2. Brief overview of SCT and AT

SCT, associated with the work of Vygotsky (Reference Vygotsky1978, Reference Vygotsky and Wertsch1981), views human cognitive development as occurring in mediated social interaction. In recent years, scholars such as Lantolf and Swain have shown how the theory can be applied to research in L2 teaching and learning (e.g. Lantolf Reference Lantolf and Lantolf2000; Swain Reference Swain and Lantolf2000; Lantolf & Thorne Reference Lantolf and Thorne2006; Swain Reference Swain and Byrnes2006; Swain, Kinnear & Steinman Reference Swain, Kinnear and Steinman2015). The key and interrelated constructs in SCT and AT that are of pertinence to a discussion of WCF are: (1) ZPD and scaffolding, (2) mediating tools, and (3) the notion of an activity.

From an SCT perspective, cognitive development takes place in social interaction, whereby an expert member of the society (e.g. parent, teacher, more knowledgeable peer) provides appropriate forms of assistance to a novice. However, not all forms of assistance are helpful and lead to development. According to SCT, for development to take place, the assistance needs to take into consideration the novice's current and potential levels of competence, the so-called Zone of Proximal Development. The potential level of competence is assessed in terms of the novice's ability to take advantage of the assistance offered. In this sense, the assistance needs to be dynamic, responsive to the novice's evolving needs rather than being static and predetermined. Lantolf & Aljaafreh (Reference Lantolf and Aljaafreh1995: 62) refer to assistance within the ZPD as ‘an act of negotiated discovery’. The two key traits of effective assistance are that it is graduated and contingent. Graduated assistance is based on the assumption that too much assistance may be as detrimental as too little assistance because it could discourage the novice from becoming independent. Contingent assistance implies that the feedback provided needs to be responsive to the novice's performance, and ultimately removed when the novice displays an ability to function independently.

The metaphor that is commonly used to describe this negotiated and dynamic assistance is scaffolding (Wood, Bruner & Ross Reference Wood, Bruner and Ross1976). From this perspective, WCF is seen as a form of assistance and any discussion of whether the WCF provided to the learner is effective needs to consider whether it represents scaffolded assistance within the learner's ZPD.

A related point to a discussion of effective WCF within the ZPD is what constitutes evidence of language development. Research on WCF informed by cognitive perspectives has tended to look at improved accuracy on new pieces of writing or in the use of certain linguistic structures targeted by the feedback (i.e. a focus on the product) as evidence of language development. From an SCT perspective, evidence of development also needs to take into consideration any change in the nature of the assistance the learner requires. Thus, if over time the learner requires less direct feedback (e.g. signalling an error without specifying its exact location) in order to self-correct, this is also taken as indicative of L2 development. Viewing the nature of assistance and responsiveness to assistance as evidence of development is discussed in the literature on dynamic assessment (see Poehner Reference Poehner2009; Poehner & Lantolf Reference Poehner and Lantolf2013; Lantolf & Poehner Reference Lantolf and Poehner2014). In dynamic assessment, the learner's overall score is made up of a score on task performance (current competence) and a score on the assistance required to complete the task (potential competence).

Another key construct in SCT is mediating tools. Vygotsky's fundamental claim is that human cognitive development is mediated by culturally constructed means or tools. These tools can be material (e.g. abacus, computer) or symbolic (e.g. language, gestures). The tools facilitate and shape actions, including the regulation of our thinking processes. For example, language, a symbolic tool, enables interaction between the expert and the novice. These tools evolve over time and new tools may impact on human action and ultimately on cognitive development. In the case of WCF, tools can impact on the provision of feedback and on the learners’ processing of the feedback. For example, when WCF is delivered via track changes and marginal comments we may see an effect on the quantity and the nature of the feedback comments the teacher provides and also on how learners engage, process and appropriate the feedback.

