eld mustmove forward to operationalize and validate thesephenom-
ena and investigate them empirically (see also Brown et al., 2018; Nicklin & Plonsky, 2020;
Vitta & Al-Hoorie, 2021). These ndings suggest the importance of greater transparency
and rigor in the design choices of future CDST research, and also underscore the need for
study designs to clarify the ways in which they are informed by CDST (see Gass et al., 2021,
for a detailed discussion of study quality).
As our inclusion criteria show, our search cast a wide net by including all self-labeled
CDST studies in the report pool. However, our analysis highlighted the fact that this self-
labeling may not always be robust or that reports did not always warrant a CDST label.
Many studies in this pool appeared to operate within a CDST perspective but did not
unambiguously articulate how, or only called attention to the fact indirectly or fairly late.
Some studies were not substantively conceived of or designed as CDST research in any
major sense of what might be expected (i.e., a focus on relational and dynamic phenomena
in context). Specifying that studies explicitly identify themselves as adopting a CDST
perspective or design added clarity to our report pool, but many studies went no further.
What is, therefore, unclear from our review, and rarely transparent from reports them-
selves, is whether studies in our pool approached the phenomena of interest in an
exploratory fashion and discovered that CDST principles t their data and accounted
for these phenomena well, or if studies were in fact looking for evidence of such principles
in their data and so applied these ex ante. By not discussing how CDST informs the design
and methods, studies like these run the risk of spurious assumptions of complex phe-
nomena from a dataset that may not support these claims. This limitation points to the
need for CDST research to take up preregistration and other open science initiatives in
research methods designed to increase study quality (see Hiver & Al-Hoorie, 2020a).
Future applications of CDST research must be transparent about the reasons for
choosing to adopt the CDST metatheory and specify why situating a study within this
perspective is a sound theoretical and empirical choice (Larsen-Freeman, 2017). Artic-
ulating how CDST informs their approach to research explicitly can help researchers
situate the design of their study, their research questions, data analyses, and the results and
Complex Dynamic Systems Theory in Language Learning 935
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
discussion more clearly within this perspective (Lowie, 2017). This can also guard against
using CDST too looselyin the sense that anything with multiple interacting parts can be
construed as CDST researchand in an opportunistic, post hoc manner.
CONCLUSION
Even though it has been a quarter of a century since it was introduced to the eld, CDST is
still a relatively new paradigm. The limitations we reviewed in this article are therefore a
TABLE 9. Nine tenets for CDST research
Tenet Purpose
For an individual study
1. Provide a rationale for why adopting a CDST
research perspective is a sound choice
Helps guard against overly loose and opportunistic
applications of CDST (i.e., applications that are
purely semantic or metaphorical)
2. Articulate how CDST informs the design and
methods
Helps to establish how a study substantively draws
from CDST in its conception and design
Helps avoid spurious assumptions of complex,
dynamic phenomena
3. Specify the aim(s), unit(s) of analysis, and the
outcome(s) or process(es) under investigation
Helps to increase transparency
Helps to leverage an integrative design (see Tenet 8)
4. Adopt methods of data elicitation and analysis
that are driven by the aim(s), unit(s) of analysis,
and the outcome(s) or process(es) under
investigation
Helps to ensure that methods adopted are suited to
investigating connections in context and are
appropriate for investigating change and
development
5. Specify information about the role of particular
contextual factors in particular processes or
outcomes
Helps to incorporate contextual detail more clearly and
more concretely in evidentiary accounts and
explanations
Helps to develop an understanding of contextual
inuences that go beyond the unique instance
For a program of research
6. Identify areas for complex interventions Allows researchers to focus on inuencing, intervening
in, and generating positive change (e.g., in systems)
that is complex, situated, and adaptive
Helps to build robust explanatory conclusions of
complex, systemic change
7. Develop criteria for designing and evaluating
these systemic interventions
Helps to account for objectives targeted by systemic
interventions
Helps to appropriately frame and assess the efcacy of
adaptive interventions for various levels of systems
(e.g., individual, group, or organizational levels)
8. Adopt more integrative designs Allows researchers to integrate exploratory
falsicatory aims, individual group analyses, and
qualitative quantitative methods
Helps drive ongoing methodological innovation
9. Become comfortable with a more problem-based,
transdisciplinary orientation
Helps to avoid rigid, paradigm-driven research
Allows CDST research to address issues in socially
useful and participant-relevant ways
Allows researchers to work in transdisciplinary ways
and teams
936 Phil Hiver, Ali H. Al-Hoorie, and Reid Evans
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
natural part of its growth and more mainstream acceptance of this meta-theory. Yet, as is
also apparent, methodological advances and applications now exist that point the way
forward for the eldparticularly those allowing researchers to tap into the system of
within-person dynamics and draw inferences about the underlying patterns of language
development (e.g., Kliesch & Pfenninger, 2021; Murakami, 2016,2020; Pfenninger,
2020; Verspoor et al., 2021). We acknowledge that the insights and guidelines CDST
offers can be overwhelming, and this can slow the progress of the eld. We have therefore
synthesized the methodological lessons we obtained in this review, and refer to them here,
as the nine tenetsof CDST research. Table 9 presents these tenets and the purpose of
each.
