Abstract Engaging a complexity theory view of learning, this study examined an atypical timescale for tracking L2 authorial voice development through the interaction of cognitive processes that inform voice construction. A microgenetic analysis of seven adult Japanese learners of English in a three-week writing course designed to help students develop their authorial voices revealed learning dimensions that were (i) wide in breadth, (ii) isomorphic in their rate, (iii) triggered by repeated tasks in a teaching-and-learning cycle facilitated by stylistic analyses, (iv) variegated across learners, and (v) erratic and nonlinear. Interactions also showed signs of stabilizing during the final phase of the intervention. These findings are consistent with a complexity theory view of L2 development, demonstrating that repeated and similar learning tasks implicate emergentist interpretations of language and literacy development. This article contributes to understanding authorial voice construction across atypical timescales and invites L2 studies to apply timescales of development more relativistically. This study also emphasizes the importance of further exploring microgenetic interactions for understanding the ontogenesis of authorial voice and for conceptualizing its development inside and outside the classroom. Research on L2 authorial voice has repeatedly documented the difficulty of capturing a workable definition of the construct across research contexts (Stock and Eik-Nes 2016). In the present work, voice is considered with respect to a single learning context that emphasizes lexico-rhetorical choices related to stance and engagement. For example, using the modal verb ‘must’ may represent authorial positionality, while employing the first-person ‘we’ may engage the reader. Voice is defined in this study as an amalgam of lexico-rhetorical choices, outlined below, that work to project something of a writer’s identity and positionality through rhetorical features of stance and engagement. Studies on authorial voice have commonly focused on voice as a socially mediated, dialogic representation of self. Studies have investigated how large-scale, macrolevel phenomena such as learners’ negotiations with disciplinary discourses influence voice development over the long term (Hirvela and Belcher 2001; Kim et al. 2006). These studies have emphasized that constructing authorial voice is a highly complex, longitudinal process that implicates individual and social aspects of voice, and increasingly, dialogic aspects of voice in the Bakhtinian tradition (Tardy 2012; Matsuda 2015). In short, these and related studies rightly suggest that authorial voice emerges from a process of enculturation—and often acculturation—that initiates in learners’ personal histories whereby learners interact with language and culture on a macroscale. However, the study of authorial voice (and identity, more broadly) lacks microlevel accounts of the functions and interactions of cognitive activities implicated during these dialogic interactions and, similarly, lacks studies that explore its construction across limited periods of development. Exploring these dimensions is useful for understanding how cognitive functions help facilitate the dialogic process of voice development, for revealing microlevel L2 learning processes, including at micro timescales, and for assisting with pedagogical interventions related to constructing voice in classroom contexts where time constraints are regularly present. The objective of the present study was to outline some of the microlevel dynamics informing cognitive processes associated with authorial voice construction over a short, atypical timescale for tracking voice development. In doing so, this work focused first on describing the dynamics of change in learners’ voice development at the microlevel (rather than predicting performance) and, second, on interindividual learner differences in developmental trajectories. THEORETICAL AND ANALYTIC PERSPECTIVE As macrolevel studies of voice demonstrate, constructing an authorial voice occurs within highly complex spaces that are informed by competing and complementary variables (i.e. individual, social, and dialogic aspects of voice). Taking this cue, this study engaged a complex dynamic systems theory (CDST) view of second language development as a framework for interpreting microlevel changes in the development of learners’ authorial voices. This focus on microlevel change is consistent with CDST research (Hiver and Al-Hoorie 2016) that emphasizes the benefit of exploring a sub-system of interacting variables (here, cognitive processes) for understanding more complex phenomenon (voice as a dialogic assemblage). CDST, outlined below, interprets language learning as a dynamic, nonlinear, and emergent process that arises from individuals interacting with each other and with their communities and, subsequently, individuals interacting with and on language (Larsen-Freeman and Cameron 2008; Ellis 2008). As a result of this complexity, variability is widespread and is congruent with how recent studies on voice have conceptualized its ontogenesis (as an amalgam of affordances and constraints that interact with unique learners in equally unique learning contexts—i.e. learning that is varied and situated). Accordingly, CDST provides a useful perspective as a descriptive tool for revealing the dynamics of developmental processes engaged in dialogic spaces—that is, this study does not compete with present sociocultural orientations to voice but rather, as I later argue, aims to provide a complementary vantage point for surveying voice. To examine how L2 writers develop their authorial voices over a short period of time, microgenesis can provide an analytic lens for scrutinizing how cognitive learning mechanisms are engaged during socially mediated interaction (Lavelli, Pantoja, Hsu, Messinger, and Fogel 2005). Microgenetic analyses have been employed successfully to track instances of development across various research domains including educational psychology (Fischer and Granott 1995), child psychology (Bruner 1983), and second-language acquisition (Gánem-Gutiérrez 2008; Knouzi et al. 2010). Moreover, microgenesis is designed to capture change occurring in short spurts (Vygotsky 1986; Poehner 2008) and facilitates the research of complex systems within limited timeframes. Research on microgenetic processes also typically interprets cognitive activity along five dimensions: path (the sequence of cognitive behavior), source (the cause of the change in behavior), breadth (the transferability of the learning feature), rate (the time invested in the change), and variability (how behavior varies across learners) (Siegler 2006). The present study focuses on these five dimensions. In the L2 context, microgenesis has repeatedly demonstrated its utility for exploring change and development. For example, Knouzi et al. (2010) followed two adult learners of French and tracked the path, breadth, and rate of mental activities that contributed to developing their conceptual understanding of grammatical voice. This analytic tool highlighted how learners iteratively access previous knowledge to build new information (the path of development), differences in intraindividual variability concerning strategy use for conceptual development, (the breadth of change), and meaningful reflective pauses in learners that facilitated learning (the rate of change). Tracking these aspects of microgenetic change underscored qualitative changes in learners’ conceptual development of grammatical voice. Other L2 studies using microgenesis have provided similarly useful findings regarding microlevel cognitive processing. Gánem-Gutiérrez (2008) explored how 18 adolescent learners of Spanish as a foreign language advanced their understanding of personal pronouns and infinitive verbs, among other constructs. Gánem-Gutiérrez showed how a microgenetic account of the path, source, and variability of change revealed phases of development (the path of learning) facilitated by collaborative activity (the source of change) that demonstrated differences (variability) among learners. This analytic tool provided meaningful insights into psychological processes that inform language learning during dyadic interaction. These and other studies employing microgenesis highlight how cognitive processes function as a subset of mental activities to inform learning; as such, microgenesis may be useful for exploring cognitive activities that inform voice development. A CDST VIEW OF DEVELOPMENT Authorial voice development engages a dialogic, sociocultural space rooted in learners’ experiences—language related and otherwise—across the life span. Extensive studies into voice development have explored ‘the person behind the written words’ (Hirvela and Belcher 2001: 85), emphasizing the constraints and affordances that occupy L2 learners in the larger, macrolevel context of their lived experiences. However, as noted at the outset, largely absent from the literature are reports documenting microlevel processes that learners engage during development—that is, the processes whereby learners internalize macrolevel dynamics. To appreciate how insights into these developmental