The third key construct in discussion of WCF from an SCT perspective is that of activity. The notion of activity was originally discussed by Vygotsky (Reference Vygotsky1978), but was developed into a fully fledged theory, referred to as activity theory, by Leont'ev (Reference Leont'ev1978, Reference Leont'ev and Wertsch1981). The theory was further revised and represented as complex models by Engeström (Reference Engeström1987, Reference Engeström2001). These models can be used to analyse and explain observable human behaviour. The underlying premise of AT is that in order to gain a better understanding of human behaviour we need to investigate the goals that drive human actions as well as the context (referred to as community) in which the activity takes place, rules of appropriate behaviour, and hierarchies (referred to as division of labour) in that context. For example, AT can be used to explain teacher feedback practices by considering the teachers’ goals, institutional policies, and the tools used to deliver the feedback. Similarly, AT can be used to analyse students’ responses to feedback as an activity, taking into consideration the learners’ own goals, their relationship with the teacher and expected norms of behaviour.

In what follows I discuss the three areas requiring further investigation and I link each of these areas to the theoretical constructs in SCT or AT. These three areas are:

1. 1. What constitutes effective WCF from an SCT perspective (Section 3)?

2. 2. What is the impact of new mediating tools on the provision of and response to WCF (Section 4)?

3. 3. How do individual and context-related factors impact on WCF as an activity (Section 5)?

In my discussion I suggest a number of research tasks, the kind of data that need to be collected and describe qualitative approaches to data analysis that are in line with these theoretical perspectives. I should also note here, and as discussed in my concluding section, that although these three areas are presented as separate research areas, a more productive and theoretically sound approach to investigating WCF would in fact be to include an examination of individual and context-related dimensions in the first two research areas.

3. Investigating what constitutes effective WCF from an SCT perspective

Research into the most effective form of WCF is important for teachers who spend an inordinate amount of time on providing such feedback and for learners who wish to improve their grammatical accuracy. Most of the research on identifying effective WCF has focused on comparing the benefits of direct versus indirect WCF. It should be noted that the distinction between these forms of WCF is not always clear-cut. Nonetheless, these two approaches to providing WCF have very different underlying assumptions. The provision of direct WCF is based on the assumption that the learner is incapable of self-correcting. Indirect feedback, on the other hand, is based on the assumption that the learner is capable of self-correcting in response to some sort of signal. To date the findings of studies investigating whether direct or indirect WCF is best have been mixed, and could perhaps be attributable to different research design features. For example, in a classroom-based study by Chandler (Reference Chandler2003) where students received direct and indirect feedback on consecutive pieces of writing, no significant differences between the two types of WCF were found, whereas the study by Van Beuningen, De Jong & Kuiken (Reference Van Beuningen, De Jong and Kuiken2008, Reference Van Beuningen, De Jong and Kuiken2012), a carefully designed experimental study (with control and treatment groups), found that direct WCF was superior to indirect WCF. Unlike these two studies where feedback was provided on a range of errors, in the study by Bitchener, Young & Cameron (Reference Bitchener, Young and Cameron2005), feedback was given only on errors in the use of three structures, and the researchers found that the efficacy of the type of WCF depended on the grammatical structure corrected.

SCT presents a different way of viewing direct and indirect WCF. Rather than viewing them as dichotomous alternatives, direct and indirect WCF can be viewed as two ends of a continuum of scaffolded feedback. Scaffolded feedback needs to be graduated and contingent; that is, it needs to be attuned and responsive to the evolving linguistic needs and abilities of the learner, with the ultimate aim of handing over the responsibility for correction to the learner (van Lier Reference van Lier2004). Applying this metaphor to the provision of WCF means that, for example, initial errors in the use of particular structures (e.g. frequent errors) should receive indirect feedback (e.g. underline the errors and denote the type of errors using codes). If in a redraft or in a one-on-one conference, the learner demonstrates ability to self-correct in response to the indirect WCF (e.g. ability to correct some errors in the use of these structures), any subsequent feedback provided on these errors (in subsequent drafts or new writing) should, from this theoretical perspective, become increasingly more indirect (e.g. the errors are just underlined and not coded or a brief end comment is made about persistent errors). However, if the learner shows inability to take advantage of the indirect feedback provided, the feedback on subsequent drafts should be more direct. A learner's increasing ability to self-correct and to generalize the knowledge gained to new texts are evidence of the learner's language development.

There is now a general consensus among scholars in the field of L2 learning that in order to gauge the learner's evolving ability, studies need to be longitudinal. Language development takes time and may not proceed in a linear fashion (see for example Larsen-Freeman Reference Larsen-Freeman2006). Thus the research question that should guide research on WCF is not which form of WCF is more effective (direct or indirect) for learners as a group but whether the form of WCF provided was appropriate to specific learners at their specific developmental point (ZPD). As Aljaafreh & Lantolf (Reference Aljaafreh and Lantolf1994: 480) point out, ‘all types of feedback are potentially relevant for learning, but their relevance depends on where in the learner's ZPD a particular property of the L2 is situated’.