We might think of CDST research in the eld as now being at a crossroads. As CDST
research assesses how far it has come, with one eye to the future, it is important not to
simply scrutinize and critique without also offering alternatives. We hope to have done
both in this paper, and our results have shown that there is robust empirical evidence as
well as ample methodological guidance on which future work can build. We hope that
future CDST research will draw on these lessons and continue to offer substantive insights
to the eld of language learning and development.
COMPETING INTERESTS
At the time this paper was initially submitted for review, Ali Al-Hoorie had not yet taken
up duties on the SSLA editorial board.
SUPPLEMENTARY MATERIALS
To view supplementary material for this article, please visit http://doi.org/10.1017/
S0272263121000553.
NOTES
1
We use this term to mean a frame of reference for thinking that provides guiding notions for methods of
scientic inquiry.
2
While many group-based designs are also cross-sectional, these two terms should not be conated. Cross-
sectional research designs examine a sample of individuals at a particular point in time, and whereas they do not
seek to establish temporal sequence, they may investigate changes in focal variables (e.g., by taking synchronic
measurements in groups with different lengths of exposure). Group-based designs need not be cross-sectional in
nature; they may be longitudinal.
3
While this was necessary for obvious reasons, the more than 70 conceptual articles are additional
testament to the robustness of the eld.
4
As one reviewer pointed out, prior to the fairly recent adoption of the term CDST,the eld used CTor
DSTand even chaos theory,though not always as entirely interchangeable concepts.
5
Of these 74 studies, 25 were from the article subpool and 49 were from the dissertations subpool.
6
An example from a parallel eld might be psychotherapy in which the content of each consultation is
tailored to the individual needs of patients, where each client responds in different ways to treatment, and the
treatment is adapted as the program of consultations unfolds.
Complex Dynamic Systems Theory in Language Learning 937
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
REFERENCES
Al-Hoorie, A. H., Hiver, P., Kim, T.-Y., & De Costa, P. I. (2021). The identity crisis in language motivation
research. Journal of Language and Social Psychology,40,118.
Al-Hoorie, A. H., & Vitta, J. P. (2019). The seven sins of L2 research: A review of 30 journalsstatistical quality
and their CiteScore, SJR, SNIP, JCR Impact Factors. Language Teaching Research,23, 727744.
Alexander, P. (2020). Methodological guidance paper: The art and science of quality systematic reviews. Review
of Educational Research,90,623.
Andringa, S., & Godfroid, A. (2020). Sampling bias and the problem of generalizability in applied linguistics.
Annual Review of Applied Linguistics,40, 134142.
Arksey, H., & OMalley, L. (2005). Scoping studies: Towards a methodological framework. International
Journal of Social Research Methodology,8,1932.
Bastardas-Boada, A. (2013). Language policy and planning as an interdisciplinary eld: Towards a complexity
approach. Current Issues in Language Planning,14, 363381.
Blommaert, J. (2014). From mobility to complexity in sociolinguistic theory and method. Tilburg Papers in
Culture Studies,103,124.
Brown, A. V., Plonsky, L., & Teimouri, Y. (2018). The use of course grades as metrics in L2 Research: A
systematic review. Foreign Language Annals,51, 763778.
Brutus, S., Gill, H., & Duniewicz, K. (2010). State-of-science in industrial and organizational psychology: A
review of self-reported limitations. Personnel Psychology,63, 907936.
Byrne, D., & Callaghan, G. (2014). Complexity theory and the social sciences: The state of the art. Routledge.
Byrne, D., & Ragin, C. C. (Eds.) (2009). The SAGE handbook of case-based methods. SAGE.
Cowley, S. J. (Ed.) (2011). Distributed language. John Benjamins.
de Bot, K. (2008). Introduction: Second language development as a dynamic process. The Modern Language
Journal,92, 166178.
Dewaele, J.-M. (2019). The vital need for ontological, epistemological and methodological diversity in applied
linguistics. In C. Wright, L. Harvey, & J. Simpson (Eds.), Voices and practices in applied linguistics:
Diversifying a discipline (pp. 7188). White Rose University Press.
Ellis, N. C., & Larsen-Freeman, D. (Eds.) (2009). Language as a complex adaptive system. Wiley-Blackwell.