processes can inform theories about voice construction, briefly illustrating how CDST conceptualizes change is useful (see Verspoor et al. (2011), for a thorough rendering). In CDST, a key driver of development is variability. It is a property of a changing system (a unit of study) that exhibits different states of stability and flux. Variability is characterized as meaningful and necessary, constituting a space where learners, among other activities, experiment with learning strategies: where learners ‘explore and select’ (Verspoor et al. 2008: 217). Variability can also function as a ‘harbinger of change’ (Thelen and Smith 1994: 342), where high levels of progress and regress can precede phase transitions—where learning reorganizes itself ‘into a higher-level state’ (Baba and Nitta 2014: 4); that is, where learning becomes internalized. This CDST view of L2 development has been realized across numerous contexts, shedding light on how L2 learning emerges. A partial review of the literature, focusing on writing development, demonstrates how such dynamic system changes emerge in L2 writers and how a CDST perspective on change can contribute to understanding microlevel changes in voice development. In their research into L2 writing development, Larsen-Freeman (2006), Baba and Nitta (2014), and Verspoor and colleagues all reported patterns of development typical of dynamic systems: increased system variability that precedes phase transitions. For example, Larsen-Freeman, after examining linguistic accuracy measures in adult Chinese, ESL learners of English, described how high levels of progress and regress (variability) facilitated the emergence of stable linguistic performance. Similarly, Verspoor et al. (2008) documented in one Dutch learner of English meaningful periods of stability and flux, or increased variability, which preceded a phase transition. In another study, Spoelman and Verspoor (2010) explored 54 writing samples composed over three years from one Dutch university student learning Finnish. One finding from their work documented the development of several measures of accuracy and complexity that were characterized by high levels of variability near phase transitions. Continuing, Verspoor et al. (2012) examined 437 texts composed by 489 Dutch-adolescent, L2 learners of English with the dual aim of establishing objective measures of L2 writing and of advancing awareness of L2 developmental processes. The researchers coded 64 variables across the word, phrase, and sentence level. They noted that seven lexical and syntactic constructions typically associated with different written proficiency levels were present, and that each of these constructions demonstrated development consistent with CDST: dynamic interactions between variables and nonlinear development informed by variability. Still in the Dutch context, Verspoor and Smiskova (2012) compared over two years the development of formulaic sequences, or chunks, in writing samples from two sets of secondary school learners of English. One group received two hours of English input per week; the other group received 15 hours of input per week. Their work did not detect significant differences between the two groups; however, a detailed variability analysis on one learner from each group revealed that the learning trajectory of the high-input learner contained significant variability that meaningfully impacted her chunk development. Most recently, Baba and Nitta investigated the development of writing fluency in two Japanese, EFL university students over one academic year. Their study noted that meaningful changes in fluency were triggered by slight stimulations and that fluency developed alongside phase transitions that were sudden and accompanied by increased variability. These studies all highlight the process of development around phase transitions, noting how meaningful change related to complexity, accuracy, and fluency in L2 writers is preceded by high levels of variability. However, from a CDST perspective, knowledge about writing processes is limited primarily to these three measures. Revealing the mental activities informing how voice emerges can both expand general CDST understandings of L2 writing development and—importantly for this study—underscore how learners construct their writerly voices at the microlevel, thus complementing present understandings of how voice emerges. The prospect of these contributions is especially important given how little is known about CDST’s impact on advancing theories about voice construction and L2 writing development in general. THE PRESENT STUDY The present study examined the development of authorial voice in the writing of English language learners engaged in a learning-to-write for a high-stakes academic writing context and aimed to provide a microgenetic glimpse into the path, source, breadth, rate, and variability regarding voice development across a limited period of study. To trace development, I designed a three-week intensive writing course modeled on private language school course offerings that claim to prepare students for the TOEFL test. Rather than employing an intact classroom, participants were recruited whose purpose was to develop their L2 writing proficiency on the prototypical, five-paragraph argumentative essay that informs the independent writing task of the TOEFL. In doing so, this design is consistent with microgenetic approaches to tracking development that aim to facilitate learning ‘under laboratory conditions to understand how people appropriate and internalize different forms of mediation’ (Hulstijn et al. 2014: 372). Following microgenetic accounts of learning described above, cognitive processes are defined in this study as a set of mental activities informing microgenetic development that are traceable along the five dimensions of change conventionally associated with microgenesis (path, source, breadth, rate, and variability). As noted beforehand, microgenetic studies contribute positively to understanding how cognitive processes function as a subset of mental activities that inform learning. Moreover, to distinguish this work from Vygotskian sociocultural orientations to development, this study takes as given the presence of sociocultural (i.e. dialogic) influences on voice development and instead focuses on ‘the adaptive activity of human mental systems’ (Larsen-Freeman and Cameron 2008: 35). METHOD Participants Seven university-aged L1 Japanese learners of English were recruited for this study from a large Canadian city. Each participant had written the TOEFL Internet-based test previously and the mean, self-reported score of their latest TOEFL tests was 71.9 (SD = 8.8) (only two students could recollect their score on the writing components of the TOEFL, so only overall scores are reported here). The participants were also screened to ensure that they lacked experience with pedagogical stylistics (Widdowson 1975)—expanded on below, and the primary tool employed to mediate authorial voice. Procedure The intervention occurred over 12 days spread over three successive weeks (totaling 16 instructional hours) wherein I taught the course and conducted the research. During this time, the study tracked the development of learners’ authorial voice through classroom activities and audio recordings of group discussions, post-intervention interviews conducted in English, and through written responses to the independent writing task of the TOEFL. The pedagogy informing the intervention focused on voice as an amalgam of the uses of hedges, boosters, attitude markers, authorial self-mentions, and direct reader-references. Table 1 provides examples of each rhetorical feature and, borrowing from Hyland (2008) and Zhao (2013), describes their rhetorical function. Table 1: Examples and function of lexical items coded as a rhetorical feature of authorial voice Rhetorical Feature Examples Function Hedge perhaps, maybe, tend to Enable writers to distance themselves from a claim. Booster definitely, indeed, absolutely, must Enable writers to express certainty about a position. Attitude marker disbelief, concurrence, dissatisfaction Convey the writer’s attitude toward a claim. Authorial self-mention I, me, my Enable writers to engage the reader. Direct reader reference we, us, you, our Enable writers to engage the reader. Rhetorical Feature Examples Function Hedge perhaps, maybe, tend to Enable writers to distance themselves from a claim. Booster definitely, indeed, absolutely, must Enable writers to express certainty about a position. Attitude marker disbelief, concurrence, dissatisfaction Convey the writer’s attitude toward a claim. Authorial self-mention I, me, my Enable writers to engage the reader. Direct reader reference we, us, you, our Enable writers to engage the reader. Table 1: Examples and function of lexical items coded as a rhetorical feature of authorial voice Rhetorical Feature Examples Function Hedge perhaps, maybe, tend to Enable writers to distance themselves from a claim. Booster definitely, indeed, absolutely, must Enable writers to express certainty about a position. Attitude marker disbelief, concurrence, dissatisfaction Convey the writer’s attitude toward a claim. Authorial self-mention I, me, my Enable writers