To date, only a small number of studies have applied the metaphors of scaffolding and ZPD when investigating corrective feedback on learners’ writing (e.g. Aljaafreh & Lantolf Reference Aljaafreh and Lantolf1994; Nassaji & Swain Reference Nassaji and Swain2000; Erlam, Ellis & Batstone Reference Erlam, Ellis and Batstone2013). These studies employed a regulatory scale of assistance developed by Aljaafreh & Lantolf (Reference Aljaafreh and Lantolf1994) which categorizes feedback from the least to the most direct form of feedback. However, it is important to note that in these studies the feedback was provided in oral form, in one-on-one conferences. The conferences began with the tutor offering the least direct form of feedback (e.g. asking the learner to self-identify their errors), and increasing the explicitness and specificity of the feedback depending on the learner's response. Nassaji & Swain (Reference Nassaji and Swain2000) reported that scaffolded feedback was more effective than non-scaffolded feedback (i.e. random feedback that was not graduated and contingently responsive).

However, in most L2 writing contexts, feedback is more likely to be provided as written comments on students’ drafts rather than orally in intensive one-on-one conferences. The study by Morton, Thompson & Storch (Reference Morton, Thompson and Storch2014) is one of the few studies to have employed the metaphor of scaffolding to examine feedback provided in written form. This longitudinal study, conducted over a five-week period, investigated retrospectively the amount and the nature of the feedback comments provided by a supervisor (Storch) to her MA student on three successive drafts of the student's literature review chapter. What further distinguished this study is that it applied the metaphor of scaffolding not only to the nature of the WCF but all the feedback comments, including feedback on content and organization, and that it took the rhetorical forms of these comments (e.g. directives, suggestions) into consideration. The study found that, although the supervisor provided feedback on all aspects of writing, a large volume of the feedback was WCF and that there was relatively little change in the degree of directness of the feedback given over time. However, since the study only considered the nature of the feedback and did not analyse the learner's response to the feedback, it is difficult to ascertain whether the learner's ability to take advantage of the feedback provided, including WCF, changed over the time and therefore whether the feedback was indeed contingently responsive to the learner's needs. It also did not take into account the learner's goals and whether she attributed much importance to improved accuracy (see Hyland Reference Hyland1998).

In what follows, I outline three related research tasks. All three employ scaffolding as an analytical framework to examine the WCF activity. The first task builds on the small-scale study conducted by Morton et al. (Reference Morton, Thompson and Storch2014). It proposes to examine the nature of WCF teachers provide to learners on their writing over time. The second task addresses the shortcoming identified in the Morton et al. study. It sets out to investigate whether the WCF provided by the teacher (gathered in Research task 1) aligns with the learners’ needs as evident in their successive drafts and new texts. The third research task measures learners’ L2 development over time, using the type of scaffolding the learners required as one measure of development.

All three research tasks are longitudinal and conducted in-situ, that is, in authentic contexts where WCF is provided as a regular part of the pedagogical programme. These research tasks also assume that the learners produce extended writing and are thus best implemented with senior high school or college/university students. Given the detailed data analysis they require, a case study approach may be appropriate for some of these research tasks.

This study can be undertaken with L2 learners in a university or college writing programme or with graduate students submitting drafts of their thesis chapters to their supervisors. The main requirement for this longitudinal study is that students submit many pieces of writing on which they receive feedback, including multiple drafts of the same text. The teacher's feedback on students’ successive drafts and new texts is analysed for volume (i.e. number of comments), focus (what items of language are corrected), and most importantly, the nature of the correction (direct versus indirect). The use of successive drafts could make the task of tracing teacher feedback on the same kind of errors easier (as well as in tracing learners’ responses to the feedback). In coding for the nature of the correction, the researcher needs to establish a scale, adapting perhaps the scale developed by Aljaafreh & Lantolf (Reference Aljaafreh and Lantolf1994) for oral one-on-one conferences, to capture the directness/indirectness of the written feedback provided. As mentioned above, in the Morton et al. Reference Morton, Thompson and Storch2014 study, the scale of directness included not only the nature of the feedback (e.g. reformulations, deletions) but also the rhetorical forms of feedback comments (e.g. suggestions, questions, directives). The findings would reveal whether the WCF changed over time: has the amount of feedback changed? Has the focus (what is targeted by the feedback) changed? Has the nature of the feedback altered (e.g. become less direct)? The findings of these retrospective studies can help raise the consciousness of WCF providers (teachers, supervisors) about their existing WCF practices and whether these practices accord with their pedagogical beliefs and intentions (see also studies framed by AT which consider individual and context-related factors in Research tasks 6, 7.