Eskildsen, S. (2009). Constructing another languageUsage-based linguistics in second language acquisition.
Applied Linguistics,30, 335357.
Evans, R. (2019). Bifurcations, fractals, and non-linearity in second language development: A complex dynamic
systems perspective (Unpublished doctoral dissertation). University at Buffalo, New York.
Gass, S., Loewen, S., & Plonsky, L. (2021). Coming of age: the past, present, and future of quantitative SLA
research. Language Teaching,54, 245258.
Han, Z.-H. (2020). Usage-based instruction, systems thinking, and the role of Language Mining in second
language development. Language Teaching. Advance online publication. https://doi.org/10.1017/
S0261444820000282
Han, Z.-H. (Ed.) (2019). Proling learner language as a dynamic system. Multilingual Matters.
Herdina, P., & Jessner, U. (2002). A dynamic model of multilingualism. Multilingual Matters.
Hilpert, J., & Marchand, G. (2018). Complex systems research in educational psychology: Aligning theory and
method. Educational Psychologist,53, 185202.
Hiver, P., & Al-Hoorie, A. H. (2016). A dynamic ensemble for second language research: Putting complexity
theory into practice. The Modern Language Journal,100, 741756.
Hiver, P., & Al-Hoorie, A. H. (2020a). Reexamining the role of vision in second language motivation: A
preregistered conceptual replication of You, Dörnyei, and Csizér (2016). Language Learning,70,48102.
Hiver, P., & Al-Hoorie, A. H. (2020b). Research methods for complexity theory in applied linguistics.
Multilingual Matters.
Hiver, P., Al-Hoorie, A. H., & Larsen-Freeman, D. (2021). Toward a transdisciplinary integration of research
purposes and methods for Complex Dynamic Systems Theory: Beyond the quantitativequalitative divide.
International Review of Applied Linguistics in Language Teaching. Advance online publication. https://
doi.org/10.1515/iral-2021-0022.
Hu, Y., & Plonsky, L. (2021). Statistical assumptions in L2 research: A systematic review. Second Language
Research,37, 171184.
938 Phil Hiver, Ali H. Al-Hoorie, and Reid Evans
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
Hulstijn, J. (2020). Proximate and ultimate explanations of individual differences in language use and language
acquisition. Dutch Journal of Applied Linguistics,9,2137.
Hult, F. (2010). The complexity turn in educational linguistics. Language, Culture and Curriculum,23,
173177.
Kang, E. Y., Sok, S., & Han, Z. (2018). Thirty-ve years of ISLA on form-focused instruction: A meta-analysis.
Language Teaching Research,23, 428453.
Kaplan, A., Cromley, J., Perez, T., Dai, T., Mara, K., & Balsai, M. (2020). The role of context in educational
RCT ndings: A call to redene evidence-based practice.Educational Researcher,49, 285288.
Ke, J., & Holland, J. H. (2006). Language origin from an emergentist perspective. Applied Linguistics,27,
691716.
Kliesch, M., Pfenninger, S. E. (2021). Cognitive and socio-affective predictors of L2 micro-development in late
adulthood: A longitudinal intervention study. The Modern Language Journal,105, 237266.
Koopmans, M. (2020). Education is a complex dynamical system: Challenges for research. The Journal of
Experimental Education,88, 358374.
Kramsch, C., & Whiteside, A. (2008). Language ecology in multilingual settings: Towards a theory of symbolic
competence. Applied Linguistics,29, 645671.
Kretzschmar, W. (2015). Language and complex systems. Cambridge University Press.
Larsen-Freeman, D. (1994, October). On the parallels between chaos theory and second language acquisition
[Paper presentation]. Second Language Research Forum, McGill University, Montreal.
Larsen-Freeman, D. (2007). On the complementarity of chaos/complexity theory and dynamic systems theory in
understanding the second language acquisition process. Bilingualism: Language and Cognition,10,3537.
Larsen-Freeman, D. (2015). Ten lessonsfrom complex dynamic systems theory: What is on offer. In Z.
Dörnyei, P.D. MacIntyre, & A. Henry (Eds.), Motivational dynamics in language learning (pp. 1119).
Multilingual Matters.
Larsen-Freeman, D. (2016a). Classroom-oriented research from a complex systems perspective. Studies in
Second Language Learning and Teaching,6, 377393.
Larsen-Freeman, D. (2016b). Thoughts on the launching of a new journal: A complex dynamic systems
perspective. Journal for the Psychology of Language Learning,1,6782.
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 (pp. 1150). John Benjamins.
Larsen-Freeman, D. (2018). Resonances: Second language development and language planning and policy from
a complexity theory perspective. In F. Hult, T. Kupisch, & M. Siiner (Eds.), Bridging language acquisition
and language policy (pp. 203217). Springer.