to engage the reader. Direct reader reference we, us, you, our Enable writers to engage the reader. Rhetorical Feature Examples Function Hedge perhaps, maybe, tend to Enable writers to distance themselves from a claim. Booster definitely, indeed, absolutely, must Enable writers to express certainty about a position. Attitude marker disbelief, concurrence, dissatisfaction Convey the writer’s attitude toward a claim. Authorial self-mention I, me, my Enable writers to engage the reader. Direct reader reference we, us, you, our Enable writers to engage the reader. Operationalizing voice in this manner was based on Zhao’s reformulation of Hyland’s interactionist model of voice. Zhao reformulated Hyland’s model to match the writing demands of the TOEFL independent writing task. After examining authorial voice in 480 ETS (Educational Testing Service)-provided writing samples by subjecting them to statistical analyses, Zhao concluded that the independent writing task of the TOEFL had constraints (time and topic, among them) that ruled out rhetorical features (directives, questions, knowledge references, and asides) typically associated with authorial voice in other genres. Given that the present work also uses the independent writing task of the TOEFL as one measure of voice, it was deemed useful for the purposes of this study to employ Zhao’s reformulated version of Hyland’s model, now adapted to the TOEFL writing context. Operationalizing voice in this manner reveals two concerns. First, this underscores tension between prescriptive accounts of voice atypical of recent literature on voice and sociocultural attempts to understand authorial voice construction. However, the situated context of the intervention (TOEFL test preparation) and the learning goals of the participants (improving their scores on the independent writing task of the TOEFL) needed to be balanced with this narrowly construed treatment of voice. Second, the decision to employ Zhao’s (2013) framework also raises concerns about excluding rhetorical features (e.g., directives and questions, among others noted above) that appeared infrequently in her corpus and that nonetheless may be employed by learners in this study. Limiting the construct of voice in this way and in this context follows ‘operational closure’ (Luhmann 1989), an inherent limitation in CDST research wherein researchers must consider ‘the specific conditions in a specific context at a specific time’ (Cilliers 2001: 8). Such considerations are a necessary condition of CDST research, allowing researchers to pragmatically manage the system under study. Accordingly, given the similarities between Zhao’s work and this study (both explore voice in the TOEFL writing context), Zhao’s research provides some justification for limiting the construct of voice to the rhetorical features noted in Table 1: this treatment is consistent with recent research on authorial voice in the TOEFL independent writing task, as Zhao’s work demonstrates, and this study’s theoretical perspective. The teaching method was based on Galperin’s interpretation of concept-based learning (Galperin 1969), a mode of operationalizing Vygotsky’s theory of spontaneous and scientific concepts (Vygotsky 1986; see Fogal 2017 for a description of concept-based learning). As part of the initial three days of the intervention, I explicitly outlined the aim of the class and simultaneously introduced students to pedagogical stylistics—an analytic technique for examining text-based discourse features (in this case, literary texts)—as the primary mediating tool for developing authorial voice. Pedagogical stylistics—unlike, for example, reader-response style reactions to literary texts that are predicated by macrolevel discourse structures such as theme, character, or symbolism (Vandrick 2003)—foregrounds microlevel structures such as lexis, semantics, or syntax (Verdonk 2013) and begins by assuming that language precedes content (Carter 1997); thus, foregrounding language use and meaning rather than a reader’s personal reaction to a given text. Accordingly, stylistics requires that language learners analyze texts systematically to uncover how language use generates meaning with the aim of assisting language learners in becoming ‘linguistically aware readers who can perceive the qualities of language which are manipulated for particular effects’ (Clark and Zyngier 2003: 342). Stylisticians argue that this approach to literature translates into increased textual understanding, and in some cases with L2 learners, increased language proficiency (Fogal, 2015). While two participants in this study had some experience studying literature at the tertiary level, stylistics was new to them. Given the unique focus of stylistics on language use rather than personal response (the latter representative of how these students approached literature previously), it was deemed that such experiences would impact minimally on the study. As noted, students examined literary texts during the intervention to help improve their authorial voices and writing proficiency in an academic writing context. The potential for a transfer of knowledge about writing across genres (here, from fictional works to argumentative essay writing) is first noted in Carter and Nash (1990), who argue that linguistic features are not the purview of types of genres; that is, metaphors (or, just as readily, rhetorical features of voice) are used successfully in poetry and in argumentative essay writing. Given the flexibility and utility of linguistic features across genres—and the ubiquity of rhetorical features of voice in literary texts and in argumentative essays (Hyland 2008; Zhao 2013)—a larger research project from which this study emerged, hypothesized that knowledge about rhetorical features associated with authorial voice could transfer from studying literature into argumentative writing. The initial three days of the intervention also intended to acculturate students to the task-type (stylistic analyses). This step was devised in part to limit fluctuations in performance associated with learning to do stylistics to this initial three-day period, when data were not collected. This was designed to restrict the amount of variability that could be attributed to participants attending to a new task-type. Following van Dijk et al. (2011), this measure helped to limit overestimating variability. During the remaining nine days of the intervention the participants practised internalizing the target rhetorical features by interacting with various classroom activities. Primary among these activities were daily stylistic analyses of different literary texts (e.g. Bob Dylan’s ‘Oxford Town’, Charles Bukowski’s ‘trashcan lives’, and Margaret Atwood’s ‘Doorway’). A worksheet (Supplementary Data A) adapted from Lin (2010) guided the stylistic analyses and elicited different modes of languaging (Swain 2006) about and with the target features, while other exercises included collaborative peer feedback on the quality of voice in participant essays, opportunities for participants to rework their essays based on peer and instructor feedback, and exercises that asked participants to analyze model essays (individually and in groups) available on the ETS website with the target features in mind. These activities facilitated microgenetic research by providing learners with opportunities to express cognitive processes verbally and nonverbally via interactive activities. This approach allowed me to ‘track and correlate specific moment-to-moment changes to the changes observed in the participants’ cognitive performance’ (Lavelli et al. 2005: 44). Data collection and analysis Nine separate sittings of the TOEFL writing task formed part of the dataset. The first three compositions were collected prior to the intervention to establish a baseline. Four written compositions were collected during the intervention (on Days 4, 6, 8, and 10). The remaining two samples were collected from a posttest (on Day 12) and a delayed posttest (four weeks after the posttest). The essay tasks (Supplementary Data B) featured randomly selected writing prompts designed by ETS (n.d.), presumed from pretesting and other analyses by ETS to be equivalent to each other in content and rhetorical form. To track the dynamics of learners’ authorial voice development, this study examined two variables: changes in the quality of voice in academic writing and evidence of voice development. Regarding the former, three outside raters scored the essays for their overall quality of voice. The raters were certified English language teaching professionals with at least three years of experience scoring essays associated with standardized English proficiency tests. They had all gone through extensive rater training and were unaware of the sequence of the writing tasks and the identity of the individual writers. Scoring was performed using a voice scoring rubric (Fogal 2015) adapted from Zhao (2013). To further explore the dynamics of voice construction, and alongside the outside raters, I looked for indicators of voice development in (i) the written compositions, (ii) the stylistic analysis worksheets, and (iii) audio recordings of participants discussing their stylistic analyses in groups of three or four, recorded daily. To examine interactions across these sets of data, this work adapted a coding and scoring framework (Brooks et al. 2010) for identifying both evidence of development and activities associated with development. See Table 2 for explanations of coding and scoring. Table 2: Coding scheme for tracking development of authorial voice Code Definition Score OIE (outside context, identified, and explained) Participant correctly used the conceptual unit outside of the context in which it was learned. Moreover, he/she correctly identified the conceptual unit on the stylistic worksheet. His/her analysis of the conceptual unit also demonstrated an understanding of its function(s). 