If we view effective WCF as scaffolded assistance, the type of WCF provided needs to be calibrated and responsive to the learner's needs and expertise. Successive drafts of the same text and new texts can provide evidence of whether these needs are changing in line with developing expertise in the use of some linguistic structures (see Aljaafreh & Lantolf Reference Aljaafreh and Lantolf1994). This research task thus requires a simultaneous examination of the learners’ writing and the WCF provided to answer questions such as: is there enough or too much feedback provided? Is the feedback direct when indirect feedback would have been sufficient or vice versa, is the feedback indirect when direct feedback was needed?

To respond to these questions, a case study or multiple case studies approach is recommended. The case study learner's successive drafts and new texts are analysed for the number and type of errors, whether these are new or repeated errors on which the learner had received feedback in previous drafts (evident from findings from Research task 1), and if so, the nature of the feedback provided on previous drafts on these errors (i.e. in terms of directness as in the Morton et al. Reference Morton, Thompson and Storch2014 study). In sum, the analysis in this research task considers errors evident in the learner's texts and the nature of the feedback provided on these errors in successive drafts. Thus to make the task more manageable, it may be best to limit the analysis to the most frequent errors.

The findings of such a study can be very informative for teachers, both novice and experienced teachers. They can be used to determine whether the teacher's WCF accords with the notion of scaffolding within the ZPD. If, for example, the study finds that the case study learner is progressively making fewer errors in the use of a particular structure, but is still receiving direct WCF on those errors, such findings would suggest that the feedback is not contingently responsive to the learner's evolving L2 expertise. Similarly, if the study finds that indirect WCF is provided despite the persistence of the errors whereas direct forms of feedback (including perhaps an explanation or sample sentences including the correct use of the structure) would have been more suitable, such findings would indicate that the feedback provided does not represent effective scaffolded assistance.

Research task 3 uses the same data as that used in Research task 2 but the focus here is on assessing the learner's L2 developmental trajectory. As noted previously, from an SCT perspective, development is evident not only in improved accuracy scores in the use of particular structures on new pieces of writing (performance) but also in terms of the kind of feedback the learner requires.

Thus, in this research task, the learner's L2 development is gauged using two measures. The first measure is based on an accuracy score in the use of specific structures on a new writing task and a general accuracy score. The second measure takes into consideration the type of feedback the learner requires on successive writing tasks (revised drafts and new texts). Together, the two measures provide a more comprehensive assessment of the learner's developmental trajectory. Yet it is important to note that the use of these measures poses a number of challenges.

Measuring accuracy is notoriously difficult (see Polio Reference Polio1997; Pallotti Reference Pallotti2009), with limitations inherent in the use of any measure of accuracy. The measures that can be used to gauge accuracy is the percentage of target-like use of specific structures as well as more general accuracy scores (e.g. total errors per words, proportion of error free clauses). In addition to choosing suitable measures of accuracy, the researcher will also need to decide and be consistent in dealing with repeated errors of the same kind or partial correct use of structures (e.g. correct choice of verb aspect but an incorrect auxiliary).

Tracing a learner's response to feedback to determine whether it constitutes signs of evolving expertise is a labour intensive exercise. It is further complicated by the underlying assumption that the teacher's feedback on successive texts has been responsive to the learner's needs. One way of addressing these challenges is to use a computer-mediated form of WCF. Based on the work of Poehner & Lantolf (Reference Poehner and Lantolf2013), a predetermined scale ranging from indirect to direct WCF on the use of a range of linguistic structures can be provided in a computerized format. The computerized system has a number of advantages: the level of assistance provided is responsive to the learner's need, the system can deal with a number of targeted structures, and it can be offered to large cohorts of students. These features of computer-mediated forms of WCF address some of the challenges mentioned above. However, using the computer to deliver WCF may have other, less desirable effects on learner engagement with WCF, and these are discussed in the next research issue.