Larsen-Freeman, D. (2020). Complexity theory: Relational systems in interaction and in interlocutor differences
in second language development. In L. Gurzynski-Weiss (Ed.), Cross-theoretical explorations of interloc-
utors and their individual differences (pp. 189208). John Benjamins.
Larsen-Freeman, D., & Cameron, L. (2008). Complex systems and applied linguistics. Oxford University Press.
Larson-Hall, J., & Plonsky, L. (2015). Reporting and interpreting quantitative research ndings: What gets
reported and recommendations for the eld. Language Learning,65, 127159.
Levac, D., Colquhoun, H., & OBrien, K. K. (2010). Scoping studies: Advancing the methodology. Imple-
mentation Science,5, 69.
Levine, G. S. (2020). A human ecological language pedagogy [Monograph issue]. The Modern Language
Journal,104,1130.
Lowie, W. (2017). Lost in state space? Methodological considerations in complex dynamic theory approaches to
second language development research. In L. Ortega & Z. Han (Eds.), Complexity theory and language
development: In celebration of Diane Larsen-Freeman (pp. 123141). John Benjamins.
Lowie, W., & Verspoor, M. (2019). Individual differences and the ergodicity problem. Language Learning,69,
184206.
Lowie, W., Verspoor, M., & de Bot, K. (2010). A dynamic view of second language development across the
lifespan. In K. de Bot & R.W. Schrauf (Eds.), Language development over the lifespan (pp. 125146).
Routledge.
MacIntyre, P. D., MacKay, E., Ross, J., & Abel, E. (2017). The emerging need for methods appropriate to study
dynamic systems: Individual differences in motivational dynamics. In L. Ortega & Z. Han (Eds.), Complexity
theory and language development: In celebration of Diane Larsen-Freeman (pp. 97122). John Benjamins.
Complex Dynamic Systems Theory in Language Learning 939
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
Marsden, E., Thompson, S., & Plonsky, L. (2018). A methodological synthesis of self-paced reading in second
language research. Applied Psycholinguistics,39, 861904.
Massip-Bonet, À., Bel-Enguix, G., & Bastardas-Boada, A. (Eds.) (2019). Complexity applications in language
and communication sciences. Springer.
McHugh, M. L. (2012). Interrater reliability: The kappa statistic. Biochemia Medica,22, 276282.
Molenaar, P. C. (2015). On the relation between person-oriented and subject-specic approaches. Journal for
Person-Oriented Research,1,3441.
Molenaar, P. C., & Campbell, C. (2009). The new person-specic paradigm in psychology. Current Directions
in Psychological Science,18, 112117.
Mufwene, S. S., Coupé, C., & Pellegrino, F. (Eds.) (2017). Complexity in language: Developmental and
evolutionary perspectives. Cambridge University Press.
Munn, Z., Peters, M. D. J., Stern, C., Tufanaru, C., McArthur, A., & Aromataris, E. (2018). Systematic review or
scoping review? Guidance for authors when choosing between a systematic or scoping review approach.
BMC Medical Research Methodology,18, 143.
Murakami, A. (2016). Modeling systematicity and individuality in nonlinear second language development:
The case of English grammatical morphemes. Language Learning,66, 834871.
Murakami, A. (2020). On the sample size required to identify the longitudinal L2 development of complexity
and accuracy indices. In W. Lowie, M. Michel, A. Rousse-Malpat, M. Keijzer, & R. Steinkrauss (Eds.),
Usage-based dynamics in second language development (pp. 2049). Multilingual Matters.
Nicklin, C., & Plonsky, L. (2020). Outliers in L2 research: A synthesis and data re-analysis from self-paced
reading. Annual Review of Applied Linguistics,40,2655.
Osberg, D., & Biesta, G. (2010). The end/s of school: Complexity and the conundrum of the inclusive
educational curriculum. International Journal of Inclusive Education,14, 593607.
Overton, W. F. (2007). A coherent metatheory for dynamic systems: Relational organicism-contextualism.
Human Development,50, 154159.
Overton, W. F., & Lerner, R. M. (2014). Fundamental concepts and methods in developmental science: A
relational perspective. Research in Human Development,11,6373.
Overton, W. F. (2015). Taking conceptual analyses seriously. Research in Human Development,12, 163171.
Page, M. J., McKenzie, J. E., Bossuyt, P. M., Boutron, I., Hoffmann, T. C., Mulrow, C. D., et al. (2021). The
PRISMA 2020 statement: An updated guideline for reporting systematic reviews. The British Medical
Journal,372, n71. https://doi.org/10.1136/bmj.n71.