5 OINE (outside context, identified, but not explained) Participant correctly used the conceptual unit outside of the context in which it was learned. Moreover, he/she correctly identified the conceptual unit on the stylistic worksheet. However, he/she did not provide an accompanying explanation on the stylistic worksheet or the accompanying explanation did not clearly explain the function of the conceptual unit. 4 OU (outside context, unidentified) Participant correctly used the conceptual unit outside of the context in which it was learned. However, he/she was unable to identify the conceptual unit on the stylistic worksheet. He/she also did not provide an accompanying explanation of the conceptual unit or the accompanying explanation did not clearly explain the function of the conceptual unit. 3 IE (identified and explained) Participant correctly identified the conceptual unit. Moreover, his/her analysis of the conceptual unit demonstrated an understanding of its function(s). However, he/she was unable to employ the conceptual unit outside of the context in which it was learned. 2 INE (identified, but not explained) Participant correctly identified the conceptual unit. However, he/she did not provide an accompanying explanation of the conceptual unit or the accompanying explanation did not clearly explain the function of the conceptual unit. He/she was also unable to employ the conceptual unit outside of the context in which it was learned. 1 U (unidentified) Participant failed to identify the conceptual unit. He/she was also unable to employ the conceptual unit outside of the context in which it was learned 0 Code Definition Score OIE (outside context, identified, and explained) Participant correctly used the conceptual unit outside of the context in which it was learned. Moreover, he/she correctly identified the conceptual unit on the stylistic worksheet. His/her analysis of the conceptual unit also demonstrated an understanding of its function(s). 5 OINE (outside context, identified, but not explained) Participant correctly used the conceptual unit outside of the context in which it was learned. Moreover, he/she correctly identified the conceptual unit on the stylistic worksheet. However, he/she did not provide an accompanying explanation on the stylistic worksheet or the accompanying explanation did not clearly explain the function of the conceptual unit. 4 OU (outside context, unidentified) Participant correctly used the conceptual unit outside of the context in which it was learned. However, he/she was unable to identify the conceptual unit on the stylistic worksheet. He/she also did not provide an accompanying explanation of the conceptual unit or the accompanying explanation did not clearly explain the function of the conceptual unit. 3 IE (identified and explained) Participant correctly identified the conceptual unit. Moreover, his/her analysis of the conceptual unit demonstrated an understanding of its function(s). However, he/she was unable to employ the conceptual unit outside of the context in which it was learned. 2 INE (identified, but not explained) Participant correctly identified the conceptual unit. However, he/she did not provide an accompanying explanation of the conceptual unit or the accompanying explanation did not clearly explain the function of the conceptual unit. He/she was also unable to employ the conceptual unit outside of the context in which it was learned. 1 U (unidentified) Participant failed to identify the conceptual unit. He/she was also unable to employ the conceptual unit outside of the context in which it was learned 0 Table 2: Coding scheme for tracking development of authorial voice Code Definition Score OIE (outside context, identified, and explained) Participant correctly used the conceptual unit outside of the context in which it was learned. Moreover, he/she correctly identified the conceptual unit on the stylistic worksheet. His/her analysis of the conceptual unit also demonstrated an understanding of its function(s). 5 OINE (outside context, identified, but not explained) Participant correctly used the conceptual unit outside of the context in which it was learned. Moreover, he/she correctly identified the conceptual unit on the stylistic worksheet. However, he/she did not provide an accompanying explanation on the stylistic worksheet or the accompanying explanation did not clearly explain the function of the conceptual unit. 4 OU (outside context, unidentified) Participant correctly used the conceptual unit outside of the context in which it was learned. However, he/she was unable to identify the conceptual unit on the stylistic worksheet. He/she also did not provide an accompanying explanation of the conceptual unit or the accompanying explanation did not clearly explain the function of the conceptual unit. 3 IE (identified and explained) Participant correctly identified the conceptual unit. Moreover, his/her analysis of the conceptual unit demonstrated an understanding of its function(s). However, he/she was unable to employ the conceptual unit outside of the context in which it was learned. 2 INE (identified, but not explained) Participant correctly identified the conceptual unit. However, he/she did not provide an accompanying explanation of the conceptual unit or the accompanying explanation did not clearly explain the function of the conceptual unit. He/she was also unable to employ the conceptual unit outside of the context in which it was learned. 1 U (unidentified) Participant failed to identify the conceptual unit. He/she was also unable to employ the conceptual unit outside of the context in which it was learned 0 Code Definition Score OIE (outside context, identified, and explained) Participant correctly used the conceptual unit outside of the context in which it was learned. Moreover, he/she correctly identified the conceptual unit on the stylistic worksheet. His/her analysis of the conceptual unit also demonstrated an understanding of its function(s). 5 OINE (outside context, identified, but not explained) Participant correctly used the conceptual unit outside of the context in which it was learned. Moreover, he/she correctly identified the conceptual unit on the stylistic worksheet. However, he/she did not provide an accompanying explanation on the stylistic worksheet or the accompanying explanation did not clearly explain the function of the conceptual unit. 4 OU (outside context, unidentified) Participant correctly used the conceptual unit outside of the context in which it was learned. However, he/she was unable to identify the conceptual unit on the stylistic worksheet. He/she also did not provide an accompanying explanation of the conceptual unit or the accompanying explanation did not clearly explain the function of the conceptual unit. 3 IE (identified and explained) Participant correctly identified the conceptual unit. Moreover, his/her analysis of the conceptual unit demonstrated an understanding of its function(s). However, he/she was unable to employ the conceptual unit outside of the context in which it was learned. 2 INE (identified, but not explained) Participant correctly identified the conceptual unit. However, he/she did not provide an accompanying explanation of the conceptual unit or the accompanying explanation did not clearly explain the function of the conceptual unit. He/she was also unable to employ the conceptual unit outside of the context in which it was learned. 