4. The impact of new mediating tools on the provision of and response to WCF

From an SCT perspective, mediating tools play an important role in human activity. Tools not only enable and enhance human actions (e.g. calculations, composing texts) but also shape actions (e.g. how we perform mental calculations now that we use calculators or compose texts now that we use computers rather than pen and paper). SCT draws a distinction between psychological tools (e.g. language) and material tools (e.g. calculators, computers). In this section I focus on material tools and specifically the use of computers to deliver WCF.

In recent years we have witnessed rapid developments in the range of tools available to provide feedback, such as audio annotations, computer-mediated means of providing WCF (e.g. insert pre-populated comments, track changes) as well as more sophisticated programs that can generate automated scoring and feedback and which draw on research in the field of Natural Language Processing (Heift & Schulze Reference Heift and Schulze2007). According to Stevenson & Phakiti (Reference Stevenson and Phakiti2014), commercially available automated scoring and feedback programs (e.g. Criterion), which were previously used mainly in large-scale proficiency tests, are increasingly being used in writing classrooms in the USA.

As in the case of all new tools, computer-mediated means of providing WCF are likely to impact on the activity of providing and responding to feedback. For example, peer feedback studies suggest that when the peer feedback is delivered via the computer it may increase the amount of the feedback (e.g. Liu & Sadler Reference Liu and Sadler2003) but may discourage substantive engagement with the feedback comments by the recipients (e.g. Guardado & Shi Reference Guardado and Shi2007; Chang Reference Chang2012). Research on the impact of computer-delivered teacher feedback on the quantity and quality of WCF is surprisingly scarce, with key details such as the amount of feedback provided and the nature of the feedback often undisclosed.

Stevenson & Phakiti (Reference Stevenson and Phakiti2014) conducted a critical review of studies investigating the impact of automated scoring and feedback programs on L2 development and concluded that as yet it is difficult to determine whether automated feedback is effective because of the limited number of studies, the different writing tasks used and learners included in these studies, as well as the flaws in the research design of these studies. Most of these studies seem to focus on the product: on how many redrafts the students submit and the accuracy scores of these redrafts. There has been little attention paid to the process of revision, that is, to how learners process and respond to the feedback provided, or to the impact of this feedback on subsequent writing. There are some indications though that automated feedback may make learners more reliant on the feedback provided, decreasing self-initiated corrections (Warden 2000, cited in Stevenson & Phakiti Reference Stevenson and Phakiti2014).

Clearly what we need is well-designed studies on the range of tools available to provide WCF, including automated feedback, which examine the impact of these tools on processes: on the nature of the feedback provided, on how learners engage with that feedback, and ultimately on L2 learning outcomes.

There are a number of ways to investigate the impact of new tools for delivering WCF on the nature of the feedback provided. One way is to compare the impact that different modalities of delivering WCF have on the teacher's WCF. These modalities could include feedback delivered in the form of digital audio files, screen-capture technology with voice-overs, or what may be considered nowadays as more traditional forms of computer-mediated WCF, such as using the feedback functionality available in online systems such as learning management systems (LMS), which are used in many universities. LMS has a number of facilities, including the ability to provide written or audio feedback. The other advantage of LMS is that the teacher has an archival file of drafts and feedback for each student (accessible to the students for the duration of the teaching term). When providing written feedback, the teacher can also create and then draw on a bank of comments and symbols. The system can also display a summary of the number of written feedback comments made on each assignment.

In this study, feedback provided as written commentary and orally via digital audio files (both delivered via LMS) would constitute the main sources of data. This data should be supplemented by in-depth teacher interviews. A comparison between the feedback provided in the two modes (written versus audio) would help determine the impact of the tool on the WCF provided: whether one means of delivery increases the quantity of the comments, whether it skews the feedback to particular structures, or whether it affects the nature of the comments – making them perhaps more impersonal. Teachers’ views about the relative ease or difficulty of providing feedback in the two modes are also important to take into consideration as they could help explain the findings (see also discussion on Research tasks 6, 7).