Paquot, M., & Plonsky, L. (2017). Quantitative research methods and study quality in learner corpus research.
International Journal of Learner Corpus Research,3,6194.
Petticrew, M., & Roberts, H. (2006). Systematic reviews in the social sciences. Blackwell.
Pfenninger, S. E. (2020). The dynamic multicausality of age of rst bilingual language exposure: Evidence from
a longitudinal CLIL study with dense time serial measurement. The Modern Language Journal,104,
662686.
Pham, M. T., Rajić, A., Greig, J. D., Sargeant, J. M., Papadopoulos, A., & McEwen, S. A. (2014). A scoping
review of scoping reviews: Advancing the approach and enhancing the consistency. Research Synthesis
Methods,5, 371385.
Plonsky, L. (2013). Study quality in SLA: An assessment of designs, analyses, and reporting practices in
quantitative L2 research. Studies in Second Language Acquisition,35, 655687.
Plonsky, L. (2014). Study quality in quantitative L2 research (19902010): A methodological synthesis and call
for reform. The Modern Language Journal,98, 450470.
Rothstein, H. R., & Hopewell, S. (2009). Grey literature. In H. Cooper, L. V. Hedges, & J. C. Valentine (Eds.),
The handbook of research synthesis and meta-analysis (2nd ed., pp. 103125). SAGE.
Schmid, M.S., Köpke, B., & de Bot, K. (2013). Language attrition as complex, non-linear development.
International Journal of Bilingualism,17, 675683.
Siddaway, A. P., Wood, A. M., Hedges, L. V. (2019). How to do a systematic review: A best practiceguide for
conducting and reporting narrative reviews, meta-analyses, and meta-syntheses. Annual Review of Psychol-
ogy,70,124.
Steenbeek, H., & van Geert, P. (2015). A complexity approach toward mindbraineducation (MBE):
Challenges and opportunities in educational intervention and research. Mind, Brain, and Education,9,
8186.
940 Phil Hiver, Ali H. Al-Hoorie, and Reid Evans
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
Ushioda, E. (2021). Doing complexity research in the language classroom: A commentary. In R. Sampson & R.
Pinner (Eds.), Complexity perspectives on researching language learner and teacher psychology
(pp. 245256). Multilingual Matters.
van Geert, P., & Steenbeek, H. (2014). The good, the bad and the ugly? The dynamic interplay between
educational practice, policy and research. Complicity: An International Journal of Complexity and Educa-
tion,11,2239.
van Lier, L. (2004). The ecology and semiotics of language learning: A sociocultural perspective. Kluwer.
Van Orden, G., Holden, J. G., & Turvey, M. T. (2003). Self-organization of cognitive performance. Journal of
Experimental Psychology: General,132, 331350.
Verspoor, M., & de Bot, K. (2021). Measures of variability in transitional phases in second language
development. International Review of Applied Linguistics in Language Teaching. Advance online publica-
tion. https://doi.org/10.1515/iral-2021-0026.
Verspoor, M., de Bot, K., & Lowie, W. (Eds.) (2011). A dynamic approach to second language development:
Methods and techniques. John Benjamins.
Verspoor, M., Lowie, W., & van Dijk, M. (2008). Variability in L2 development from a dynamic systems
perspective. The Modern Language Journal,92, 214231.
Verspoor, M., Lowie, W., & Wieling, M. (2021). L2 developmental measures from a dynamic perspective. In
B. Le Bruyn & M. Paquot (Eds.), Learner corpus research meets second language acquisition
(pp. 172190). Cambridge University Press.
Vitta, J. P., & Al-Hoorie, A. H. (2020). The ipped classroom in second language learning: A meta-analysis.
Language Teaching Research. Advance online publication. https://doi.org/10.1177/1362168820981403.
Vitta, J. P., & Al-Hoorie, A. H. (2021). Measurement and sampling recommendations for L2 ipped learning
experiments: A bottom-up methodological synthesis. The Journal of Asia TEFL,18, 682692.
Complex Dynamic Systems Theory in Language Learning 941
https://doi.org/10.1017/S0272263121000553 Published online by Cambridge University Press
... An intelligent algorithm not only adapts the student's language skills to a specific language norm but also adapts itself to the student's characteristics, speeding up learning and contributing to its individualisation (Kindermans et al. 2014). There is a significant body of research focused on developing more dynamic approaches to language learning as a transfer of language systems and language skills in specific environments (Hiver et al. 2021). These approaches are especially important when studying the Chinese language, the structure of which differs sharply from most of the widespread Indo-European languages. ...