1 U (unidentified) Participant failed to identify the conceptual unit. He/she was also unable to employ the conceptual unit outside of the context in which it was learned 0 The coding scheme was used as an interpretive guide for tracking states—rather than stages—of development. This framework was employed to code (i) the effectiveness of how the target features were used for each composition—determined by voice scores assigned to each essay—and (ii) the depth of understanding of the function of the target features, recorded in learners’ stylistic worksheets and in the audio recordings. For example, if a learner only identified a target rhetorical feature in the text but did not expand on its function and did not employ the rhetorical feature in a context unique from the text under analysis, then, following the framework outlined in Table 2, the instance of learning was coded as INE (feature Identified, but function Not Explained and feature Not Employed) and was given a score of one. Moreover, the frequency of occurrences of the target rhetorical features in each audio recording did not impact the coding; rather, the coding focused on accuracy. Coding in this manner enabled the tracking of ‘form–meaning–use composites’ (Larsen-Freeman 2011: 53) and the interactions informing their development. Beyond the framework outlined in Table 2, the outside raters also tracked the frequency of the number of target features used by tallying every use of each item in every composition. While not an indicator of the quality of voice (Zhao 2013), tracking usage can be useful as a heuristic for tracing development. Alongside acceptable levels of rater reliability (κ = 0.89), each of these instruments was successfully piloted. Data were also collected from post-intervention stimulated recall and semi-structured interviews, conducted with each participant. Alongside the in-class discussions that were audio-recorded, both sets of verbal data were coded using an open coding framework (Gass and Mackey 2000) adopted from Braun and Clarke (2006) that allowed for a thematic analysis of the data. This entailed exploring the transcripts holistically for evidence of interaction and development. For example, at one stage verbal data highlighted learner interactions concerned with using hedges as a potential marker of a weak authorial voice. This led me to revisit other data (the stylistic worksheets and the essay writing tasks) to examine the dynamics of how learners co-constructed their understanding of hedges at that time and how that was later manifest in their writing. This thematic analysis provided insight into all five dimensions of microgenetic change. Finally, line graphs and moving min-max graphs (van Geert and van Dijk 2002) were employed as visual, descriptive aids (consistent with CDST studies, enabling heuristic analyses of change whereby meaningful patterns may be identified—Byrne (2002)). The line graphs were used to examine individual patterns in voice scores from essay writing tasks and then from voice scores regarding development of authorial voice from learning tasks across the dataset. The min-max graphs were employed to depict general developmental patterns and to help detect temporary changes in the path of voice development. Given the limited data points, a moving window span of three consecutive data points was used to generate the min-max graphs. The frequency of uses of rhetorical features for each learner was also plotted to depict individual usage and developmental tendencies. This line graph visually documented how participants altered their use of the target rhetorical features in responses to TOEFL writing prompts over the course of the study. FINDINGS The path of interactions The individual voice scores displayed in Figure 1 show that the quality of learners’ authorial voices fluctuated over time and was primarily erratic at the early and middle phases of the intervention. Figure 1: View largeDownload slide Individual quality of voice scores on essay compositions (holistic scores) Figure 1: View largeDownload slide Individual quality of voice scores on essay compositions (holistic scores) For example, Figure 1 shows a sharp peak in the quality of Anny’s authorial voice on Day 4 and a drop on Day 6, a deep valley in Angela’s voice score on Day 10, and repeated peaks and valleys with Pi’s scores (Day 4, Day 6, Posttest, and Delayed Posttest). Fluctuations across the dataset were greatest between the baseline and Day 6 and between Day 8 and Day 10. However, homologous trajectories are documented between Day 10 and the delayed posttest. Figure 1 suggests that development was initially erratic while displaying characteristics of stability in the later phases of the study. Except for Angela, the quality of learners’ authorial voices also shows improvement across the dataset. Notwithstanding Angela, Supplementary Data C represents qualitative changes in writing typical across the dataset, and provides evidence of voice development in Maru’s writing from his third baseline and then posttest writing samples. These compositions highlight shifts in Maru’s attempts to engage the reader, with the latter writing sample revealing form–meaning–use composites. Figure 2 documents the frequency of use of the target rhetorical features across participants’ TOEFL writing compositions. Figure 2: View largeDownload slide Frequency of use of the target rhetorical features across participants. Baseline tallies are mean tallies derived from three baseline-writing tasks Figure 2: View largeDownload slide Frequency of use of the target rhetorical features across participants. Baseline tallies are mean tallies derived from three baseline-writing tasks These tallies display erratic usages of the target rhetorical features across the study, with evidence of stability appearing in the final phases of the intervention, similar to Figure 1. Figure 3 provides another perspective on the rate of change. Figure 3: View largeDownload slide Min-max graphs depicting individual development of conceptual understanding of voice (the y-axis indicates voice development scores from 0 to 5; the x-axis indicates the day the data were collected) Figure 3: View largeDownload slide Min-max graphs depicting individual development of conceptual understanding of voice (the y-axis indicates voice development scores from 0 to 5; the x-axis indicates the day the data were collected) The min-max graphs show the breadth of fluctuations in learners’ conceptual understanding of voice, highlighting unique degrees of variation across all learners. For example, the period between Day 8 and Day 10 indicates Kim’s greatest variability and the period between Day 8 and Day 9 indicates similar variability with Angela. With the exception of Pi, all learning stabilizes in the final days of the intervention. Analogous with Figure 1 and Figure 2, the min-max graphs underscore erratic stages in the path of development that stabilize in the latter days of the intervention period. Figures 1, 2, and 3 show erratic and variegated paths of development that stabilize toward the end of the study. Interview data provide insight into this stability: ‘I was getting better at decision-making about my voices [target rhetorical features] near the end [of the study]’ (Maru, post-intervention interview); ‘It was faster for me to use them [rhetorical features]’ (Anne, post-intervention interview); and ‘I could understand better [toward the end of the study] for me how to make a stronger voice’ (Am, post-intervention interview). Moreover, these and similar insights are complemented by the in-class discussions, particularly those comprising the end of the study. For example, at this later stage learners’ discussions focused on the function of the rhetorical features (as opposed to focusing on identifying them in the earlier discussions): Anny: I got a hedge in line four, but I’m not sure…. The highway speed [referencing an image in a poem]? I dunno. Pi: Hmmm. I see it. What about…? But maybe his thinking is slow. Maybe not the cars. Anny: So the uncertainty…. Pi: Right. Maybe. The slow thinking is makes him have caution. Kim: I think that can be it, maybe. The hedge is for his thinking. Anny: That makes sense. The writer goes slow with his thinking like the car moving slow. Is that it? Kim: Yea. I think so. (Day 8, in-class discussion) These students suggest that a hedge describing a vehicle’s velocity is a metaphor for the thought process of the poem’s narrator; that is, the hedge is interpreted as a mode of cautious expression, one that is reflected in the slow-moving vehicle. This excerpt reveals how students focused attention on the function of the hedge (to mimic the thought process of the author) rather than on merely identifying it. Among other samples, this excerpt also demonstrates how learners co-constructed their understanding of the target concept along a recurrent path, where learners provided sample responses, refined their responses, and then attempted again to describe the function of situated, rhetorical features until learners reached a consensus. This behavior is also present in how learners manifest the target features in their writing: I think the introduction and the conclusion I could use booster to say, to mention about my opinion. But little bit I think in body paragraph I couldn’t use booster or hedge or those [other rhetorical features] very well so I think I should have written more example or more detail about me and my idea. (Kim, post-intervention interview) The conceptual development of the target rhetorical features, as Kim’s comments and her written compositions indicate, followed a cyclical course: In writing collected on Day 8 and Day 10, Kim was able to successfully employ the target rhetorical features; during the posttest (on Day 12) she was unable to do so—and struggled to explain why that was the case—despite being cognizant of how such features could benefit her writing; in the delayed posttest (collected on Day 40), Kim was again able to successfully employ these features. Similarly, Pi noted that while writing she would reflect on a literary text examined in class to help her select an appropriate rhetorical feature for her writing; thus, retrieving and then employing previous knowledge during the writing process. In both instances, her behavior displayed a recurrent