The analysis adopted in this study could be similar to that suggested in Research task 2, to help determine whether some tools for providing WCF encourage teachers to provide WCF which accords with the traits of scaffolding.

Research task 5 considers the students’ actions, that is, their engagement and response to two different modes of feedback provision (in Research task 4). Data collection tools could include think-aloud protocols, retrospective interviews, key stroke analysis and screen capture (depending on the feedback tool used). It is important, however, to be cognizant of the fact that some of these data collection tools could have an impact on the activity itself (see, for example, research on the reactivity of think-aloud protocols in Sachs & Polio Reference Sachs and Polio2007). One way of dealing with this design issue is to use a range of data collection tools and triangulating the data.

Data analysis would include measuring the time spent on considering each feedback point (using think-aloud data and key stroke analysis), the response to the feedback (accept/reject/ignore) and ultimately whether new pieces of writing show evidence of language learning (ability and/or inclination to use the structures on which the students received WCF correctly in new texts). If think-aloud protocols are used in both modes, then a qualitative analysis of the nature of deliberations about the feedback provided could also be undertaken. Retrospective interviews with students may reveal the students’ perceptions of these new tools and what beliefs, goals, and preferences drive their purposeful behaviour (see also Research tasks 6, 7) and again help explain findings on how learners engage with the feedback delivered in the two modes and whether the mode impacts on learners’ response to the feedback provided. These are important considerations. Learners’ response to the feedback may explain why WCF may or may not lead to L2 development. A small number of case studies have shown that the depth of processing of feedback provided (e.g. Qi & Lapkin Reference Qi and Lapkin2001) and learners’ affective response to the feedback (Storch & Wigglesworth Reference Storch and Wigglesworth2010a) may impact on whether the information supplied by the feedback is retained.

5. Individual and context-related dimensions of the WCF activity: An AT perspective

One of the criticisms levelled at studies on WCF is that they are often decontextualized, presenting the feedback activity as linear: the teacher provides feedback and the learner revises (Goldstein Reference Goldstein, Silva and Matsuda2001, Reference Goldstein, Hyland and Hyland2006). Details provided about the learners are often limited to age, gender, L1 and L2 proficiency. No details are given about the learners’ or teachers’ goals, the relationship between the learner and the teacher, the classroom context, or the broader context. This represents a limited view of WCF as an activity: of feedback provision and response to feedback. As Hyland & Hyland (Reference Hyland and Hyland2006: 86) correctly point out, ‘feedback is not simply a disembodied reference to student text but an interactive part of the whole context of learning’. Individual and context-specific dimensions, in addition to the nature of the WCF, will enable us to extrapolate research findings from one context to other L2 learning contexts. More importantly, given rapid developments in digital technology, such studies will provide us with a deeper understanding of learner response to WCF when delivered by different tools.

If we consider WCF from an AT perspective, we view it as a human, situated activity. We view all participants in the activity as active agents who come into the activity with their own history, beliefs and goals, and whose actions are driven by their desire to achieve these goals. Furthermore, in a situated activity the participants’ actions will also be shaped by cultural and institutional norms and expectations, as well as by the relationships humans form with each other.

The second generation model of AT, proposed by Engeström (Reference Engeström1987, Reference Engeström2001), provides us with a theoretical heuristic that can be used to map the information we gather about a WCF activity. The model enables researchers to identify the key elements in an activity and the relationships among these interacting elements in a systematic and coherent manner. These key elements are: the participants whose actions are driven by their goals, the tools used to complete the activity, and the context in which the activity takes place (the community with its norms and division of labour). Furthermore, once we populate this model with information from the activity we investigate (e.g. see Lee Reference Lee2014), we can not only gain a broader perspective of the activity but also identify the source of any problems and make appropriate recommendations, including the kind of innovations that could be considered (see Thorne Reference Thorne, St John, van Esch and Schalkwijk2004).

The following research task suggests how AT can be employed to investigate WCF as an activity. It is a replication of the Lee (Reference Lee2014) study that used AT to interpret the findings from her earlier research about the nature of teacher feedback in her particular context (a Hong Kong secondary school) and then to make appropriate recommendations. The proposed research task diverges from Lee's study in that AT informs its design from the outset, and it consists of two parts: the first investigates teachers’ beliefs and WCF practices, the second students’ beliefs and response to the WCF. Although the two parts may appear as two separate studies, it is best to view them as one research task, albeit a complex one, reflecting the complexity of WCF as an activity. The proposed research task is exploratory, with data collected from a range of sources.