Article
This study aims to study the comparative results of the application of the classical academic and modern dynamic types of studying Chinese writing in its pragmatic aspect. Chinese language and writing are among the most difficult to learn as a second language, requiring advanced teaching methods with wider use of elements of the digital environment and the student’s independent active use of web services for active language learning. This aspect is also important when using more and more mass distance learning. The study involved 90 students aged 23 years old to 35 years old, who were equally divided into experimental and control groups. The experimental group used active learning methods and the control group used the usual academic methods of teaching Chinese writing, in accordance with the university course. The training was allocated four hours per week for remote training (using the Google Meet platform). The results demonstrated that the dynamic approach is effective in teaching hieroglyphics and pragmatic aspects of the language (reading, typing on a computer, listening comprehension, recognising words and written hieroglyphs). The academic classroom method is more suitable for the formation of correct pronunciation, taking into account the tonal type of Chinese phonetics.
... These studies need not be narrowly experimental (i.e., a linear pre-post design) in the conventional sense, as there are multiple appropriate designs that can be used to study the experimental effects of educational interventions in ways that incorporate important initial conditions and emergent influences (e.g., teachers' practices) on student engagement and study their reciprocal effects. Some experimental methods that are highly congruent with a complex dynamic framing of language learning include case-based research methods, design-based intervention research, experimental ethnography, and single-case designs (Hiver, Al-Hoorie, & Evans, 2021). ...
Full-text available
Article
In this study we investigated how student engagement and disengagement change over the course of a semester in the L2 classroom. We modeled change at the inter- and intra-individual levels, using time-variant predictors to examine differences in student classroom engagement and disengagement trajectories. In addition to these temporal dynamics, we also examined what motivational antecedents are related to these changes in engagement and disengagement over time. We collected data from 686 students enrolled in general-purpose English courses at two publicly funded universities in mainland China at three waves in a 17-week semester, and tested a series of multi-level, mixed-effects growth models. Our analyses showed that students who reported higher initial classroom engagement or disengagement levels had lower growth rates than their counterparts as the semester proceeded. Students’ classroom engagement in language learning dipped to its lowest point around the middle of the semester and peaked toward the end of the semester. Motivational antecedents were also strong predictors of student engagement and disengagement in the language classroom at both within- and between-person levels. We discuss the implications of these temporal dynamics of learner engagement in the language classroom.
... Thus, a longitudinal perspective on individual differences necessitates examining individual differences on multiple timescales and over multiple developmental windows to identify the variables that regulate learning. Fortunately, researchers are becoming increasingly aware of the diverse longitudinal designs that can be brought to bear to understand L2 development at various levels of granularity ( Nagle, 2021 b), which is due in part to the growing conceptual and methodological impact that complex dynamic systems theory is having on the field (e.g., Hiver et al., 2021 ). On one end of the spectrum are measurement-intensive longitudinal designs, also known as time series analyses or "shortitudinal " research, where the goal is to collect data very frequently over a shorter window to understand the dynamics of the variable -or system -being studied. ...
Article
There is more interest in individual differences than ever before. Researchers continue to expand the conceptual basis of individual difference work, introducing and validating new variables and models. Methodologically, signs of disciplinary growth also abound, especially related to how individual differences can be measured from a dynamic, time-sensitive perspective. It therefore comes as no surprise that several individual differences have been studied longitudinally as both dependent and independent variables to understand how they change over time and the extent to which they predict the development of a range of language learning behaviors and skills. Yet, longitudinal measurement is an often overlooked area of individual difference research. Traditionally, even in recent longitudinal studies, researchers have focused on individual differences as time-invariant, or between-subjects, predictors of learning, even though most individual differences are time-varying and would be most fruitfully studied as such. Moreover, most individual differences that are the basis of current (longitudinal) research-motivation, anxiety, etc.-are latent factors that are assessed using self-report surveys. As a result, if researchers aim to make claims about how individuals change on these latent factors and include them in complex, multivariate models of behavioral and linguistic outcomes, then they must ensure that the measurement model holds over time, an issue known as longitudinal measurement invariance. In this paper, I present a framework for engaging in longitudinal individual differences research and offer a tutorial on formulating and evaluating measurement invariance models in R.
... Some of these include how to operationalize the system, how to assess the influence of contextual factors on the system, as well as macro-and micro-structure considerations . Given the inherent complexities of analyzing dynamic cause-effect relationships between systems and their components, there has been much discussion about suitable methodologies and suggestions of how to enhance our CDST toolbox (de Bot, 2011;, 2020aHiver et al., 2022). Hilpert and Marchand (2018) distinguish between three conceptual perspectives to studying complex systems and their accompanying research designs: timeintensive, relation-intensive, and time-relation intensive approaches. ...