pattern. These sentiments were consistent across participants. The verbal data and the figures thus far presented, then, are complementary in demonstrating how learners’ paths of learning and cognitive interactions were nonlinear, recurrent, and erratic. The source of interactions I relied on the verbal data to highlight the source of changes in learners’ voice scores and changes in their conceptual understanding of voice. These data suggest that consciousness-raising (particularly, of awareness of the importance of projecting a strong authorial voice) triggered by stylistic analyses of literary texts was the primary driver of change. Six of seven participants noted this, including the following sample commentary: I think it’s [stylistics] a really good idea because they [the analyses] influence me to use hedge and booster and each [feature]. Each author has a habit of writing. So when I saw many types of habit I could see my habit as well. ((Can you explain more?)) In my case there are not enough voices [target rhetorical features] but author use very well those voices [rhetorical features] to say their opinion or feeling so I thought, ‘Oh. I have to steal this skill’. (Maru, post-intervention interview) The verbal data are replete with similar sentiments. Only Angela expressed some doubt about the efficacy of stylistics at triggering changes in her writing, despite increased development in her conceptual understanding of voice (Figure 4, below). When queried to explain her position, Angela asserted that stylistic analyses of academic texts—as opposed to literary texts—would have been more beneficial for her voice development. Moreover, Angela also revealed that she received guidance from an instructor unrelated to this study who suggested she avoid using personal pronouns in her writing. She received this guidance concurrent with the latter half of the intervention. Angela noted that this advice influenced the last three essays she composed for this study, and later reflected that said advice ‘probably effected my voice style in my essays’ (Angela, post-intervention interview). Figure 4: View largeDownload slide Individual voice development scores depicting conceptual understanding of voice Figure 4: View largeDownload slide Individual voice development scores depicting conceptual understanding of voice The breadth of interactions The breadth of learning displayed by the group was wide. Changes in the quality of learners’ authorial voices over the intervention period and the verbal data highlighted that learners, with the exception of Angela, transferred their learning from one context (literary text analysis) to another context (high-stakes essay writing preparation), and engaged the target features across modalities (reading or analyzing, writing, and discussing). Evidence for this change from their writing (Figure 1) and their conceptual understanding of voice (Figure 4, below) is corroborated by interview data. Here I cite two largely representative samples from the interview data: ‘At first I didn’t think we could do this [transfer learning from literary texts to essay writing] but now I am confidence with it’ (Pi, post-intervention interview); and: After first couple of classes I talked with others after class on subway. We were not sure about this kind of learning. ((Can you be more specific, please? What kind of learning?)) Uhm… I mean with poems and those kinds of things. Like, we thought about how can we improve the TOEFL test like this? But after studying like this, I can see how. (Anny, post-intervention interview) As these and other comments indicate, then, learners (notwithstanding Angela) could successfully transfer their understanding of authorial voice from literary text analysis to their academic writing, indicating a wide breadth of learning. The rate of interactions Figure 4 documents changes in how learners conceptualized and constructed their authorial voices during the intervention phase of the study (accordingly, baseline and posttest data points are absent from Figure 4). The data, collected using Table 2, presents development through learners’ interactions with their writing, stylistic analysis worksheets, and peers during in-class discussions. Figure 4 highlights that all leaners improved their conceptual understanding of voice. Moreover, Figure 4 shows a high level of interindividual variation concerning voice development scores from Day 4 until Day 9, when differences between learners then begins to stabilize—highlighting a potential developmental jump around Day 9, but otherwise suggesting that the rate of learning was isomorphic. As noted, the rate of change is similar across learners; verbal data provide some evidence that the rate was also gradual and ongoing. Some of the verbal data reported above note that learners were skeptical of their learning about authorial voice during the initial phases of the study, but that they later came to realize the benefits of the intervention; for Angela, Anne, and Am understanding how to do a stylistic analysis was gradual as well, contributing to a potential gradual rate of change in how they conceptualized and projected an authorial voice. Moreover, Kim suggested that three weeks was insufficient time to internalize the target rhetorical features, suggesting that her development is ongoing. However, there is insufficient quantitative evidence to collaborate these findings. Rather, the data reveal that the rate of change was isomorphic. Variability of interactions Unlike Figure 1, Figure 4 documents less variability regarding how learners’ oriented to the target concept, with the exception of the initial phase of the intervention where variability was widest. However, examined together (Figures 1 and 4), these data indicate some variation between learners’ voice scores in their essays (Figure 1) and their overall voice development scores (Figure 4). For example, Anny and Pi struggled to transfer their conceptual understanding of voice (Figure 4) into an effective voice in their final compositions (Figure 1). In contrast, Maru successfully transferred his conceptual understanding of authorial voice (documented in Figure 4) into his final composition (see Figure 1), demonstrating a strong form–meaning–use composite. Discussion Drawing on the findings, the microgentic analysis of learners studying and constructing their authorial voices suggests that voice development was: (i) nonlinear and erratic in its initial path, though exhibiting stability in the final period of the study, and was initiated primarily through a stylistics-based pedagogy that triggered learners’ awareness of authorial voice and how to develop their authorial voices; (ii) consistent (i.e. participants demonstrated nonlinear learning and participants suggested stylistics was the main factor influencing change); (iii) wide in its breadth (notwithstanding Angela, learners transferred their learning across modes and mediums of expression); and (iv) isomorphic in its rate, with some evidence that it was gradual (and ongoing). These findings document developmental behavior characteristic of CDST that emphasizes individual variability (Larsen-Freeman and Cameron 2008; Verspoor et al. 2008) as well as variability concerning changes in L2 writing (Spoelman and Verspoor 2010; Verspoor and Smiskova 2012; Baba and Nitta 2014) and expands these contexts to include variability in authorial voice development in a learning-to-write preparatory course for high-stakes writing. Changes in learners’ quality of authorial voice and the erratic uses of the target rhetorical features represented by Figure 2, and the subsequent nonlinear path of cognitive behavior also represent changes (and regarding the latter, uses) that are congruent with a usage-based emergentism theory of L2 development that emphasizes development through repeated engagement with target language feature(s) (see Ellis 2008), realized daily in this study by the repetition of stylistic analyses. These factors (changes and uses) highlight patterns of ‘form–meaning–use composites’ (Larsen-Freeman 2011: 53) present in participants’ latter writing samples (see Writing Sample 2 in Supplementary Data C), and corroborate similar findings on emergence in other L2 learning-to-write contexts such as those described in Baba and Nitta (2014) and Eckstein et al. (2011). A unique contribution of this study, however, is the appearance of patterns typical of emergence at the earliest stages of voice construction, and from an ontogenetic perspective contributes to understanding the onset of authorial voice development across limited periods of time and in classroom contexts. Revealing the dynamics of voice development at these early stages is particularly relevant for classroom learning given how little is known about how classroom interactions influence authorial voice construction. In this sense, findings here also provide evidence for the teachability of authorial voice to L2 learners—a topic often debated in the literature (Tardy 2016). However, despite evidence of changes and uses that complement emergent perspectives of L2 development, as the verbal data emphasize learners did not have sufficient contact time with the target features to trigger a full phase transition (van