To date, studies on teachers’ WCF practices have relied mainly on surveys (e.g. Evans, Hartshorn & Tuioti Reference Evans, Hartshorn and Tuioti2010) and interviews, at times supplemented by students’ texts containing the feedback comments (e.g. Al Shahrani & Storch Reference Al Shahrani and Storch2014). In Research task 6, I propose that data should include transcripts of in-depth interviews with teachers (and/or supervisors) as well as administrators, classroom observations, policy documents (government, institutional) and a sample of student texts containing the teachers’ feedback comments. The amount of data and the qualitative analysis required means that a single or multiple case study approach is the most suitable.

In this project, teacher interviews could elicit the goals that drive teachers’ feedback action (e.g. whether it is to help their students achieve higher accuracy), their beliefs about WCF (e.g. whether it is beneficial), the factors that shaped these beliefs (e.g. their own L2 learning experiences, training received), as well as the kind of feedback they think they provide their students and why (e.g. students’/parents’ expectations, workloads, tools used). An analysis of the feedback teachers provide is an important means to triangulate the interview data. Research has shown that teachers (and supervisors) may be unaware of the amount and kind of feedback they provide (e.g. Al Shahrani & Storch Reference Al Shahrani and Storch2014; Basturkmen, East & Bitchener Reference Basturkmen, East and Bitchener2014; Morton et al. Reference Morton, Thompson and Storch2014) and that their perceptions and beliefs may not match their feedback practices.

Interviews with administrators and policy documents provide important sources of information about overt policies or implied expectations about appropriate forms of WCF. Institutional views concerning good teacher practices (see Sullivan Reference Sullivan and Lantolf2000) and the value attributed to accuracy (see Lee Reference Lee2014) may influence teacher WCF practices as much as mandated WCF policies (see Al Shahrani & Storch Reference Al Shahrani and Storch2014). Classroom observations would enable the researcher to establish not only the type of information explicitly and implicitly communicated to the students about what is important in L2 writing but also the kind of relationships the teacher builds with her students. The findings of such a study could explain, for example, why teachers may provide or fail to provide scaffolded feedback (cf. Research task 1) as well as how and why new feedback provision tools impact on the feedback provided (cf. Research task 4).

The second part of this project, Research task 7, focuses on learners’ actions, and perceptions, goals and expectations that drive their actions. The assumption underlying this research task is that learners, and particularly adult learners, exercise some volitional control over what they notice in any feedback they receive on their writing and how they respond to the feedback (see Hyland Reference Hyland1998; Storch & Wigglesworth Reference Storch and Wigglesworth2010a, Reference Storch, Wigglesworth and Batstoneb; Han & Hyland Reference Han and Hyland2015).

Data collection tools include the learners’ revised writing in response to WCF provided on earlier drafts, in-depth retrospective interviews with the learners, and observations of classroom behaviour. Analysis of the revised drafts would yield findings about how learners respond to the feedback received (e.g. how each draft is revised and what proportion of the feedback comments is accepted, whether other unsolicited revisions are made). The interviews aim to elicit the learners’ goals, including the importance they attribute to accuracy (over other aspects of writing), their beliefs (e.g. about how best to improve their writing), and the factors that may have shaped their beliefs (e.g. previous L2 learning experience, the views of significant others). The interviews should also probe about the relationship they establish with the feedback provider (see Lee & Schallert Reference Lee and Schallert2008). Interview data is supplemented by classroom observations. Lee's (Reference Lee2008) study, for example, showed that a teacher's teaching style played an important role in explaining students’ response to WCF provided. Broader cultural factors also need to be considered (e.g. expectations regarding student behaviour) as they too may explain certain response behaviour.

Together, the two parts of this larger research project (Research tasks 6 & 7) constitute a systematic analysis of the WCF activity that can demonstrate how individual variables, context-related variables, and the interaction of these two sets of variables, can explain the outcome of this activity. It is important to conduct such studies longitudinally, because relationships between the participants take time to establish and may change over time. A multiple case study design could yield rich data, providing a comprehensive picture of the WCF activity in a particular context.