Full-text available
Article
Network analysis is a method used to explore the structural relationships between people or organizations, and more recently between psychological constructs. Network analysis is a novel technique that can be used to model psychological constructs that influence language learning as complex systems, with longitudinal data, or cross-sectional data. The majority of complex dynamic systems theory (CDST) research in the field of second language acquisition (SLA) to date has been time-intensive, with a focus on analyzing intraindividual variation with dense longitudinal data collection. The question of how to model systems from a structural perspective using relation-intensive methods is an underexplored dimension of CDST research in applied linguistics. To expand our research agenda, we highlight the potential that psychological networks have for studying individual differences in language learning. We provide two empirical examples of network models using cross-sectional datasets that are publicly available online. We believe that this methodology can complement time-intensive approaches and that it has the potential to contribute to the development of new dimensions of CDST research in applied linguistics.
... Another important idea that is front and center in recent language learning research is that context shapes development and learning outcomes (Hiver, Al-Hoorie, & Evans, 2021). It is widely accepted that language learning and interaction are embedded in social-cultural contexts and that language use fulfills important social-interactive functions (Kramsch, 2008;van Lier, 2004). ...
Full-text available
Chapter
Task-based approaches to L2 instruction have become de rigueur in many learning contexts, and learners routinely encounter tasks in the course of regular L2 instruction. The reality of many instructed L2 contexts is that the same task or sequence of tasks can provoke varying responses when presented to students within the same group or classroom. Engagement is a useful lens for L2 researchers seeking to understand how and why individuals focus on, interact within, and learn from tasks. Task engagement can vary across students who are doing the same task, even if that task is highly stimulating. In addition, there may be important differences in how individual engagement manifests among students who have the same overarching level of engagement; these differences have implications for L2 learning and for researching tasks. This chapter is divided into three parts. In the first part, we define task engagement and provide a brief overview of existing work on the topic. As our review shows, task engagement represents the level and quality of a learner’s integrated mental and physical activity, as well as their affective experience, within a task. In the second part, we compare task engagement with task motivation, another framework for looking at students’ involvement in TBLT. We emphasize that task motivation can be thought of as either a precursor of task engagement or as the by-product of engaging in a task. We end our chapter by suggesting ideas for task engagement research that treats individuals’ task engagement as a holistic, situated, adaptive, and momentary phenomenon. Our position is that confusion in understanding task engagement may arise when macro-level information (i.e., general engagement tendencies in a collective of learners across a course of task-based language learning) is used to capture micro-level insights about the time (momentary), task (an individual task), and agent (the individual learner). In response, we propose ways to reconfigure the unit of analysis and the level of granularity at which task engagement is conceptualized, observed, and measured.
... We began the investigation with the LGCM analysis, which showed that learners' motivation did not significantly change over time, whereas achievement did. The insignificant changes observed in the growth of students' motivation could be interpreted in light of the idea of sensitive dependence on initial conditions that comes from the propositions of the Complex Dynamic Systems Theory as it views second language acquisition as a dynamic, complex, non-linear process that is sensitive to initial conditions and that the changes a construct may undergo are sensitive on the initial conditions of this construct (Hiver et al. 2021). In light of this theory, it seems that the insignificant growth in student motivation in our case could be attributed to the fact that the trajectory of system behaviour tends to be dependent on the initial condition of the system in that the high initial endorsement of motivation exhibited at the beginning of the semester (wave 1) did not allow much change to take place at subsequent stages (i.e. ...
Full-text available
Article
It is well known that successful second language (L2) learners are motivated individuals. Accordingly, L2 researchers have tested the predictive power of different motivational constructs on language learning outcomes. However, this perspective appears to neglect the assessment of achievement as a predictor of future motivation. To assess this possibility, we first employed the latent growth curve model (LGCM) to evaluate the initial values and growth rates of the two variables. We further applied a newly developed statistical method, the random-intercept cross-lagged panel model (RI-CLPM), to study the causal relationship. A total of 226 language students were monitored for 17 weeks at three time points. The analysis showed an increasing trend in the association between the growth levels of both variables. However, students' autonomous motivation at Time 1 appears to affect achievement at Time 2. Further, the second wave of the RI-CLPM illustrated that achievement at Time 2 impacted autonomous motivation at Time 3, while motivation failed to predict scores on achievement at Time 3.
Article
Complex Dynamic Systems Theory (CDST), an instantiation in applied linguistics of complexity epistemology that transcends disciplinary boundaries, has gained much traction and momentum over the last decade, finding expressions in a fast-growing number of empirical second language developmental studies. However, the literature, while rapidly expanding, has displayed much confusion, notably oscillating between invoking CDST as a metatheory and as an object theory. Then, too, the metaphorical genesis of CDST—the metaphorical adoption of complexity epistemology from physical sciences—has seemed to invite miscellaneous interpretations, rendering CDST an ostensibly all-in-one conceptual prism. This article explores the epistemology of CDST, tracing its ontology and examining its role in second language developmental research. This enables a more nuanced understanding of CDST, while at once surfacing critical issues and directions for future research, as it moves toward a pluralistic approach to investigating CDST as a potentially unique lens on second language development.
Article
Learner emotions represent sudden, dynamic, and complex adaptations to the language classroom environment. Recent Second Language Acquisition research calls for a more holistic perspective in approaching classroom emotions, one that considers emotional variations between and across learners, and which foregrounds the interconnections among emotions and between emotions and the learning environment. This paper approaches emotions from a complex dynamic systems perspective and investigates the classroom emotions of five university students of Korean as a foreign language using a Q methodology intensive single-case study design. Overall results show that students have sometimes similar, sometimes different emotional reactions depending on classroom events, indicating different levels of interaction between a learner’s emotional system and other individual characteristics. Additionally, a more fine-grained analysis at the level of individual learners reveals clusters of emotions triggered by the same event and foregrounds the relevance of epistemic emotions for instructed foreign language learning. The results are discussed focusing on new hypotheses to inform future SLA emotion research and classroom practices.
Full-text available
Preprint
Self-determination theory is one of the most established motivational theories both within second language learning and beyond. This theory has generated several mini-theories, namely: organismic integration theory, cognitive evaluation theory, basic psychological needs theory, goal contents theory, causality orientations theory, and relationships motivation theory. After providing an up-to-date account of these mini-theories, we present the results of a systematic review of empirical second language research into self-determination theory over a 30-year period (k = 111). Our analysis of studies in this report pool showed that some mini-theories were well-represented while others were underrepresented or absent from the literature. We also examined this report pool to note trends in research design, operationalization, measurement, and application of self-determination theory constructs. Based on our results, we highlight directions for future research in relation to theory and practice.
Full-text available
Article
With the advent of COVID-19, learning has transitioned from the classroom to online platforms (Tang et al., 2020). Interestingly, this forced migration to online teaching has coincided with the rise of the popularity of flipped learning in education in general (Låg & Sæle, 2019) and ELT and L2 in particular (Mehring, 2018). Flipped learning, in a general sense, inverts the traditional learning paradigm by presenting new content to students before and outside of the class with subsequent class time used for interacting and engaging with said content (for extensive consideration of L2 flipped learning; see Mehring, 2018). Unsurprisingly, recent research has investigated the potential of flipped classrooms against the backdrop of online teaching in response to COVID-19. Tang and colleagues (2020), for instance, investigated Chinese students’ perceptions of flipped learning in online environments vis-à-vis traditional methods. It is likewise reasonable to expect L2 researchers to enhance their already substantial interest in flipped learning in response to the current migration to online teaching. What is more is that even after COVID-19 subsides, there is no reason to assume the L2 academic community’s interest in flipped learning will wane as interest in it had been growing before the pandemic (see e.g., Bonyadi, 2018).
Full-text available
Article
The Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) statement, published in 2009, was designed to help systematic reviewers transparently report why the review was done, what the authors did, and what they found. Over the past decade, advances in systematic review methodology and terminology have necessitated an update to the guideline. The PRISMA 2020 statement replaces the 2009 statement and includes new reporting guidance that reflects advances in methods to identify, select, appraise, and synthesise studies. The structure and presentation of the items have been modified to facilitate implementation. In this article, we present the PRISMA 2020 27-item checklist, an expanded checklist that details reporting recommendations for each item, the PRISMA 2020 abstract checklist, and the revised flow diagrams for original and updated reviews.
Full-text available
Article
The question of cognition in second language (L2) acquisition later in life is of importance inasmuch as L2 learning is largely mediated by domain‐general cognitive capacities. While a number of these capacities have been shown to decline with age, individual differences in cognition increase over the lifespan. This microdevelopment study investigates the L2 trajectories of 28 older German‐speaking adults (age 64+) who participated in a combined computer‐assisted and classroom‐instructed 7‐month Spanish training for beginners. We made use of generalized additive mixed models (GAMMs) to quantify linear and nonlinear learner trajectories as well as any predictors thereof. Participants were assessed on a range of behavioral, L2, socioaffective, and background variables. We found a significant (linear and nonlinear) increase across all measures of L2 proficiency. Between‐subject cognitive, socioaffective, and background variables significantly predicted the overall level of L2 proficiency as well as developmental patterns over time. Daily variances in cognitive performance and socioaffect had little impact on fluctuations in L2 performance. Findings are discussed against the backdrop of complex dynamic systems theory and highlight the necessity for dense longitudinal research designs to capture nonlinearity in third‐age L2 learning.