Dijk and van Geert 2007; Baba and Nitta 2014). Rather, the changes in learners’ voice scores (depicted in Figure 1) and their conceptual development of authorial voice (see Figures 3 and 4) more closely resemble the initial phases of self-organizing systems (Larsen-Freeman and Cameron 2008) and the initial development that preceded phase transitions in the writing of the two students examined in Baba and Nitta’s study (and that is characteristic of phase transitions in general—see Verspoor et al. 2008 and Verspoor and Smiskova 2012). This trend suggests that the present intervention may have produced the preamble to a phase transition (consistent with low-level stability defined by erratic performance and variability, as reported in Verspoor et al. (2008)) and that it stopped short of triggering the kind of transition characteristic of L2 appropriation. However, this should be interpreted with caution, given the difficulty associated with predicting phase transitions before they happen. Interpreting the data cautiously would also be consistent with the strict parameters CDST uses to characterize phase transitions (see Baba and Nitta 2014) as well as with (i) the CDST-informed studies just noted, (ii) the gradual rate of change implied in this study, and (iii) research documenting the need for extended periods of time (often years) to develop L2 academic writing proficiency (Leki 2007). In addition, unlike previous scholarship on voice that has focused on its long-term development (Ivanič 1997; Dressen-Hammouda 2014), these findings promote how studies on short-term development can expand present conceptualizations of voice construction. While the timeframe explored in this study does not capture the ongoing process of voice construction longitudinally, the findings here emphasize this process at the earliest stages of its ontogenesis, when the concept of authorial voice is first introduced to writers, and provide insight into initial conditions that contribute to the dynamic process of voice development. For example, this study highlights variegated learning trajectories that are typical of initial conditions of complex systems (Verspoor 2015) and the importance of repeated learning tasks (here, stylistic analyses) in shorter developmental periods to facilitate ongoing voice construction. Moreover, small-scale interactions related to Angela’s voice development—advice from an instructor outside the present study—were sufficient to alter her writing behavior. This instance emphasizes both the range of factors impacting learning systems as they co-adapt to their surroundings even at the microlevel and the need for researchers to attend to and account for unexpected influences on systems under study. Attention to smaller scale interactions, then, complements earlier descriptions of voice development that cast voice construction as primarily facilitated by large-scale interactions across the life span (e.g. the influence of the L1, enculturation, and personal beliefs about writing, developed over time) by underscoring smaller scale interactions that support its development. The present findings also contribute to reformulating how timescales are applied to track voice development—and more broadly, L2 development. In the CDST context (and on studies of L2 development in general—see, e.g. Leki et al. 2008), longitudinal research is repeatedly called for as a way forward (Larsen-Freeman 2011; Baba and Nitta 2014). However, as this study shows there is a reason for applying timescales more relativistically—in studies of authorial voice development and beyond—as no one timescale can represent ‘the scale for language development’ (emphasis in the original) (de Bot 2015: 31). Reconfiguring how timeframes are interpreted in developmental research is all the more important given, as we know, individual rates of language development can fluctuate for numerous reasons (e.g. target language, initial conditions, motivation, interactions). To develop research into micro timescales, continued studies may consider the benefit of microgenetic analyses. As this and other studies (Gánem-Gutiérrez 2008; Knouzi et al. 2010) have demonstrated, microgenesis is helpful for unpacking detailed moment-to-moment changes that may foreground phase transitions and for unpacking what is already known of macrolevel phenomena. Furthermore, continued research into voice construction on the micro timescale would eliminate the risk of conceptualizing voice construction across a single, longitudinal timescale, thus expanding understanding of its development. As this study shows, while the time period explored was insufficient for tracking interactions that resulted in full-scale phase transitions, the timeframe did reveal patterns of interaction that precede these transitions and the classroom dynamics that foster them. Microgenesis, then, is useful for exploring interactions across limited timeframes and can contribute to revealing unique levels of granularity across learning contexts and constructs. CONCLUSION This study explored microlevel cognitive processes that informed the earliest stages of authorial voice construction in one academic writing context. The findings revealed that all learners improved their conceptual understanding of voice, and that voice development was triggered by repeated tasks focused on stylistic analyses. Moreover, development was initially characterized as nonlinear and erratic with a period of stability emerging in the latter stages of the intervention, suggesting that interindividual variation decreased with time. Moreover, development unfolded at a rate that was isomorphic across learners. In revealing these processes, this work first furthers a CDST perspective on L2 writing development and demonstrates the value of applying timeframes in developmental research on a more relativistic scale, taking care to account for changes at the microlevel that can be captured via microgenesis. Second, this study complements present conceptualizations of voice construction by highlighting microlevel components of its development and underscoring the teachability of voice in classroom contexts. In this vein, this work also expands Matsuda’s (2015) review article on identity (and voice). Aptly absent therein is any discussion about cognitive processes that inform voice construction. As an addendum to Matsuda’s suggestions for future studies while simultaneously corroborating the present findings, further work is needed that explores how cognitive processes are constructed as these mechanisms interact with varying learning contexts. In this manner (and to borrow Matsuda’s metaphor) researchers move to equilibrium between social and cognitive approaches to conceptualizing L2 voice (and writing) development. This is especially relevant if it is accepted that the ‘dynamics of language learning are inextricably linked to the dynamics of consciousness, in neural activity and in the social world’ (Ellis 2008: 242)—a balance afforded to understanding change in L2 learning by a CDST approach to second language development. Reconceptualizing voice may thus lead to changes in how voice is described in the literature. For example, in Tardy’s (2012) work voice is expounded under the headings ‘Individual aspects of voice’ (35), characterized as a ‘writer’s unique and recognizable imprint’ (37), ‘Social aspects of voice’ (37), and ‘Voice as dialogic’ (39). Duly absent is a subheading that captures how learners’ cognitive behaviors interact to construct voice and that future research can work towards: Cognitive aspects of voice. Gary G. Fogal is an Assistant Professor in the Faculty of Liberal Arts, Sophia University, Japan. His primary research examines developmental processes informing L2 writing and draws from a dynamic systems theory account of additional language development. His work also explores the interface between emergentism and Vygotsky's sociocultural theory as well as the utility of literary texts for developing L2 proficiency. Address for correspondence: Gary G. Fogal, Faculty of Liberal Arts, Sophia University, Tokyo, Japan. <email@example.com> Acknowledgements Thanks are due to the journal editors and anonymous readers for valuable comments on this paper, and to Alister Cumming who provided valuable insights on multiple drafts of this work. This research was supported by the Social Sciences and Humanities Research Council of Canada, Grant No. 75220131291. SUPPLEMENTARY DATA Supplementary material is available at Applied Linguistics online. REFERENCES Baba K., Nitta R.. 2014. ‘ Phrase transitions in development of writing fluency from a complex dynamic systems perspective,’ Language Learning 64: 1– 35. Google Scholar CrossRef Search ADS Braun V., Clarke V.. 2006. ‘ Using thematic analysis in psychology,’ Qualitative Research in Psychology 3: 77– 101. Google Scholar CrossRef Search ADS Brooks L., Swain M., Lapkin S., Knouzi I.. 2010. ‘ Mediating between scientific and spontaneous concepts through languaging,’ Language Awareness 19: 89– 110. Google Scholar CrossRef Search ADS Bruner J. 1983. Child’s Talk: Learning to Use Language . Norton. Byrne D. 2002. Interpreting Quantitative Data . SAGE. Google Scholar CrossRef Search ADS de Bot K. 2015. ‘Rates of change: timescales in second language development’ in Dörnyei Z., MIntyre P. D., Henry A. (eds): Motivational Dynamics in Language Learning . Multilingual Matters, pp. 29– 37. Carter R. 1997. Investigating English Discourse: Language, Literacy and Literature . Longman. Carter R., Nash W.. 1990. Seeing Through Language: A Guide to Styles of English Writing . Basil Blackwell. Cilliers P. 2001. ‘ Boundaries, hierarchies, and networks in complex systems,’ International Journal of Innovation Management 5: 135– 47. Google Scholar CrossRef Search ADS Clark U., Zyngier S.. 2003. ‘ Towards a pedagogical stylistics,’ Language and Literature 12: 339– 51. Google Scholar CrossRef Search ADS Dressen-Hammouda D. 2014. ‘ Measuring the voice of disciplinarity in scientific writing: a longitudinal exploration of experienced writers in geology,’ English for Specific Purposes 34: 14– 25. Google Scholar CrossRef Search ADS Educational Testing Service. (n.d.). Writing Topics. Retrieved from http://www.ets.org/Media/Tests/TOEFL/pdf/989563wt.pdf. Eckstein G., Chariton J., McCollum R. M.. 2011. ‘ Multi-draft composing: an iterative model for academic argument writing,’ Journal of English for Academic Purposes 10: 162– 72. Google Scholar CrossRef Search ADS Ellis N. C. 2008. ‘ The dynamics of second language emergence: cycles of language use, language change, and language acquisition,’ The Modern Language Journal 92: 232– 49. Google Scholar CrossRef Search ADS Fischer K. W., Granott N.. 1995. ‘ Beyond one-dimensional change: parallel, concurrent, socially distributed processes in learning and development,’ Human Development 38: 302– 14. Google Scholar CrossRef Search ADS Fogal G. G. 2015. Pedagogical stylistics and concept-based instruction: An investigation into the development of voice in the academic writing of Japanese university students of English. Unpublished doctoral dissertation, University of Toronto, Toronto, Canada. Fogal G. 2017. ‘ Developing concept-based instruction, pedagogical content knowledge: Implications for teacher educators and L2 instructors,’ Language and Sociocultural Theory 4/1: 53– 75. Google Scholar CrossRef Search ADS Galperin P. I. 1969. ‘Stages in the development of mental acts’ in Cole M., Maltzman I. (eds): A Handbook of Contemporary Soviet Psychology . Basic Books, pp. 248– 73. Gánem-Gutiérrez A. 2008. ‘ Microgenesis, method and object: a study of collaborative activity in a Spanish as a foreign language classroom,’ Applied Linguistics 29: 120– 48. Google Scholar CrossRef Search ADS Gass S. M., Mackey A.. 2000. Stimulated Recall Methodology in Second Language Research . Lawrence Erlbaum. Hirvela A., Belcher D.. 2001. ‘ Coming back to voice: the multiple voices and identities of mature multilingual writers,’ Journal of Second Language Writing 10: 83– 106. Google Scholar CrossRef Search ADS Hiver P., Al-Hoorie A. H.. 2016. ‘ A dynamic ensemble for second language research: putting complexity theory into practice,’ The Modern Language Journal 100: 741– 56. Google Scholar CrossRef Search ADS Hulstijn J. H., Young R. F., Ortega L., Bigelow M., DeKeyser R., Ellis N. C., Lantolf J. P., Mackey A., Talmy S.. 2014. ‘ Bridging the gap: cognitive and social approaches to research in second language learning and teaching,’ Studies in Second Language Acquisition . 36: 361– 421. Google Scholar CrossRef Search ADS Hyland K. 2008. ‘ Disciplinary voices: interactions in research writing,’ English Text Construction 1: 5– 22. Google Scholar CrossRef Search ADS Ivanič R. 1997. Writing and Identity: The Discoursal Construction of Identity in Academic Writing . John Benjamins. Kim T.-Y., Baba K., Cumming A.. 2006. ‘Goals, motivations, and identities of three students writing in English’ in Cumming A. (ed.): Goals for Academic Writing: ESL Students and their Instructors . John Benjamins, pp. 125– 41. Google Scholar CrossRef Search ADS Knouzi I., Swain M., Lapkin S., Brooks L.. 2010. ‘ Self-scaffolding mediated by languaging: Microgenetic analysis of high and low performers,’ International Journal of Applied Linguistics 20: 23– 49. Google Scholar CrossRef Search ADS Larsen-Freeman D. 2006. ‘ The emergence of complexity, fluency and accuracy in the oral production and written production of five Chinese learners of English,’ Applied Linguistics 27/4: 590– 619. Google Scholar CrossRef Search ADS Larsen-Freeman D. 2011. ‘A complexity theory approach to second language development/acquisition’ in Atkinson D. (ed.): Alternative Approaches to Second Language Acquisition , Routledge. pp. 48– 72. Larsen-Freeman D., Cameron L.. 2008. Complex Systems and Applied Linguistics . Oxford University Press. Lavelli M., Pantoja A. P., Hsu H., Messinger D., Fogel A.. 2005. ‘Using microgenetic designs to study change processes’ in Teti D. M. (ed.): Handbook of Research Methods in Developmental Science . Wiley-Blackwell. pp. 40– 65. Google Scholar CrossRef Search ADS Leki I. 2007. Undergraduates in a Second Language. Challenges and Complexities of Academic Literacy Development . Lawrence Erlbaum. Leki I., Cumming A., Silva T.. 2008. A Synthesis of Research on Second Language Writing in English . Routledge. Lin H. 2010. ‘The taming of the immeasurable: An empirical assessment of language awareness’ in Paran A., Sercu L. (eds.): Testing the Untestable in Language Education . Multilingual Matters, pp. 191– 216. Luhmann N. 1989. Ecological Communication . University of Chicago Press. Matsuda P. K. 2015. ‘ Identity in written discourse,’ Annual Review of Applied Linguistics . 35: 140– 59. Google Scholar CrossRef Search ADS Poehner M. E. 2008. Dynamic assessment: a Vygotskian approach to understanding and promoting L2 development . Springer. Siegler R. S. 2006. ‘Microgenetic analyses of learning’ in Kuhn D., Siegler R. S. (eds): Handbook of Child Psychology, Volume 2: Cognition, Perception, and Language (6th edn.) . Wiley, pp. 464– 510. Stock I., Eik-Nes N. L.. 2016. ‘ Voice features in academic texts: a review of empirical studies,’ Journal of English for Academic Purposes 24: 89– 99. Google Scholar CrossRef Search ADS Swain M. 2006. ‘Languaging, agency and collaboration in advanced language proficiency’ in Byrnes H. (ed.): Advanced Language Learning: The Contribution of Halliday and Vygotsky . Continuum, pp. 95– 108. Spoelman M., Verspoor M.. 2010. ‘ Dynamic patterns in development of accuracy and complexity: A longitudinal case study in the acquisition of Finnish,’ Applied Linguistics 31: 532– 53. Google Scholar CrossRef Search ADS Tardy C. M. 2012. ‘Current conceptions of voice’ in Hyland K., Sancho Guinda C. (eds.): Stance and Voice in Written Academic Genres . Palgrave Macmillan, pp. 34– 48. Google Scholar CrossRef Search ADS Tardy C. M. 2016. ‘Voice and identity’ in Manchón R. M., Matsuda P. K. (eds.): Handbook of Second and Foreign Language Writing . De Gruyter, pp. 349– 64. Thelen E, Smith L. B.. 1994. A Dynamic Systems Approach to the Development of Cognition and Action . MIT Press. van Dijk M., van Geert P.. 2007. ‘ Wobbles, humps and sudden jumps: A case study of continuity, discontinuity and variability in early language development,’ Infant and Child Development 16: 7– 33. Google Scholar CrossRef Search ADS van Dijk M., Verspoor M., Lowie W.. 2011. ‘Variability and DST’ in Verspoor M. H., de Bot K., Lowie W. (eds): A Dynamic Approach to Second Language Development . John Benjamins, pp. 55– 84. Google Scholar CrossRef Search ADS Vandrick S. 2003. ‘Literature in the teaching of second language composition’ in Kroll B. (ed.): Exploring the Dynamics of Second Language Writing . Cambridge University Press, pp. 263– 83. Google Scholar CrossRef Search ADS van Geert P., van Dijk M.. 2002. ‘ Focus on variability: New tools to study intra-individual variability in developmental data,’ Infant Behavior & Development 25: 340– 74. Google Scholar CrossRef Search ADS Verdonk P. 2013. The Stylistics of Poetry: Context, Cognition, Discourse, History . Bloomsbury. Verspoor M. 2015. ‘Initial conditions’ in Dörnyei Z., MacIntyre P. D., Henry A. (eds): Motivational Dynamics in Language Learning . Multilingual Matters, pp. 38– 46. Verspoor M., Lowie W., van Dijk M.. 2008. ‘ Variability in second language development from a dynamic systems perspective,’ Modern Language Journal 92: 214– 31. Google Scholar CrossRef Search ADS Verspoor M., de Bot K., Lowie W.. (eds.) 2011. ‘ A dynamic systems approach to second language development,’ John Benjamins. . Verspoor M., Schmid M., Xu X.. 2012. ‘ A dynamic usage based perspective on L2 writing,’ Journal of Second Language Writing 21: 239– 63. Google Scholar CrossRef Search ADS Verspoor M., Smiskova H.. 2012. ‘Foreign language writing development from a dynamic usage based perspective’ in Manchón R. M. (ed.): L2 Writing Development: Multiple Perspectives. De Gruyter, pp. 17– 46. Vygotsky L. S. 1986. Thought and Language (Revised edn). MIT Press. Widdowson H. G. 1975. Stylistics and the Teaching of Literature . Longman. Zhao C. G. 2013. ‘ Measuring authorial voice strength in L2 argumentative writing: the development and validation of an analytic rubric,’ Language Testing 30: 201– 30. Google Scholar CrossRef Search ADS © Oxford University Press 2017
Applied Linguistics – Oxford University Press
Published: Oct 18, 2017
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera